Merge branch 'feature/question-strategy-alignment-v0.6'

This commit is contained in:
2026-08-03 07:38:50 +01:00
31 changed files with 7406 additions and 213 deletions
+8
View File
@@ -80,6 +80,14 @@ export default function DiagnosticsView({ result }) {
? `${validationIcons.valid} valid`
: `${validationIcons.invalid} invalid`,
},
{
label: "Investigation strategy",
value:
diagnostics.investigationStrategy?.key ||
diagnostics.investigationStrategy ||
result.selectedQuestion?.strategy ||
"?",
},
];
const errors = [
+44
View File
@@ -28,6 +28,8 @@ export default function GraphUpdateView({ updateResult }) {
proposal,
previousSituationGraph,
updatedSituationGraph,
reasoningState,
previousReasoningState,
} = updateResult;
const newlySurfacedUnknownNodeIds = (proposal.addedNodes || [])
@@ -64,6 +66,11 @@ export default function GraphUpdateView({ updateResult }) {
<div className="text-xs text-gray-600">
{node.kind} · {node.confidence}
</div>
{node.confidenceAssessment && (
<div className="text-xs text-gray-600">
evidence {node.confidenceAssessment.evidenceConfidence} · completeness {node.confidenceAssessment.completenessStatus} · conclusion {node.confidenceAssessment.conclusionConfidence}
</div>
)}
{(update?.previousStatus || update?.newStatus || node.status) && (
<div className="text-xs text-gray-700">
{update?.previousStatus ? `Previous status: ${update.previousStatus}` : null}
@@ -109,6 +116,23 @@ export default function GraphUpdateView({ updateResult }) {
: null,
].filter(Boolean);
const previousComparabilityStatus =
previousReasoningState?.comparabilityStatus ||
previousSituationGraph?.reasoningState?.comparabilityStatus ||
null;
const newComparabilityStatus =
reasoningState?.comparabilityStatus ||
updatedSituationGraph?.reasoningState?.comparabilityStatus ||
null;
const relationshipStatus =
reasoningState?.relationshipStatus ||
updatedSituationGraph?.reasoningState?.relationshipStatus ||
null;
const reasoningStagesAfter =
reasoningState?.reasoningStages ||
updatedSituationGraph?.reasoningState?.reasoningStages ||
[];
return (
<div className="space-y-4">
<section className="rounded-lg border border-blue-200 bg-blue-50 p-4">
@@ -134,12 +158,32 @@ export default function GraphUpdateView({ updateResult }) {
{selectedQuestion.question}
</div>
)}
{previousComparabilityStatus && newComparabilityStatus && (
<div>
<span className="font-medium">Comparability:</span>{" "}
{previousComparabilityStatus} {newComparabilityStatus}
</div>
)}
{relationshipStatus && (
<div>
<span className="font-medium">Relationship status:</span>{" "}
{relationshipStatus}
</div>
)}
{!selectedQuestion?.question && !newActiveUnknownNodeId && previousActiveUnknownNodeId && (
<div>
<span className="font-medium">Next question status:</span> No next question selected yet.
</div>
)}
</div>
{reasoningStagesAfter.length > 0 && (
<div className="mt-3 text-sm text-blue-950">
<span className="font-medium">Reasoning stages:</span>{" "}
{reasoningStagesAfter
.map((stage) => `${stage.stage}: ${stage.status}`)
.join(" → ")}
</div>
)}
</section>
<ListSection
+11
View File
@@ -40,6 +40,11 @@ function NodeGroup({
<span className="font-medium text-gray-900">{node.label}</span>
<NodeBadge tone="blue">{node.status}</NodeBadge>
<NodeBadge tone="green">{node.confidence}</NodeBadge>
{node.confidenceAssessment?.completenessStatus && (
<NodeBadge tone="purple">
completeness: {node.confidenceAssessment.completenessStatus}
</NodeBadge>
)}
{resolvedNodeIds.has(node.id) && (
<NodeBadge tone="red">resolved unknown</NodeBadge>
)}
@@ -59,6 +64,12 @@ function NodeGroup({
{node.description && node.description !== node.label && (
<p className="mt-1 text-gray-600">{node.description}</p>
)}
{node.confidenceAssessment && (
<p className="mt-1 text-xs text-gray-500">
evidence: {node.confidenceAssessment.evidenceConfidence} ·
conclusion: {node.confidenceAssessment.conclusionConfidence}
</p>
)}
</li>
))}
</ul>
+40
View File
@@ -0,0 +1,40 @@
# v0.6 Ambiguity Generalisation
## Hypothesis
If the selector truly handles unjustified contradiction ties generically, it should return ambiguity across multiple domains without preferring one explanation by wording alone.
## Scenarios
1. Revenue increased by 18%, but cash in the bank fell over the same period.
2. Customer satisfaction scores increased, but complaints also increased.
3. Average delivery time decreased by 25%, but order cancellations increased.
4. Website traffic doubled, but sales remained unchanged.
5. Production output increased by 30%, but quality defects also increased.
## Observed behaviour
All five fixtures produced the same pattern:
- candidate count: 2
- selector status: `ambiguous`
- tie reason: `No justified distinction between leading unknowns.`
- no explanation was favoured
- one broad investigation question was produced from the central contradiction
- neutral label renaming did not collapse ambiguity into a winner
## Repeated failure patterns
None observed across two or more scenarios.
The current ambiguity handling generalised cleanly across the five contradiction fixtures.
## Corrections
No production correction was required in this task.
## Lessons learned
- The current ambiguity path appears domain-agnostic when structure and semantic weights remain intentionally non-discriminating.
- Central-statement-based tie questions are broad enough to avoid prematurely backing one branch.
- The most useful regression signal is whether ambiguity survives neutral relabelling, not whether one label sorts ahead of another in display order.
+211
View File
@@ -0,0 +1,211 @@
# v0.6 Atomicity Experiment
## Hypothesis
After deterministic unknown selection, the engine should assess whether the selected unknown is already atomic or is still too composite to ask directly.
If the unknown is atomic, the engine should proceed exactly as before.
If the unknown is composite, the engine should not ask that parent unknown directly. Instead, it should decompose it into a small set of explicit child unknowns representing broad, independent candidate dimensions that a non-expert could understand.
## Constraints
- No graph redesign
- No persistence
- No UI redesign
- No selection-weight tuning
- No Ollama calls in unit tests
## Deterministic rule introduced
Atomicity assessment is **not** a new investigation strategy.
It runs in the graph update path at this seam:
```text
unknown selection -> atomicity assessment -> optional decomposition -> deterministic reselection -> question formulation
```
The implementation uses deterministic text and graph-shape checks:
- focused unknowns like denominator / threshold / definition / baseline / evidence remain **atomic**
- broad relationship-explanation unknowns and broad “possible causes / what changed / explanation for why X but Y” unknowns become **composite**
## Decomposition behavior
When a selected unknown is composite:
1. The parent unknown remains unresolved.
2. Between 2 and 5 child unknowns are created or reused deterministically.
3. Children become explicit graph nodes.
4. Children link back to the parent with existing `depends_on` edges.
5. Children inherit the same “why it matters” discipline in their descriptions.
6. Deterministic selection reruns across the updated graph.
For the current relationship-explanation experiment, the broad child dimensions are:
- Whether the two observations reflect different timing
- How the two observations were measured
- Change affecting signal A more than signal B
- Change affecting signal B more than signal A
- One-off event during the period
These are intentionally non-jargon and broad enough to generalise across scenarios like:
- Revenue up / Cash down
- Customer satisfaction up / Complaints up
- Delivery time down / Cancellations up
- Traffic up / Sales flat
- Production up / Defects up
## Diagnostics added
The orchestrator now reports:
- `atomicityAssessment`
- `atomicityDecisionReason`
- `decompositionDepth`
- `decompositionAttempted`
- `decompositionAccepted`
- `decompositionStoppedReason`
- `proposedChildCount`
- `acceptedChildCount`
- `rejectedChildren`
- `selectedChildNodeId`
- `childQualitySummary`
- `propagationPerformed`
- `resolvedChildNodeId`
- `parentNodeId`
- `parentStatusBefore`
- `parentStatusAfter`
- `parentConfidenceBefore`
- `parentConfidenceAfter`
- `affectedAncestorIds`
- `nextSelectedSibling`
- `parentResolved`
- `decompositionPerformed`
- `childUnknownCount`
- `childNodeIds`
- `atomicityReason`
This sits alongside the existing explicit-emergent-unknown diagnostics.
## Observed outcome
The experiment was useful.
Before this change, the engine could select a broad explanation unknown and ask it directly.
After this change:
- the broad explanation parent remains explicit in the graph
- the engine decomposes it into child unknowns first
- the next asked question is backed by a more focused child unknown
- repeated updates reuse the same decomposition children deterministically
- child-quality checks reject compound or duplicate children before they enter the graph
- decomposition stops deterministically once a selected child is directly answerable
- resolving one child does not resolve the parent immediately
- resolved child evidence now propagates upward to the parent and ancestor chain deterministically
- parent status and confidence change conservatively after child resolution
- the next sibling becomes eligible for normal deterministic selection without recreating the resolved child
In the revenue-versus-cash case, the selected next question becomes:
> What evidence would clarify how the two observations were measured?
rather than asking the full broad explanation node directly.
## Upward propagation and reconstruction
Recursive reasoning is complete only when decomposition and reconstruction are both deterministic.
Confidence must not outrun completeness or evidence.
For this experiment, reconstruction now behaves as follows:
- when a child unknown resolves, that child keeps its own resolved status and answer evidence
- the parent is updated, but remains unresolved unless the deterministic completion rule is satisfied
- only the ancestor chain connected to that child is updated
- unrelated branches remain unchanged
- the deterministic selector then chooses the next justified unresolved sibling or related follow-up
For the current conservative completion rule:
- **one resolved child** → parent becomes `provisional` with higher confidence, but remains unresolved
- **all direct child unknowns resolved** → parent resolves deterministically with `high` confidence
The confidence model is now explicitly separated into:
- **evidence confidence**: how trustworthy the currently attached support is
- **completeness**: whether the required direct child structure is empty, partial, or complete
- **conclusion confidence**: how strongly the current parent state is justified given both evidence and completeness
Deterministic propagation rules now enforce:
- one resolved child may raise evidence confidence
- unresolved direct children cap conclusion confidence
- contradictory direct children block high conclusion confidence
- duplicate evidence does not increase confidence
- status changes do not raise confidence on their own
- parent resolution still requires the separate completion rule
## Cross-branch corroboration
The next confidence experiment adds deterministic branch interaction checks without changing the graph model.
The engine now distinguishes between:
- **multiple evidence**: more than one branch exists
- **independent corroboration**: distinct resolved branches support the same parent without sharing the same evidence key
- **duplicate evidence**: the same evidence key appears through multiple branches and must not be double-counted
- **conflicting evidence**: branches support incompatible positions, such as `recognised correctly` vs `recognised incorrectly`
Deterministic branch rules:
- corroboration only counts when branches are distinct and their evidence sources differ
- duplicate evidence groups never count as corroboration
- conflicts cap conclusion confidence and prevent a higher confidence upgrade
- independent branches remain interaction-neutral
Additional diagnostics now expose:
- `corroboratingBranchCount`
- `conflictingBranchCount`
- `duplicateEvidenceCount`
- `independentBranchCount`
- `interactionSummary`
- `confidenceAdjustmentReason`
Observed effect:
- independent corroboration can raise `evidenceConfidence`
- duplicate evidence produces no extra confidence increase
- conflicting evidence lowers or caps `conclusionConfidence`
- completeness rules still dominate whether a parent may become highly justified
Example progression:
- parent before: `unknown`, `medium`
- after resolving `How the two observations were measured`: parent becomes `provisional`, `medium`
- evidence confidence becomes `high`, completeness becomes `partial`, conclusion confidence becomes `medium`
- next sibling becomes selectable and the engine moves on without recreating the resolved child
## Interpretation
This supports the idea that recursive decomposition is a fundamental part of graph-backed questioning, not just a prompt refinement.
The main remaining limitation is that sibling selection still inherits the existing deterministic scorer. That means some domains may advance to a justified sibling that is not the intuitively expected next child, even though the propagation itself remains deterministic and graph-valid.
## Validation run
Covered by:
- `tests/graph/atomicity-assessment.test.js`
- `tests/graph/decomposition-quality.test.js`
- `tests/graph/upward-propagation.test.js`
- `tests/graph/apply-proposal.test.js`
- `tests/graph/orchestrator.test.js`
- `tests/graph/question-formulator.test.js`
- `tests/ui/scenario-form.test.jsx`
And then by the broader requested validation pass with lint and build.
+48
View File
@@ -0,0 +1,48 @@
# v0.6 Comparability Experiment
## Hypothesis
The engine should confirm that observations are comparable before treating their difference as a contradiction that needs explanatory follow-up.
## Fixtures
1. Revenue increased by 18%, but cash in the bank fell over the same period.
2. Complaints increased. Production increased.
3. Average delivery time decreased by 25%, but order cancellations increased.
4. Customer satisfaction increased, but complaints increased.
5. Temperature increased. Ice melted.
6. Sales doubled. Sales doubled.
## Results
- The first four scenarios repeated the same failure pattern: contradiction-level investigation could begin before comparability was established.
- A deterministic comparability gate corrected that by producing one comparison question first.
- Confirmed comparability did not by itself imply contradiction.
- Temperature increased / Ice melted was reclassified as a compatible relationship, so no contradiction question was asked.
- Sales doubled / Sales doubled was reclassified as duplicate observations, so no follow-up question was asked.
## Relationship classification stage
After comparability assessment, observations now pass through a deterministic relationship classification stage:
- `contradictory`
- `compatible`
- `potentially_related`
- `duplicate`
- `insufficient_information`
## Whether comparability should become a permanent reasoning stage
Yes, in minimal deterministic form.
The repeated pattern appeared in four scenarios, so a small pre-contradiction comparability assessment is justified.
## Two-step experiment result
A comparison question is useful only if its answer advances the reasoning stage rather than merely adding more text.
In the revenue-versus-cash scenario, the first question now confirms whether the figures are comparable, and the answer resolves that existing uncertainty instead of creating a parallel note. After that update, the engine progresses from comparability assessment to cautious relationship assessment and can select one broad non-expert follow-up question.
Every justified next question should correspond to an explicit unresolved graph node.
The earlier fallback-only path has now been removed from the normal successful progression. After comparability is resolved and a further investigation question is justified, the engine creates or reuses an explicit unresolved reasoning unknown and lets deterministic selection and question formulation proceed through the standard graph pipeline. A fallback is now only acceptable as an explicit failure case, not as the normal source of the next question.
+375
View File
@@ -0,0 +1,375 @@
# v0.6 Reasoning Architecture
## Purpose
This document describes the implemented deterministic reasoning architecture on branch `feature/question-strategy-alignment-v0.6`.
It is written for future developers who need to understand how v0.6 actually executes, what invariants it relies on, where the recursive loops are, and what the system deliberately does **not** attempt to do.
## End-to-end pipeline
The implemented runtime pipeline is:
```text
Scenario input
LLM analysis / reconstruction
Initial graph build
Deterministic unknown selection
Selected question
User answer
LLM graph-update proposal
Proposal parsing / normalisation
Proposal compatibility validation
Deterministic graph update application
Reasoning-state rebuild
Comparability assessment
Relationship classification
Explicit emergent unknown creation / reuse (if required)
Deterministic reselection
Atomicity assessment
Optional decomposition into child unknowns
Deterministic reselection
Resolved-child propagation upward
Confidence / completeness / corroboration update
Next active unknown
Question formulation
```
## Deterministic stages
### 1. Scenario analysis / reconstruction
- **Purpose**: obtain structured reconstruction material from scenario text
- **Input**: scenario, prompt version
- **Output**: analysis payload containing reconstruction, evidence, diagnostics, and optional next question
- **Why it exists**: provides the initial structured substrate from which the graph is built
- **What breaks if removed**: the graph builder has no structured reconstruction to convert into nodes and edges
### 2. Initial graph build
- **Purpose**: convert reconstruction output into an initial `SituationGraph`
- **Input**: reconstruction + evidence
- **Output**: graph nodes and edges, then `makeGraph(...)` wraps them with active/resolved/summary state
- **Why it exists**: all later reasoning is graph-based, not free text
- **What breaks if removed**: no explicit unknown nodes, no deterministic selection, no validated update loop
### 3. Deterministic unknown selection
- **Purpose**: choose the next active unknown from unresolved graph nodes
- **Input**: graph, resolved node IDs
- **Output**: selected candidate or explicit ambiguity result
- **Why it exists**: the system needs a deterministic next investigation target
- **What breaks if removed**: question ordering becomes arbitrary or hidden in prompts
### 4. Selected question exposure
- **Purpose**: expose the chosen unknown as the next question to the user
- **Input**: selected unknown + question formulation or tie-resolution logic
- **Output**: selected question object
- **Why it exists**: the user-facing loop must ask a concrete next question
- **What breaks if removed**: the system can build a graph but cannot continue interaction coherently
### 5. LLM graph-update proposal
- **Purpose**: transform a user answer into a proposed graph change set
- **Input**: current graph, previous question, answer, prompt version
- **Output**: raw JSON-like proposal
- **Why it exists**: the LLM is limited to proposing changes; it does not mutate the graph directly
- **What breaks if removed**: answers cannot affect the graph except through manual hard-coded logic
### 6. Proposal parsing / normalisation
- **Purpose**: parse JSON, remove null array items, apply known aliases, fill omitted optional fields
- **Input**: raw model response
- **Output**: validated `graphUpdateSchema` payload or structured parser failure
- **Why it exists**: model outputs are not trusted as-is
- **What breaks if removed**: malformed or partially missing model output would reach graph logic directly
### 7. Proposal compatibility validation
- **Purpose**: ensure the proposal is graph-safe and semantically valid before application
- **Input**: current graph + proposed update
- **Output**: accepted proposal or compatibility errors
- **Why it exists**: protects graph integrity and reasoning invariants
- **What breaks if removed**: duplicate IDs, missing references, fake selected questions, and no-op updates could corrupt the graph
### 8. Deterministic graph update application
- **Purpose**: apply only validated graph changes to a copied graph
- **Input**: graph + validated proposal
- **Output**: updated nodes, edges, resolved node IDs
- **Why it exists**: separates safe application from generation
- **What breaks if removed**: no explicit, replayable state transition exists
### 9. Reasoning-state rebuild
- **Purpose**: derive fresh comparability/relationship state from the updated graph
- **Input**: updated graph + optional override state
- **Output**: `reasoningState`
- **Why it exists**: reasoning stages are derived from graph state, not stored blindly
- **What breaks if removed**: comparability and relationship decisions drift from actual graph contents
### 10. Comparability assessment
- **Purpose**: decide whether supported observations are comparable enough for relationship reasoning
- **Input**: graph observations + central statement + optional stored override
- **Output**: comparability status/reason + contradiction permission
- **Why it exists**: relationship reasoning is gated by comparability
- **What breaks if removed**: contradiction or relationship reasoning would run over incomparable observations
### 11. Relationship classification
- **Purpose**: classify observation relationships once comparability permits it
- **Input**: graph + comparability result
- **Output**: relationship status, reason, whether a follow-up question is justified
- **Why it exists**: determines whether explanation-style follow-up is needed
- **What breaks if removed**: the system cannot distinguish compatible, duplicate, insufficient, and contradiction-adjacent observation sets
### 12. Explicit emergent unknown creation / reuse
- **Purpose**: ensure any justified relationship follow-up is represented by an explicit unresolved graph node
- **Input**: provisional graph + relationship assessment
- **Output**: reused or newly added explanation unknown and edges
- **Why it exists**: preserves the invariant that a question must originate from an explicit unknown
- **What breaks if removed**: relationship follow-up would revert to fallback-only question text not backed by the graph
### 13. Atomicity assessment
- **Purpose**: determine whether the selected unknown is directly investigable or too composite
- **Input**: selected unknown + graph context
- **Output**: `atomic` or `composite` decision with decomposition kind/reason
- **Why it exists**: prevents asking broad explanation unknowns directly
- **What breaks if removed**: the system asks high-level composite unknowns instead of decomposing them first
### 14. Optional decomposition
- **Purpose**: split a composite unknown into deterministic child unknowns
- **Input**: composite selected unknown + graph context
- **Output**: 25 child unknowns, edges, quality summary, rejection diagnostics
- **Why it exists**: narrows broad unknowns into explicit candidate dimensions
- **What breaks if removed**: recursive reasoning stops at broad parents and loses graph-backed substructure
### 15. Resolved-child propagation upward
- **Purpose**: move resolved child effects to parent and ancestor chain without prematurely resolving them
- **Input**: updated graph + proposal snapshot
- **Output**: parent/ancestor status and confidence updates, additional diagnostics
- **Why it exists**: decomposition requires deterministic reconstruction as well as decomposition
- **What breaks if removed**: child answers stay local and parents never become progressively better-supported
### 16. Confidence / completeness / corroboration update
- **Purpose**: derive parent-level `confidenceAssessment` from resolved children and branch interactions
- **Input**: parent child set + branch evidence/status interactions
- **Output**: `evidenceConfidence`, `completenessStatus`, `conclusionConfidence`, plus derived display `confidence`
- **Why it exists**: reasoning support must be separated from completion and contradiction state
- **What breaks if removed**: parent confidence collapses back into vague status-driven heuristics
### 17. Next active unknown + question formulation
- **Purpose**: reselect the next unresolved unknown and formulate a concrete next question
- **Input**: updated graph + selection state + graph context
- **Output**: next active unknown and question object
- **Why it exists**: closes the recursive interaction loop
- **What breaks if removed**: the system updates the graph but cannot continue investigation deterministically
## Architectural invariants
The current implementation enforces these invariants:
1. **A question must originate from an explicit unresolved unknown node.**
2. **Unknown selection is deterministic.**
3. **Alphabetical ordering is not treated as reasoning.**
4. **Relationship reasoning does not precede comparability.**
5. **The LLM never mutates the graph directly; it only proposes updates.**
6. **All graph updates are schema-validated before application.**
7. **All node/edge references must resolve to existing nodes.**
8. **Duplicate node IDs are rejected.**
9. **Duplicate added edge IDs are rejected.**
10. **A selected question cannot target a resolved unknown.**
11. **A resolved unknown updated to `resolved` must also appear in `resolvedUnknownNodeIds`.**
12. **A proposal must contain a meaningful change.**
13. **Every newly added unknown must include why-it-matters language.**
14. **Every newly added unknown must be explicitly connected to answer-derived graph structure.**
15. **Composite selected unknowns are decomposed before direct questioning when atomicity rules require it.**
16. **Parent unknowns remain unresolved until completion rules are satisfied.**
17. **Confidence must not outrun completeness.**
18. **Duplicate evidence cannot increase confidence.**
19. **Conflicting evidence caps conclusion confidence.**
20. **Cross-branch corroboration only counts for distinct branches with distinct evidence keys.**
21. **Ambiguous leading unknowns remain explicit ambiguity, not silent forced choice.**
## Recursive loops and stopping rules
### Main investigation loop
```text
Unknown
Question
Answer
Proposal
Graph update
Propagation
Next unknown
```
- **Exit condition**: no unresolved candidates remain, or no next question is justified, or proposal/application fails
- **Stopping rule**: deterministic selection returns `null` or explicit ambiguity, or update validation blocks progress
- **Completion behaviour**: continues only while the graph contains justified unresolved unknowns
### Decomposition loop
```text
Selected unknown
Atomicity assessment
If composite: decompose
Reselect child
Atomicity assessment again
```
- **Exit condition**: selected child is atomic; parent already has children; max decomposition depth reached; or decomposition quality fails
- **Stopping rule**: `MAX_DECOMPOSITION_DEPTH`, reuse instead of regeneration, or inability to produce enough valid child unknowns
- **Completion behaviour**: deterministic and bounded; no infinite recursive decomposition path is intentionally allowed
### Propagation loop
```text
Resolved child
Ancestor chain walk
Recompute parent state
Stop when no ancestor state changes
```
- **Exit condition**: no more parents in the ancestor chain or no state change
- **Stopping rule**: ancestor chain is explicit and finite; propagation does not invent new ancestors
- **Completion behaviour**: deterministic upward traversal with explicit stop on unchanged state
### Potential infinite loops reviewed
- **Unknown/question recursion**: bounded by unresolved unknown set, proposal validation, and explicit no-candidate states
- **Decomposition recursion**: bounded by max depth and child reuse rules
- **Propagation recursion**: bounded by finite ancestor chain and no-change stop condition
No intentional infinite reasoning loop is present in the implemented architecture.
## Graph lifecycle summary
### Node lifecycle
1. node created by `buildInitialGraph` or later proposal/decomposition/emergent-unknown logic
2. node validated by schema
3. node may become active unknown
4. node may be updated by proposal application
5. unknown node may become `resolved`, `provisional`, `contradicted`, or remain `unknown`
6. resolved unknown ID is tracked in `resolvedNodeIds`
### Edge lifecycle
1. edge created in initial graph or by deterministic proposal augmentation
2. edge validated against existing node IDs
3. edge may be removed only through explicit `removedEdgeIds`
4. edge relationships also update `dependsOn` / `childIds` projections during application
### Unknown lifecycle
1. initial unknown discovered from reconstruction
2. selected deterministically or left ambiguous
3. may be decomposed if composite
4. may be resolved directly by answer
5. may cause emergent reasoning unknown creation when relationship reasoning demands a new explicit question target
### Resolved lifecycle
1. proposal marks unresolved unknown resolved
2. reconciliation ensures resolution semantics are explicit
3. `resolvedUnknownNodeIds` feed graph application
4. propagation may resolve parent only when completion rule is met
### Confidence lifecycle
1. nodes begin with base `confidence`
2. parent/ancestor propagation derives `confidenceAssessment`
3. display `confidence` is derived from `conclusionConfidence`
4. completeness, duplicate evidence, contradiction, and corroboration constrain the result
### Question lifecycle
1. selected unknown becomes question target
2. `formulateQuestion` or tie-resolution logic produces question text
3. answer returns through update route
4. proposal may select a new question target or leave reselection to deterministic logic
Every major transition above is explicit in the current codebase rather than implicit in model text alone.
## Duplicated or overlapping concepts
The following concepts are intentionally close and may look duplicated:
- **status vs confidence**: status captures lifecycle/progression; confidence captures support strength
- **confidence vs confidenceAssessment**: `confidence` is now a derived display field, while `confidenceAssessment` carries separated reasoning dimensions
- **resolvedNodeIds vs node.status === resolved**: both are maintained; the first is a graph-level index, the second is node-local state
- **selectedQuestion in proposal vs selectedQuestion in final result**: proposal may omit or propose one, final result recomputes deterministic selection/questioning after graph logic
- **comparability state in reasoningState vs derived comparability from graph**: overrides may carry forward prior confirmed reasoning, but `buildReasoningState` still rebuilds from graph + override context
These are not necessarily defects, but they are the main places where future simplification pressure is likely.
## Known boundaries and deliberate exclusions
v0.6 deliberately does **not** attempt the following:
- probabilistic reasoning
- Bayesian inference
- persistence
- semantic embeddings
- fuzzy semantic similarity
- autonomous exploration outside explicit user answers
- multi-hop corroboration across unrelated subtrees without a shared direct parent
- expert-only jargon-specific reasoning modes
- UI-heavy reasoning visualisation beyond existing graph/update displays
- arbitrary non-deterministic tie breaking
## Defects found during this review
No new production defect was intentionally introduced or fixed as part of this architecture review.
## Developer notes
- `startCase` owns reconstruction → graph build → first deterministic selection.
- `updateCaseWithDependencies` owns proposal generation / parsing and delegates deterministic graph semantics to `applyValidatedProposal`.
- `applyValidatedProposal` is the main reasoning pipeline coordinator for update-time graph evolution.
- `question-formulator.js` owns comparability, relationship classification, atomicity assessment, investigation strategy selection, and question formulation.
- `utils.js` owns selection scoring, ordering, graph validation, and safe graph update application.
+89
View File
@@ -0,0 +1,89 @@
# v0.6 Release Notes
## Purpose
v0.6 turns the engine into a deterministic recursive reasoning system that keeps next questions, decomposition, propagation, and confidence updates explicitly grounded in the situation graph.
## Capabilities added
- deterministic unknown selection explanations
- explicit ambiguity handling instead of silent tie-breaking
- comparability assessment before relationship reasoning
- relationship classification after comparability
- reasoning-stage progression after comparability answers
- graph-backed next questions via explicit unknown nodes
- investigation-strategy-based question formulation
- atomicity assessment for selected unknowns
- composite-unknown decomposition into child unknowns
- child-quality validation for decomposition outputs
- upward propagation from resolved children to parents and ancestors
- separation of evidence confidence, completeness, and conclusion confidence
- deterministic cross-branch corroboration, conflict, and duplicate-evidence handling
- developer-facing reasoning architecture documentation
## Reasoning pipeline summary
```text
Scenario
→ Reconstruction
→ Initial graph
→ Deterministic unknown selection
→ Question
→ Answer
→ Proposal
→ Proposal parsing / validation
→ Graph update
→ Reasoning-state rebuild
→ Comparability assessment
→ Relationship classification
→ Emergent unknown creation / reuse
→ Atomicity assessment
→ Optional decomposition
→ Propagation
→ Confidence / completeness / corroboration update
→ Next active unknown
→ Next question
```
## Core invariants
- every asked question must originate from an explicit unresolved unknown
- unknown selection is deterministic
- ambiguity is preserved explicitly when no justified distinction exists
- relationship reasoning cannot precede comparability
- parent nodes cannot resolve before completion rules are met
- confidence cannot outrun completeness
- duplicate evidence cannot increase confidence
- conflicting evidence caps conclusion confidence
- cross-branch corroboration only counts for distinct branches with distinct evidence keys
- the LLM proposes updates but does not mutate the graph directly
## What v0.6 proved
- graph-backed questioning works better when every justified next question maps to an explicit unresolved node
- broad unknowns can be decomposed deterministically before direct questioning
- resolved child evidence can be propagated upward without prematurely resolving parent reasoning
- confidence becomes easier to reason about when evidence quality, completeness, and conclusion strength are separated
- deterministic cross-branch corroboration can improve support without double-counting repeated evidence
## Known limitations
- sibling selection still depends on the existing deterministic scorer and may choose a justified next branch that is not always the intuitively expected one
- cross-branch corroboration is limited to direct child branches of the same parent
- no multi-hop corroboration exists across unrelated subtrees
- reasoning remains bounded to explicitly represented graph structure and user-provided answers
## Deliberate exclusions
- no persistence
- no autonomous exploration
- no probabilistic reasoning
- no Bayesian reasoning
- no semantic embeddings
- no expert mode
- no multi-hop corroboration across unrelated subtrees
- no heavy graph visualisation
## Next experimental question
`Can the engine preserve and reuse successful reasoning structures across separate cases without turning prior experience into unquestioned assumptions?`
@@ -0,0 +1,50 @@
# v0.6 Selection Influence Experiment
## Hypothesis
The initial unknown selected for the revenue-versus-cash scenario may be driven more by graph structure, more by semantic keyword matches, or by both together.
## Scenario
`Revenue increased by 18%, but cash in the bank fell over the same period.`
## Actual selected node
- Node ID: `nqdzobz`
- Label: `Magnitude and nature of cash outflows (operating expenses, debt repayments, capex, or working capital shifts).`
- Deterministic investigation strategy: `definition`
- Deterministic question: `What evidence would resolve whether magnitude and nature of cash outflows (operating expenses, debt repayments, capex, or working capital shifts). is true?`
## Structural contribution
- Downstream dependency count: `0`
- Prerequisite position: no unresolved prerequisites; count `0`
- Dependency ordering / centrality: no candidate had downstream dependants or dependency depth advantage in the live graph
## Semantic contribution
- Objective: false
- Actor: false
- Criteria: false
- Measurement: false
- Terminology: false
- Constraint: false
- Pricing: false
- Implementation: false
- Optimisation: false
- Speculative: false
- Contribution list: only `downstream_dependencies` was present, with delta `0`
## Counterfactual results
- Live-shaped ordering: `nqdzobz` ranked above `niewza`, but both had score `0`, downstream `0`, and unresolved prerequisites `0`
- Links removed: ordering stayed the same, because the live graph already provided no differentiating structure between the two unknowns
- Wording neutralised: ordering flipped to the first unknown by neutral label order (`Unknown A` before `Unknown B`), showing the outcome remained tie-break-driven rather than structure-driven
## Conclusion
For this scenario, the actual winner was not selected because of graph structure and not selected because of semantic keyword weights. The live diagnostics show a complete tie on score, downstream influence, and prerequisite position, with every semantic match category false for both candidates. The winner was therefore chosen by the final tie-break rule, `label_asc`.
## Is a scoring change justified?
Not from this single experiment alone. The result shows a diagnostic gap for this scenario, but this task does not justify a scoring change by itself, and no scoring change is made.
File diff suppressed because it is too large Load Diff
+336 -9
View File
@@ -15,8 +15,13 @@ import {
import { buildInitialGraph, describeGraph } from "./builder.js";
import { applyValidatedProposal } from "./apply-proposal.js";
import { buildGraphUpdatePrompt } from "./prompt-builder.js";
import {
buildReasoningState,
formulateTieResolutionQuestion,
} from "./question-formulator.js";
import { parseGraphUpdateProposal } from "./update-proposal.js";
import {
explainUnknownSelection,
selectActiveUnknownCandidate,
validateGraphReferences,
} from "./utils.js";
@@ -31,7 +36,12 @@ function toValidationErrors(error) {
);
}
function buildDiagnostics({ analysis, graph, graphReferenceValidation }) {
function buildDiagnostics({
analysis,
graph,
graphReferenceValidation,
unknownSelectionExplanation,
}) {
return {
promptVersion: analysis?.promptVersion ?? null,
modelName: analysis?.modelName ?? null,
@@ -43,9 +53,30 @@ function buildDiagnostics({ analysis, graph, graphReferenceValidation }) {
compatibilityApplied: analysis?.compatibilityApplied ?? false,
compatibilityChanges: analysis?.compatibilityChanges ?? [],
compatibilityWarnings: analysis?.compatibilityWarnings ?? [],
unknownSelectionExplanation: unknownSelectionExplanation ?? null,
};
}
function buildUnknownSelectionDiagnostics(
graph,
resolvedNodeIds = [],
selectedQuestion = null,
) {
const explanation = explainUnknownSelection(graph, resolvedNodeIds);
if (explanation.status === "ambiguous") {
return {
...explanation,
tieResolutionQuestion:
selectedQuestion?.selectionStatus === "ambiguous"
? selectedQuestion.question
: formulateTieResolutionQuestion({ graph }).question,
alphabeticalUsedAsReasoning: false,
};
}
return explanation;
}
function buildUpdateDiagnostics({
promptVersion,
modelName,
@@ -53,6 +84,55 @@ function buildUpdateDiagnostics({
normalisationsApplied,
graph,
graphReferenceValidation,
selectedQuestion,
unknownSelectionExplanation,
previousReasoningState,
reasoningState,
resolvedReasoningNodeIds,
emergentReasoningNodeCreated,
emergentReasoningNodeId,
emergentReasoningNodeReason,
atomicityAssessment,
atomicityDecisionReason,
decompositionDepth,
decompositionAttempted,
decompositionAccepted,
decompositionStoppedReason,
proposedChildCount,
acceptedChildCount,
rejectedChildren,
selectedChildNodeId,
childQualitySummary,
propagationPerformed,
resolvedChildNodeId,
parentNodeId,
parentStatusBefore,
parentStatusAfter,
parentConfidenceBefore,
parentConfidenceAfter,
evidenceConfidenceBefore,
evidenceConfidenceAfter,
completenessBefore,
completenessAfter,
conclusionConfidenceBefore,
conclusionConfidenceAfter,
resolvedDirectChildren,
unresolvedDirectChildren,
contradictoryDirectChildren,
corroboratingBranchCount,
conflictingBranchCount,
duplicateEvidenceCount,
independentBranchCount,
interactionSummary,
confidenceCapReason,
ancestorPropagationStoppedReason,
affectedAncestorIds,
nextSelectedSibling,
parentResolved,
decompositionPerformed,
childUnknownCount,
childNodeIds,
atomicityReason,
}) {
return {
promptVersion: promptVersion ?? "v0.4",
@@ -66,6 +146,63 @@ function buildUpdateDiagnostics({
errors: [],
},
normalisationsApplied: normalisationsApplied ?? [],
investigationStrategy:
selectedQuestion?.investigationStrategy ??
selectedQuestion?.strategy ??
null,
previousComparabilityStatus:
previousReasoningState?.comparabilityStatus ?? null,
comparabilityStatus: reasoningState?.comparabilityStatus ?? null,
relationshipStatus: reasoningState?.relationshipStatus ?? null,
relationshipAssessed: reasoningState?.relationshipAssessed ?? null,
reasoningStagesBefore: previousReasoningState?.reasoningStages ?? [],
reasoningStagesAfter: reasoningState?.reasoningStages ?? [],
resolvedReasoningNodeIds: resolvedReasoningNodeIds ?? [],
emergentReasoningNodeCreated: emergentReasoningNodeCreated ?? false,
emergentReasoningNodeId: emergentReasoningNodeId ?? null,
emergentReasoningNodeReason: emergentReasoningNodeReason ?? null,
atomicityAssessment: atomicityAssessment ?? null,
atomicityDecisionReason: atomicityDecisionReason ?? null,
decompositionDepth: decompositionDepth ?? 0,
decompositionAttempted: decompositionAttempted ?? false,
decompositionAccepted: decompositionAccepted ?? false,
decompositionStoppedReason: decompositionStoppedReason ?? null,
proposedChildCount: proposedChildCount ?? 0,
acceptedChildCount: acceptedChildCount ?? 0,
rejectedChildren: rejectedChildren ?? [],
selectedChildNodeId: selectedChildNodeId ?? null,
childQualitySummary: childQualitySummary ?? [],
propagationPerformed: propagationPerformed ?? false,
resolvedChildNodeId: resolvedChildNodeId ?? null,
parentNodeId: parentNodeId ?? null,
parentStatusBefore: parentStatusBefore ?? null,
parentStatusAfter: parentStatusAfter ?? null,
parentConfidenceBefore: parentConfidenceBefore ?? null,
parentConfidenceAfter: parentConfidenceAfter ?? null,
evidenceConfidenceBefore: evidenceConfidenceBefore ?? null,
evidenceConfidenceAfter: evidenceConfidenceAfter ?? null,
completenessBefore: completenessBefore ?? null,
completenessAfter: completenessAfter ?? null,
conclusionConfidenceBefore: conclusionConfidenceBefore ?? null,
conclusionConfidenceAfter: conclusionConfidenceAfter ?? null,
resolvedDirectChildren: resolvedDirectChildren ?? 0,
unresolvedDirectChildren: unresolvedDirectChildren ?? 0,
contradictoryDirectChildren: contradictoryDirectChildren ?? 0,
corroboratingBranchCount: corroboratingBranchCount ?? 0,
conflictingBranchCount: conflictingBranchCount ?? 0,
duplicateEvidenceCount: duplicateEvidenceCount ?? 0,
independentBranchCount: independentBranchCount ?? 0,
interactionSummary: interactionSummary ?? null,
confidenceCapReason: confidenceCapReason ?? null,
ancestorPropagationStoppedReason: ancestorPropagationStoppedReason ?? null,
affectedAncestorIds: affectedAncestorIds ?? [],
nextSelectedSibling: nextSelectedSibling ?? null,
parentResolved: parentResolved ?? false,
decompositionPerformed: decompositionPerformed ?? false,
childUnknownCount: childUnknownCount ?? 0,
childNodeIds: childNodeIds ?? [],
atomicityReason: atomicityReason ?? null,
unknownSelectionExplanation: unknownSelectionExplanation ?? null,
};
}
@@ -105,14 +242,17 @@ export async function startCase(body) {
});
const currentSummary = describeGraph(initialGraph);
const deterministicSelection = selectActiveUnknownCandidate(
{
...initialGraph,
resolvedNodeIds: [],
},
[],
);
const activeUnknownNodeId =
selectActiveUnknownCandidate(
{
...initialGraph,
resolvedNodeIds: [],
},
[],
)?.nodeId ?? null;
deterministicSelection?.status === "selected"
? deterministicSelection.nodeId
: null;
const situationGraph = makeGraph({
centralStatement: scenario,
@@ -121,11 +261,30 @@ export async function startCase(body) {
activeUnknownNodeId,
resolvedNodeIds: [],
currentSummary,
reasoningState: buildReasoningState({
centralStatement: scenario,
nodes: initialGraph.nodes,
edges: initialGraph.edges,
resolvedNodeIds: [],
}),
});
situationGraphSchema.parse(situationGraph);
const graphReferenceValidation = validateGraphReferences(situationGraph);
const selectedQuestion =
deterministicSelection?.status === "ambiguous"
? {
id: "q_tie_resolution",
...formulateTieResolutionQuestion({ graph: situationGraph }),
tiedCandidateIds: deterministicSelection.tiedCandidateIds,
}
: (analysis.nextQuestion ?? null);
const unknownSelectionExplanation = buildUnknownSelectionDiagnostics(
situationGraph,
[],
selectedQuestion,
);
if (!graphReferenceValidation.valid) {
return {
success: false,
@@ -134,6 +293,7 @@ export async function startCase(body) {
analysis,
graph: situationGraph,
graphReferenceValidation,
unknownSelectionExplanation,
}),
validationErrors: graphReferenceValidation.errors,
statusCode: 500,
@@ -143,11 +303,12 @@ export async function startCase(body) {
return {
success: true,
situationGraph,
selectedQuestion: analysis.nextQuestion ?? null,
selectedQuestion,
diagnostics: buildDiagnostics({
analysis,
graph: situationGraph,
graphReferenceValidation,
unknownSelectionExplanation,
}),
};
}
@@ -269,6 +430,8 @@ async function updateCaseWithDependencies(body, dependencies = {}) {
const applicationResult = applyProposalUpdate({
situationGraph,
proposal: parsedProposal.proposal,
previousQuestion,
answer,
});
if (!applicationResult.success) {
@@ -284,6 +447,58 @@ async function updateCaseWithDependencies(body, dependencies = {}) {
normalisationsApplied: parsedProposal.normalisationsApplied,
graph: situationGraph,
graphReferenceValidation: graphReferenceValidation,
selectedQuestion: null,
previousReasoningState: buildReasoningState(situationGraph),
reasoningState: buildReasoningState(situationGraph),
resolvedReasoningNodeIds: [],
emergentReasoningNodeCreated: false,
emergentReasoningNodeId: null,
emergentReasoningNodeReason: null,
atomicityAssessment: null,
atomicityDecisionReason: null,
decompositionDepth: 0,
decompositionAttempted: false,
decompositionAccepted: false,
decompositionStoppedReason: null,
proposedChildCount: 0,
acceptedChildCount: 0,
rejectedChildren: [],
selectedChildNodeId: null,
childQualitySummary: [],
propagationPerformed: false,
resolvedChildNodeId: null,
parentNodeId: null,
parentStatusBefore: null,
parentStatusAfter: null,
parentConfidenceBefore: null,
parentConfidenceAfter: null,
evidenceConfidenceBefore: null,
evidenceConfidenceAfter: null,
completenessBefore: null,
completenessAfter: null,
conclusionConfidenceBefore: null,
conclusionConfidenceAfter: null,
resolvedDirectChildren: 0,
unresolvedDirectChildren: 0,
contradictoryDirectChildren: 0,
corroboratingBranchCount: 0,
conflictingBranchCount: 0,
duplicateEvidenceCount: 0,
independentBranchCount: 0,
interactionSummary: null,
confidenceCapReason: null,
ancestorPropagationStoppedReason: null,
affectedAncestorIds: [],
nextSelectedSibling: null,
parentResolved: false,
decompositionPerformed: false,
childUnknownCount: 0,
childNodeIds: [],
atomicityReason: null,
unknownSelectionExplanation: explainUnknownSelection(
situationGraph,
situationGraph.resolvedNodeIds || [],
),
}),
},
statusCode:
@@ -313,6 +528,65 @@ async function updateCaseWithDependencies(body, dependencies = {}) {
normalisationsApplied: parsedProposal.normalisationsApplied,
graph: applicationResult.updatedSituationGraph,
graphReferenceValidation: applicationResult.graphReferenceValidation,
selectedQuestion: applicationResult.selectedQuestion,
previousReasoningState: applicationResult.previousReasoningState,
reasoningState: applicationResult.reasoningState,
resolvedReasoningNodeIds: applicationResult.resolvedReasoningNodeIds,
emergentReasoningNodeCreated:
applicationResult.emergentReasoningNodeCreated,
emergentReasoningNodeId: applicationResult.emergentReasoningNodeId,
emergentReasoningNodeReason:
applicationResult.emergentReasoningNodeReason,
atomicityAssessment: applicationResult.atomicityAssessment,
atomicityDecisionReason: applicationResult.atomicityDecisionReason,
decompositionDepth: applicationResult.decompositionDepth,
decompositionAttempted: applicationResult.decompositionAttempted,
decompositionAccepted: applicationResult.decompositionAccepted,
decompositionStoppedReason:
applicationResult.decompositionStoppedReason,
proposedChildCount: applicationResult.proposedChildCount,
acceptedChildCount: applicationResult.acceptedChildCount,
rejectedChildren: applicationResult.rejectedChildren,
selectedChildNodeId: applicationResult.selectedChildNodeId,
childQualitySummary: applicationResult.childQualitySummary,
propagationPerformed: applicationResult.propagationPerformed,
resolvedChildNodeId: applicationResult.resolvedChildNodeId,
parentNodeId: applicationResult.parentNodeId,
parentStatusBefore: applicationResult.parentStatusBefore,
parentStatusAfter: applicationResult.parentStatusAfter,
parentConfidenceBefore: applicationResult.parentConfidenceBefore,
parentConfidenceAfter: applicationResult.parentConfidenceAfter,
evidenceConfidenceBefore: applicationResult.evidenceConfidenceBefore,
evidenceConfidenceAfter: applicationResult.evidenceConfidenceAfter,
completenessBefore: applicationResult.completenessBefore,
completenessAfter: applicationResult.completenessAfter,
conclusionConfidenceBefore:
applicationResult.conclusionConfidenceBefore,
conclusionConfidenceAfter: applicationResult.conclusionConfidenceAfter,
resolvedDirectChildren: applicationResult.resolvedDirectChildren,
unresolvedDirectChildren: applicationResult.unresolvedDirectChildren,
contradictoryDirectChildren:
applicationResult.contradictoryDirectChildren,
corroboratingBranchCount: applicationResult.corroboratingBranchCount,
conflictingBranchCount: applicationResult.conflictingBranchCount,
duplicateEvidenceCount: applicationResult.duplicateEvidenceCount,
independentBranchCount: applicationResult.independentBranchCount,
interactionSummary: applicationResult.interactionSummary,
confidenceCapReason: applicationResult.confidenceCapReason,
ancestorPropagationStoppedReason:
applicationResult.ancestorPropagationStoppedReason,
affectedAncestorIds: applicationResult.affectedAncestorIds,
nextSelectedSibling: applicationResult.nextSelectedSibling,
parentResolved: applicationResult.parentResolved,
decompositionPerformed: applicationResult.decompositionPerformed,
childUnknownCount: applicationResult.childUnknownCount,
childNodeIds: applicationResult.childNodeIds,
atomicityReason: applicationResult.atomicityReason,
unknownSelectionExplanation: buildUnknownSelectionDiagnostics(
applicationResult.updatedSituationGraph,
applicationResult.updatedSituationGraph.resolvedNodeIds || [],
applicationResult.selectedQuestion,
),
}),
};
}
@@ -328,6 +602,59 @@ async function updateCaseWithDependencies(body, dependencies = {}) {
normalisationsApplied: parsedProposal.normalisationsApplied,
graph: situationGraph,
graphReferenceValidation,
selectedQuestion: null,
previousReasoningState: buildReasoningState(situationGraph),
reasoningState: buildReasoningState(situationGraph),
resolvedReasoningNodeIds: [],
emergentReasoningNodeCreated: false,
emergentReasoningNodeId: null,
emergentReasoningNodeReason: null,
atomicityAssessment: null,
atomicityDecisionReason: null,
decompositionDepth: 0,
decompositionAttempted: false,
decompositionAccepted: false,
decompositionStoppedReason: null,
proposedChildCount: 0,
acceptedChildCount: 0,
rejectedChildren: [],
selectedChildNodeId: null,
childQualitySummary: [],
propagationPerformed: false,
resolvedChildNodeId: null,
parentNodeId: null,
parentStatusBefore: null,
parentStatusAfter: null,
parentConfidenceBefore: null,
parentConfidenceAfter: null,
evidenceConfidenceBefore: null,
evidenceConfidenceAfter: null,
completenessBefore: null,
completenessAfter: null,
conclusionConfidenceBefore: null,
conclusionConfidenceAfter: null,
resolvedDirectChildren: 0,
unresolvedDirectChildren: 0,
contradictoryDirectChildren: 0,
corroboratingBranchCount: 0,
conflictingBranchCount: 0,
duplicateEvidenceCount: 0,
independentBranchCount: 0,
interactionSummary: null,
confidenceCapReason: null,
ancestorPropagationStoppedReason: null,
affectedAncestorIds: [],
nextSelectedSibling: null,
parentResolved: false,
decompositionPerformed: false,
childUnknownCount: 0,
childNodeIds: [],
atomicityReason: null,
unknownSelectionExplanation: buildUnknownSelectionDiagnostics(
situationGraph,
situationGraph.resolvedNodeIds || [],
null,
),
}),
};
}
+784 -96
View File
@@ -11,6 +11,13 @@ function sentenceCase(value) {
return trimmed.charAt(0).toLowerCase() + trimmed.slice(1);
}
function stripTrailingPunctuation(value) {
return String(value || "")
.trim()
.replace(/[.?!:;]+$/g, "")
.trim();
}
function buildNodeMap(graph) {
return new Map((graph?.nodes || []).map((node) => [node.id, node]));
}
@@ -52,8 +59,8 @@ function collectResolvedContextValues(graph) {
function extractMeaning(node) {
const raw = `${node?.label || ""} ${node?.description || ""}`.trim();
let meaning = String(
node?.label || node?.description || "this uncertainty",
let meaning = stripTrailingPunctuation(
String(node?.label || node?.description || "this uncertainty"),
).trim();
const lowered = normaliseText(raw);
@@ -79,6 +86,608 @@ function extractMeaning(node) {
return sentenceCase(meaning);
}
function isDefinitionLikeUnknown(nodeText, text) {
return (
/\b(define|definition|meaning|term|terminology)\b/.test(nodeText) ||
(/\bdefinition\b/.test(text) && /\bdisagreement\b/.test(text)) ||
(/\b(define|definition|meaning|term|terminology)\b/.test(text) &&
/\b(unclear|ambiguous|inconsistent|undefined|used inconsistently)\b/.test(
text,
))
);
}
function isClaimLikeUnknown(node, text) {
return (
node?.kind === "reported_claim" ||
node?.kind === "conclusion" ||
/\b(claim|assertion|true|false|correct|incorrect|happened|happening)\b/.test(
text,
) ||
/^whether\b/i.test(String(node?.label || "").trim())
);
}
function sanitizeQuestionText(question) {
return String(question || "")
.replace(/\)\.\s+/g, ") ")
.replace(/\s+/g, " ")
.trim();
}
function buildNeutralClarificationQuestion(meaning) {
return `What would clarify ${stripTrailingPunctuation(meaning)} in this situation?`;
}
function buildEvidenceFallbackQuestion(meaning) {
return `What evidence would confirm or rule out ${stripTrailingPunctuation(meaning)}?`;
}
function collectObservationNodes(graph) {
return (graph?.nodes || []).filter(
(node) => node.kind === "observation" && node.status === "supported",
);
}
function analyseObservationText(text) {
const normalised = normaliseText(text);
return {
text,
normalised,
isMeasurementLike:
/\b(increase|increased|decrease|decreased|fell|rose|doubled|halved|remained|average|score|scores|rate|time|traffic|sales|output|defects|complaints|production|revenue|cash|temperature|quality)\b/.test(
normalised,
) || /%|percent/.test(String(text || "")),
timeframeMentioned:
/\b(period|timeframe|quarter|month|week|year|day|annual|daily|weekly|monthly|same period)\b/.test(
normalised,
),
scaleMentioned: /\b(average|rate|score|scores|per|percent|%)\b/.test(
normalised,
),
unitMentioned:
/\b(celsius|fahrenheit|minutes|minute|hours|hour|days|day|units|sales|traffic|cash|revenue|complaints|defects)\b/.test(
normalised,
),
};
}
export const COMPARABILITY_REASONING_NODE_ID = "reasoning:comparability";
function readStoredComparabilityState(graph) {
const reasoningState = graph?.reasoningState;
if (!reasoningState?.comparabilityStatus) return null;
return {
comparabilityStatus: reasoningState.comparabilityStatus,
reason:
reasoningState.comparabilityReason ||
"Comparability state was carried forward from earlier reasoning.",
contradictionReasoningAllowed:
reasoningState.comparabilityStatus === "confirmed",
};
}
export function assessComparability(graph) {
const storedState = readStoredComparabilityState(graph);
if (storedState) {
return storedState;
}
const observations = collectObservationNodes(graph);
const centralText = normaliseText(graph?.centralStatement || "");
const profiles = observations.map((node) =>
analyseObservationText(`${node.label} ${node.description}`),
);
if (profiles.length < 2) {
return {
comparabilityStatus: "confirmed",
reason: "Fewer than two supported observations need comparison.",
contradictionReasoningAllowed: true,
};
}
if (
profiles.every((profile) => profile.normalised === profiles[0].normalised)
) {
return {
comparabilityStatus: "confirmed",
reason: "The observations restate the same measurement.",
contradictionReasoningAllowed: false,
};
}
if (profiles.some((profile) => !profile.isMeasurementLike)) {
return {
comparabilityStatus: "confirmed",
reason: "The observations are not competing like-for-like measurements.",
contradictionReasoningAllowed: true,
};
}
const hasExplicitTimeframe =
/\b(period|timeframe|quarter|month|week|year|day|same period)\b/.test(
centralText,
) || profiles.every((profile) => profile.timeframeMentioned);
const hasSharedScale = profiles.every((profile) => profile.scaleMentioned);
const hasSharedUnits = profiles.every((profile) => profile.unitMentioned);
if (!hasExplicitTimeframe || !hasSharedScale || !hasSharedUnits) {
return {
comparabilityStatus: "uncertain",
reason:
"Comparability between the observations is not yet established across period, scale, or measurement basis.",
contradictionReasoningAllowed: false,
};
}
return {
comparabilityStatus: "uncertain",
reason:
"The observations appear comparable in form, but the basis for comparing them is still not established.",
contradictionReasoningAllowed: false,
};
}
function buildReasoningStages(comparability, relationship, deferred = false) {
return [
{
stage: "comparability",
status: comparability.comparabilityStatus,
outcome: comparability.reason,
},
{
stage: "relationship",
status: relationship.relationshipStatus,
outcome: deferred
? "not assessed until comparability is established"
: relationship.reason,
},
];
}
function extractObservationConcepts(profile) {
const concepts = new Set();
const text = profile.normalised;
const conceptPatterns = [
["sales", /\bsales\b/],
["revenue", /\brevenue\b/],
["cash", /\bcash\b/],
["complaints", /\bcomplaints?\b/],
["production", /\bproduction\b/],
["delivery_time", /\bdelivery time\b|\baverage delivery time\b/],
["cancellations", /\bcancellations?\b/],
["satisfaction", /\bsatisfaction\b/],
["temperature", /\btemperature\b/],
["ice", /\bice\b/],
["traffic", /\btraffic\b/],
["defects", /\bdefects?\b/],
["quality", /\bquality\b/],
["staffing", /\bstaff(ing)?\b/],
["availability", /\bavailable|availability|unavailable\b/],
["service", /\bservice\b/],
];
for (const [name, pattern] of conceptPatterns) {
if (pattern.test(text)) concepts.add(name);
}
return [...concepts];
}
function extractObservationDirection(profile) {
const text = profile.normalised;
if (/\bunavailable\b/.test(text)) return "unavailable";
if (/\b(increase|increased|rose|up|doubled)\b/.test(text)) return "up";
if (/\b(decrease|decreased|fell|down|halved)\b/.test(text)) return "down";
if (/\b(remained unchanged|unchanged|same)\b/.test(text)) return "flat";
if (/\bavailable\b/.test(text)) return "available";
if (/\bmelted\b/.test(text)) return "melted";
return "unknown";
}
function classifyObservationRelationshipWhenComparable(graph) {
const observations = collectObservationNodes(graph);
const profiles = observations.map((node) =>
analyseObservationText(`${node.label} ${node.description}`),
);
if (profiles.length < 2) {
return {
relationshipStatus: "insufficient_information",
reason:
"Fewer than two supported observations are available for comparison.",
contradictionReasoningAllowed: false,
questionRequired: false,
questionSuppressedReason:
"Not enough observations to classify a relationship.",
};
}
if (
profiles.every((profile) => profile.normalised === profiles[0].normalised)
) {
return {
relationshipStatus: "duplicate",
reason: "The observations repeat the same measurement and direction.",
contradictionReasoningAllowed: false,
questionRequired: false,
questionSuppressedReason:
"Duplicate observations do not justify a follow-up question.",
};
}
const conceptSets = profiles.map((profile) =>
extractObservationConcepts(profile),
);
const sharedConcepts = conceptSets.reduce((shared, concepts, index) => {
if (index === 0) return new Set(concepts);
return new Set(concepts.filter((concept) => shared.has(concept)));
}, new Set());
const directions = profiles.map((profile) =>
extractObservationDirection(profile),
);
const conceptUnion = new Set(conceptSets.flat());
const hasRevenueCashPair =
conceptUnion.has("revenue") && conceptUnion.has("cash");
if (
sharedConcepts.size > 0 &&
directions.includes("available") &&
directions.includes("unavailable")
) {
return {
relationshipStatus: "contradictory",
reason:
"The observations assert mutually incompatible states about the same subject.",
contradictionReasoningAllowed: true,
questionRequired: true,
};
}
if (
sharedConcepts.size > 0 &&
directions.every((direction) => direction !== "unknown")
) {
return {
relationshipStatus: "potentially_related",
reason:
"The observations concern the same subject but do not assert a direct contradiction.",
contradictionReasoningAllowed: false,
questionRequired: true,
};
}
if (
hasRevenueCashPair &&
directions.every((direction) => direction !== "unknown")
) {
return {
relationshipStatus: "potentially_related",
reason:
"The observations concern connected business signals but do not establish a direct contradiction or cause.",
contradictionReasoningAllowed: false,
questionRequired: true,
};
}
if (
sharedConcepts.size === 0 &&
directions.every((direction) => direction !== "unknown")
) {
return {
relationshipStatus: "compatible",
reason:
"The observations can coexist without asserting incompatible states about the same subject.",
contradictionReasoningAllowed: false,
questionRequired: false,
questionSuppressedReason:
"Compatible observations do not justify a contradiction investigation.",
};
}
return {
relationshipStatus: "insufficient_information",
reason:
"There is not enough structure to classify the relationship safely.",
contradictionReasoningAllowed: false,
questionRequired: true,
};
}
export function classifyObservationRelationship(graph) {
const comparability = assessComparability(graph);
if (comparability.comparabilityStatus !== "confirmed") {
const deferredRelationship = {
relationshipStatus: "insufficient_information",
reason:
"Relationship classification is deferred until comparability is established.",
contradictionReasoningAllowed: false,
questionRequired: comparability.comparabilityStatus === "uncertain",
questionSuppressedReason:
comparability.comparabilityStatus === "incompatible"
? "Relationship classification was not attempted because the observations are not yet comparable."
: undefined,
relationshipAssessed: false,
};
return {
...deferredRelationship,
reasoningStages: buildReasoningStages(
comparability,
deferredRelationship,
true,
),
};
}
const classified = classifyObservationRelationshipWhenComparable(graph);
return {
...classified,
relationshipAssessed: true,
reasoningStages: buildReasoningStages(comparability, classified, false),
};
}
export function buildReasoningState(graph, overrides = {}) {
const relationship = classifyObservationRelationship({
...graph,
reasoningState: {
...(graph?.reasoningState || {}),
...(overrides || {}),
},
});
return {
comparabilityStatus: relationship.reasoningStages[0]?.status ?? null,
comparabilityReason: relationship.reasoningStages[0]?.outcome ?? null,
comparabilityEvidence:
overrides.comparabilityEvidence ??
graph?.reasoningState?.comparabilityEvidence ??
[],
relationshipStatus: relationship.relationshipStatus,
relationshipReason: relationship.reason,
relationshipAssessed: relationship.relationshipAssessed,
contradictionReasoningAllowed: relationship.contradictionReasoningAllowed,
reasoningStages: relationship.reasoningStages,
};
}
function buildComparabilityQuestion(graph, assessment) {
const centralText = normaliseText(graph?.centralStatement || "");
const mentionsPeriod =
/\b(period|timeframe|quarter|month|week|year|day)\b/.test(centralText);
if (mentionsPeriod) {
return "Were these figures measured on the same basis and at the same scale?";
}
return "Were these figures measured over the same period and at the same scale?";
}
function detectContradictionContext(graph) {
const central = stripTrailingPunctuation(
graph?.centralStatement || "this situation",
);
const contradictionNode = (graph?.nodes || []).find((node) => {
const text = normaliseText(`${node.label} ${node.description}`);
return (
node.kind === "relationship" &&
/\b(contradiction|conflict|inconsistent|mismatch|divergent|opposing)\b/.test(
text,
)
);
});
return {
centralStatement: central,
contradictionLabel: stripTrailingPunctuation(
contradictionNode?.label || "",
),
};
}
function buildBroadInvestigationQuestion(graph) {
const central = sanitizeQuestionText(
stripTrailingPunctuation(graph?.centralStatement || "these observations"),
);
return `What changed during that period that could help explain why ${central}?`;
}
function isRelationshipExplanationUnknown(node, graph) {
const text = normaliseText(`${node?.label || ""} ${node?.description || ""}`);
return (
collectObservationNodes(graph).length >= 2 &&
/\b(explain|explanation|divergence|moved differently|difference between|change or event|what changed|why the observations)/.test(
text,
)
);
}
function isBroadCompositeUnknownText(text) {
return /\b(possible causes|possible reasons|root causes|causes of|drivers of|factors behind|factors affecting|what changed|explanation for why|why .* but|difference between|divergence|moved differently|broad explanation|independent dimensions)\b/.test(
text,
);
}
function hasCompoundAbstractSignals(text) {
return (
/\b(timing|measurement|basis|cost|costs|debt|stock|tax|capital spending|mix|segment)\s+(and|or)\s+\b/.test(
text,
) ||
/\b[a-z]+\/[a-z]+\b/.test(text) ||
/,\s*[a-z]+,\s*[a-z]+/.test(text)
);
}
function isFocusedAtomicUnknownText(text) {
return /\b(define|definition|meaning|term|threshold|criterion|criteria|baseline|evidence|measure|metric|denominator|rate|date|period|budget|constraint|customer|actor|owner)\b/.test(
text,
);
}
function isDirectlyAnswerableObservationChildText(text) {
return /\b(whether the two observations reflect different timing|how the two observations were measured|change mainly affecting|one off event during the period|mix shift during the period)\b/.test(
text,
);
}
export function assessUnknownAtomicity({ node, graph }) {
const nodeText = normaliseText(
`${node?.label || ""} ${node?.description || ""}`,
);
if (
isDirectlyAnswerableObservationChildText(nodeText) &&
!hasCompoundAbstractSignals(nodeText)
) {
return {
atomicity: "atomic",
reason:
"This unknown isolates one specific line of enquiry and can be investigated directly.",
decompositionKind: null,
};
}
if (isRelationshipExplanationUnknown(node, graph)) {
return {
atomicity: "composite",
reason:
"This unknown asks for a broad explanation across multiple observations, so it should be decomposed before asking a direct question.",
decompositionKind: "relationship_explanation",
};
}
if (hasCompoundAbstractSignals(nodeText)) {
return {
atomicity: "composite",
reason:
"This unknown still bundles multiple abstract uncertainties together, so it should be decomposed before asking it directly.",
decompositionKind: "compound_child",
};
}
if (
isFocusedAtomicUnknownText(nodeText) &&
!isBroadCompositeUnknownText(nodeText)
) {
return {
atomicity: "atomic",
reason:
"This unknown already targets a single concrete detail that can be investigated directly.",
decompositionKind: null,
};
}
if (isBroadCompositeUnknownText(nodeText)) {
return {
atomicity: "composite",
reason:
"This unknown combines multiple broad candidate explanations, so it should be split into smaller dimensions first.",
decompositionKind: "broad_explanation",
};
}
return {
atomicity: "atomic",
reason:
"No deterministic composite pattern was detected, so the unknown can be investigated directly.",
decompositionKind: null,
};
}
export function formulateTieResolutionQuestion({ graph }) {
const comparability = assessComparability(graph);
if (comparability.comparabilityStatus === "uncertain") {
const deferredRelationship = classifyObservationRelationship(graph);
return {
question: buildComparabilityQuestion(graph, comparability),
reason:
"Formulated to confirm whether the observations are comparable before exploring competing explanations.",
strategy: null,
investigationStrategy: null,
selectionStatus: "ambiguous",
comparabilityStatus: comparability.comparabilityStatus,
comparabilityReason: comparability.reason,
contradictionReasoningAllowed:
comparability.contradictionReasoningAllowed,
relationshipStatus: deferredRelationship.relationshipStatus,
relationshipReason: deferredRelationship.reason,
relationshipAssessed: deferredRelationship.relationshipAssessed,
questionRequired: true,
reasoningStages: deferredRelationship.reasoningStages,
};
}
const relationship = classifyObservationRelationship(graph);
if (!relationship.questionRequired) {
return {
question: null,
reason: relationship.reason,
strategy: null,
investigationStrategy: null,
selectionStatus: "ambiguous",
comparabilityStatus: comparability.comparabilityStatus,
comparabilityReason: comparability.reason,
relationshipStatus: relationship.relationshipStatus,
relationshipReason: relationship.reason,
relationshipAssessed: relationship.relationshipAssessed,
contradictionReasoningAllowed: relationship.contradictionReasoningAllowed,
questionRequired: relationship.questionRequired,
questionSuppressedReason: relationship.questionSuppressedReason,
reasoningStages: relationship.reasoningStages,
};
}
if (relationship.relationshipStatus === "potentially_related") {
return {
question: buildBroadInvestigationQuestion(graph),
reason:
"Formulated as a neutral relationship question because the observations may be related without being contradictory.",
strategy: null,
investigationStrategy: null,
selectionStatus: "ambiguous",
comparabilityStatus: comparability.comparabilityStatus,
comparabilityReason: comparability.reason,
relationshipStatus: relationship.relationshipStatus,
relationshipReason: relationship.reason,
relationshipAssessed: relationship.relationshipAssessed,
contradictionReasoningAllowed: relationship.contradictionReasoningAllowed,
questionRequired: relationship.questionRequired,
reasoningStages: relationship.reasoningStages,
};
}
const { centralStatement, contradictionLabel } =
detectContradictionContext(graph);
const focus =
centralStatement || contradictionLabel || "these conflicting signals";
const question = sanitizeQuestionText(
`What changed during the period that could explain why ${focus}?`,
);
return {
question,
reason:
"Formulated to distinguish between tied unresolved explanations without prematurely choosing one branch.",
strategy: null,
investigationStrategy: null,
selectionStatus: "ambiguous",
comparabilityStatus: comparability.comparabilityStatus,
comparabilityReason: comparability.reason,
relationshipStatus: relationship.relationshipStatus,
relationshipReason: relationship.reason,
relationshipAssessed: relationship.relationshipAssessed,
contradictionReasoningAllowed: relationship.contradictionReasoningAllowed,
questionRequired: relationship.questionRequired,
reasoningStages: relationship.reasoningStages,
};
}
function extractActionPhrase(texts) {
for (const text of texts) {
const value = String(text || "").trim();
@@ -139,7 +748,41 @@ function toGerundPhrase(phrase) {
return [gerund, ...rest].join(" ");
}
function detectStrategy({ node, graph, relatedNodes, combinedText, meaning }) {
function buildInvestigationStrategy({
key,
reason,
node,
graph,
relatedNodes,
meaning,
actionPhrase,
}) {
return {
key,
reason,
nodeId: node?.id ?? null,
nodeLabel: node?.label ?? null,
meaning,
actionPhrase,
relatedNodeIds: relatedNodes.map((relatedNode) => relatedNode.id),
centralStatement: graph?.centralStatement ?? null,
};
}
export function selectInvestigationStrategy({ node, graph, context = {} }) {
const relatedNodes = collectRelatedNodes(node, graph);
const meaning = extractMeaning(node);
const combinedText = [
node?.label,
node?.description,
...relatedNodes.map((relatedNode) => relatedNode.label),
...relatedNodes.map((relatedNode) => relatedNode.description),
graph?.centralStatement,
...(context.resolvedValues || []),
]
.filter(Boolean)
.join(" ");
const text = normaliseText(combinedText);
const nodeText = normaliseText(
`${node?.label || ""} ${node?.description || ""}`,
@@ -172,25 +815,25 @@ function detectStrategy({ node, graph, relatedNodes, combinedText, meaning }) {
nodeText,
);
if (/\b(customer|user|buyer|stakeholder|recipient|audience)\b/.test(text)) {
return { strategy: "actor/customer", meaning, actionPhrase };
}
const hasBaselineLanguage =
/\b(before|previous|baseline|prior|comparable state)\b/.test(text);
const hasPrimaryBaselineLanguage =
/\b(before|previous|baseline|prior|comparable state)\b/.test(nodeText);
if (hasBaselineLanguage && hasPrimaryBaselineLanguage) {
return { strategy: "baseline", meaning, actionPhrase };
return buildInvestigationStrategy({
key: "baseline_reconstruction",
reason:
"Selected because the unknown explicitly references a missing previous or baseline state.",
node,
graph,
relatedNodes,
meaning,
actionPhrase,
});
}
if (/\b(when|timing|timeline|duration|sequence|milestone)\b/.test(text)) {
return { strategy: "transition/timing", meaning, actionPhrase };
}
const hasDefinitionLanguage =
/\b(define|definition|meaning|term|terminology)\b/.test(text);
const hasDefinitionLanguage = isDefinitionLikeUnknown(nodeText, text);
const hasPrimaryDefinitionLanguage =
/\b(define|definition|meaning|term|terminology)\b/.test(nodeText);
const hasCriteriaLanguage =
@@ -206,84 +849,105 @@ function detectStrategy({ node, graph, relatedNodes, combinedText, meaning }) {
/\b(metric|measure|measurable|roi|revenue projection|benchmark)\b/.test(
text,
);
const hasEvidenceLanguage =
/\b(evidence|proof|validate|validation|signal|demand)\b/.test(text) ||
isClaimLikeUnknown(node, text);
const hasContradictionLanguage =
/\b(contradiction|contradict|conflict|inconsistent|inconsistency|disagree|mismatch)\b/.test(
`${text} ${relatedText}`,
) ||
relatedNodes.some(
(relatedNode) =>
relatedNode.status === "contradicted" ||
relatedNode.kind === "conclusion",
);
if (hasDecisionValueLanguage && hasMeasurementLanguage) {
return { strategy: "measurement", meaning, actionPhrase };
}
if (hasPrimaryDefinitionLanguage) {
return { strategy: "definition", meaning, actionPhrase };
if (hasPrimaryDefinitionLanguage || hasDefinitionLanguage) {
return buildInvestigationStrategy({
key: "definition",
reason:
"Selected because the unknown is primarily about clarifying what a term means in this case.",
node,
graph,
relatedNodes,
meaning,
actionPhrase,
});
}
if (hasDecisionValueLanguage || hasCriteriaLanguage) {
return { strategy: "decision criterion", meaning, actionPhrase };
return buildInvestigationStrategy({
key: "decision_threshold",
reason:
"Selected because the unknown determines the threshold for making or justifying a decision.",
node,
graph,
relatedNodes,
meaning,
actionPhrase,
});
}
if (hasConstraintLanguage && hasPrimaryConstraintLanguage) {
return { strategy: "constraint", meaning, actionPhrase };
if (hasPrimaryBaselineLanguage || hasBaselineLanguage) {
return buildInvestigationStrategy({
key: "baseline_reconstruction",
reason:
"Selected because reconstructing the prior state is the most direct way to resolve the unknown.",
node,
graph,
relatedNodes,
meaning,
actionPhrase,
});
}
if (hasDefinitionLanguage) {
return { strategy: "definition", meaning, actionPhrase };
if (hasContradictionLanguage) {
return buildInvestigationStrategy({
key: "contradiction_resolution",
reason:
"Selected because the graph context indicates conflicting claims or inconsistent states that must be reconciled.",
node,
graph,
relatedNodes,
meaning,
actionPhrase,
});
}
if (hasBaselineLanguage) {
return { strategy: "baseline", meaning, actionPhrase };
if (hasEvidenceLanguage || hasMeasurementLanguage || hasConstraintLanguage) {
return buildInvestigationStrategy({
key: "evidence_gathering",
reason:
hasConstraintLanguage && hasPrimaryConstraintLanguage
? "Selected because evidence about the practical limiting factor is needed before the unknown can be resolved."
: "Selected because resolving the unknown requires evidence, signals, or measurable confirmation.",
node,
graph,
relatedNodes,
meaning,
actionPhrase,
});
}
if (hasConstraintLanguage) {
return { strategy: "constraint", meaning, actionPhrase };
}
if (/\b(evidence|proof|validate|validation|signal|demand)\b/.test(text)) {
return { strategy: "evidence", meaning, actionPhrase };
}
if (hasMeasurementLanguage) {
return { strategy: "measurement", meaning, actionPhrase };
}
if (
/\b(objective|goal|outcome|problem|job to be done|benefit)\b/.test(text)
) {
return { strategy: "objective", meaning, actionPhrase };
}
if (
node?.kind === "reported_claim" ||
node?.kind === "conclusion" ||
/\b(claim|assertion|true|false)\b/.test(text)
) {
return { strategy: "evidence", meaning, actionPhrase };
}
return { strategy: "generic clarification", meaning, actionPhrase };
return null;
}
function buildQuestion({ strategy, meaning, actionPhrase }) {
switch (strategy) {
case "decision criterion":
return actionPhrase
? `What outcome would demonstrate enough value to justify ${toGerundPhrase(actionPhrase)}?`
function buildQuestionFromStrategy(strategy) {
switch (strategy.key) {
case "decision_threshold":
return strategy.actionPhrase
? `What outcome would demonstrate enough value to justify ${toGerundPhrase(strategy.actionPhrase)}?`
: "What outcome would be sufficient to justify this decision?";
case "definition":
return `What does ${meaning} mean in this situation?`;
case "evidence":
return `What evidence would show whether ${meaning} is true?`;
case "baseline":
return `What was the comparable state before ${meaning}?`;
case "actor/customer":
return "Who experiences the problem or receives the value in this situation?";
case "objective":
return "What outcome is this decision or effort meant to achieve?";
case "constraint":
return "What constraint most limits the available options in this situation?";
case "measurement":
return `What measure would determine whether ${meaning} is sufficient?`;
case "transition/timing":
return `When does ${meaning} become relevant in the decision or change?`;
return `What does ${strategy.meaning} mean in this situation?`;
case "evidence_gathering":
return `What evidence would clarify ${stripTrailingPunctuation(strategy.meaning)}?`;
case "baseline_reconstruction":
return `What was the comparable state before ${strategy.meaning}?`;
case "contradiction_resolution":
return `What fact would resolve the contradiction about ${strategy.meaning}?`;
default:
return `What specific fact would resolve whether ${meaning} is true?`;
return `What specific fact would resolve whether ${strategy.meaning} is true?`;
}
}
@@ -315,6 +979,9 @@ function validateFormulatedQuestion(question, meaning) {
if (/^how should uncertainty regarding\b/i.test(trimmed)) return false;
if (/^what would resolve uncertainty regarding\b/i.test(trimmed))
return false;
if (/\)\.\s+[A-Z]/.test(trimmed)) return false;
if (/\bis true\?$/i.test(trimmed) && !/^whether\b/i.test(meaning))
return false;
if (
/\bprice|pricing|price point\b/i.test(trimmed) &&
!/\bprice\b/i.test(meaning)
@@ -332,36 +999,57 @@ function validateFormulatedQuestion(question, meaning) {
}
export function formulateQuestion({ node, graph, context = {} }) {
const relatedNodes = collectRelatedNodes(node, graph);
const meaning = extractMeaning(node);
const combinedText = [
node?.label,
node?.description,
...relatedNodes.map((relatedNode) => relatedNode.label),
...relatedNodes.map((relatedNode) => relatedNode.description),
graph?.centralStatement,
...(context.resolvedValues || []),
]
.filter(Boolean)
.join(" ");
if (context.selectionState?.status === "ambiguous") {
return formulateTieResolutionQuestion({ graph });
}
const detected = detectStrategy({
const investigationStrategy = selectInvestigationStrategy({
node,
graph,
relatedNodes,
combinedText,
meaning,
context,
});
let question = buildQuestion(detected);
let question = investigationStrategy
? buildQuestionFromStrategy(investigationStrategy)
: isRelationshipExplanationUnknown(node, graph)
? buildBroadInvestigationQuestion(graph)
: buildNeutralClarificationQuestion(extractMeaning(node));
if (!validateFormulatedQuestion(question, meaning)) {
question = `What evidence would resolve whether ${meaning} is true?`;
question = sanitizeQuestionText(question);
const fallbackMeaning = extractMeaning(node);
if (
!validateFormulatedQuestion(
question,
investigationStrategy?.meaning || fallbackMeaning,
)
) {
question = sanitizeQuestionText(
investigationStrategy &&
isClaimLikeUnknown(
node,
normaliseText(
collectRelatedNodes(node, graph)
.map(
(relatedNode) =>
`${relatedNode.label} ${relatedNode.description}`,
)
.concat([node?.label, node?.description])
.filter(Boolean)
.join(" "),
),
)
? buildEvidenceFallbackQuestion(fallbackMeaning)
: buildNeutralClarificationQuestion(fallbackMeaning),
);
}
return {
question,
reason: `Formulated from graph context using the ${detected.strategy} strategy.`,
strategy: detected.strategy,
reason: investigationStrategy
? `Formulated from graph context using the ${investigationStrategy.key} investigation strategy.`
: "Formulated as a neutral clarification question because no narrower investigation strategy clearly applied.",
strategy: investigationStrategy?.key ?? null,
investigationStrategy,
};
}
+37
View File
@@ -36,6 +36,20 @@ export const ConfidenceLevel = /** @type {const} */ ({
high: "high",
});
export const CompletenessStatus = /** @type {const} */ ({
empty: "empty",
partial: "partial",
complete: "complete",
});
export const confidenceAssessmentSchema = z
.object({
evidenceConfidence: z.enum(Object.values(ConfidenceLevel)),
completenessStatus: z.enum(Object.values(CompletenessStatus)),
conclusionConfidence: z.enum(Object.values(ConfidenceLevel)),
})
.strict();
// ── SituationNode ────────────────────────────────────
export const situationNodeSchema = z.object({
@@ -45,6 +59,7 @@ export const situationNodeSchema = z.object({
kind: z.enum(Object.values(SituationKind)),
status: z.enum(Object.values(SituationStatus)),
confidence: z.enum(Object.values(ConfidenceLevel)),
confidenceAssessment: confidenceAssessmentSchema.optional(),
value: z.union([z.string(), z.number(), z.null()]).nullable().optional(),
unit: z.string().nullable().optional(),
evidenceIds: z.array(z.string()).default([]),
@@ -84,6 +99,25 @@ export const situationEdgeSchema = z.object({
// ── SituationGraph ───────────────────────────────────
const reasoningStageSchema = z.object({
stage: z.string().min(1),
status: z.string().min(1),
outcome: z.string().min(1),
});
export const reasoningStateSchema = z
.object({
comparabilityStatus: z.string().min(1).nullable().optional(),
comparabilityReason: z.string().min(1).nullable().optional(),
comparabilityEvidence: z.array(z.string()).default([]),
relationshipStatus: z.string().min(1).nullable().optional(),
relationshipReason: z.string().min(1).nullable().optional(),
relationshipAssessed: z.boolean().optional(),
contradictionReasoningAllowed: z.boolean().optional(),
reasoningStages: z.array(reasoningStageSchema).default([]),
})
.strict();
export const situationGraphSchema = z.object({
centralStatement: z.string().min(1),
nodes: z.array(situationNodeSchema).min(1),
@@ -91,6 +125,7 @@ export const situationGraphSchema = z.object({
activeUnknownNodeId: z.string().nullable(),
resolvedNodeIds: z.array(z.string()).default([]),
currentSummary: z.string().min(1),
reasoningState: reasoningStateSchema.optional(),
});
/** @typedef {z.infer<typeof situationGraphSchema>} SituationGraph */
@@ -168,6 +203,7 @@ export function makeNode(opts) {
kind: opts.kind ?? "observation",
status: opts.status ?? "unknown",
confidence: opts.confidence ?? "medium",
confidenceAssessment: opts.confidenceAssessment,
value: opts.value ?? null,
unit: opts.unit ?? null,
evidenceIds: opts.evidenceIds ?? [],
@@ -201,5 +237,6 @@ export function makeGraph(opts) {
activeUnknownNodeId: opts.activeUnknownNodeId ?? null,
resolvedNodeIds: opts.resolvedNodeIds ?? [],
currentSummary: opts.currentSummary || "",
reasoningState: opts.reasoningState,
});
}
+376 -35
View File
@@ -95,6 +95,240 @@ function classifyUnknownPriority(text) {
return matches;
}
function buildScoreContributions(
matches,
downstreamCount,
unresolvedParentUnknownCount,
) {
const contributions = [
{
rule: "downstream_dependencies",
value: downstreamCount,
weight: 4,
delta: downstreamCount * 4,
},
];
if (matches.objective) {
contributions.push({
rule: "objective_match",
value: true,
weight: 12,
delta: 12,
});
}
if (matches.actor) {
contributions.push({
rule: "actor_match",
value: true,
weight: 10,
delta: 10,
});
}
if (matches.criteria) {
contributions.push({
rule: "criteria_match",
value: true,
weight: 11,
delta: 11,
});
}
if (matches.measure) {
contributions.push({
rule: "measure_match",
value: true,
weight: 8,
delta: 8,
});
}
if (matches.terminology) {
contributions.push({
rule: "terminology_match",
value: true,
weight: 7,
delta: 7,
});
}
if (matches.constraint) {
contributions.push({
rule: "constraint_match",
value: true,
weight: 9,
delta: 9,
});
}
if (matches.pricing) {
contributions.push({
rule: "pricing_penalty",
value: true,
weight: -8,
delta: -8,
});
}
if (matches.implementation) {
contributions.push({
rule: "implementation_penalty",
value: true,
weight: -10,
delta: -10,
});
}
if (matches.optimisation) {
contributions.push({
rule: "optimisation_penalty",
value: true,
weight: -9,
delta: -9,
});
}
if (matches.speculative) {
contributions.push({
rule: "speculative_penalty",
value: true,
weight: -12,
delta: -12,
});
}
if (
matches.pricing &&
!matches.objective &&
!matches.criteria &&
!matches.actor
) {
contributions.push({
rule: "isolated_pricing_penalty",
value: true,
weight: -6,
delta: -6,
});
}
if (unresolvedParentUnknownCount > 0) {
contributions.push({
rule: "unresolved_prerequisite_penalty",
value: unresolvedParentUnknownCount,
weight: -7,
delta: unresolvedParentUnknownCount * -7,
});
}
return contributions;
}
function getMeaningfulSemanticContributions(contributions = []) {
return contributions
.filter(
(contribution) =>
contribution.rule !== "downstream_dependencies" &&
contribution.rule !== "unresolved_prerequisite_penalty" &&
contribution.delta !== 0,
)
.map((contribution) => ({
rule: contribution.rule,
delta: contribution.delta,
}));
}
function buildCandidateDisplayOrder(candidates) {
return [...candidates].sort((a, b) => {
if (b.score !== a.score) return b.score - a.score;
if (b.downstreamCount !== a.downstreamCount) {
return b.downstreamCount - a.downstreamCount;
}
if (a.unresolvedParentUnknownCount !== b.unresolvedParentUnknownCount) {
return a.unresolvedParentUnknownCount - b.unresolvedParentUnknownCount;
}
return a.label.localeCompare(b.label);
});
}
function semanticSignature(candidate) {
return JSON.stringify(
getMeaningfulSemanticContributions(candidate.contributions),
);
}
function classifyCandidateOrdering(candidates) {
const displayOrder = buildCandidateDisplayOrder(candidates);
const best = displayOrder[0] ?? null;
if (!best) {
return {
displayOrder,
best: null,
leadingCandidates: [],
status: "no_candidates",
tieType: "none",
usedAlphabeticalOrdering: false,
reason: "No unresolved unknown candidates remain.",
};
}
const topScoreCandidates = displayOrder.filter(
(candidate) => candidate.score === best.score,
);
if (topScoreCandidates.length === 1) {
return {
displayOrder,
best,
leadingCandidates: [best],
status: "selected",
tieType: "none",
usedAlphabeticalOrdering: false,
reason: `Clear winner by total score (${best.score}).`,
};
}
const topStructuralCandidates = topScoreCandidates.filter(
(candidate) =>
candidate.downstreamCount === best.downstreamCount &&
candidate.unresolvedParentUnknownCount ===
best.unresolvedParentUnknownCount,
);
if (topStructuralCandidates.length === 1) {
return {
displayOrder,
best,
leadingCandidates: [best],
status: "selected",
tieType: "structural_tie",
usedAlphabeticalOrdering: false,
reason:
"Score tie was resolved by downstream dependency count or prerequisite ordering.",
};
}
const topSemanticSignature = semanticSignature(best);
const semanticPeers = topStructuralCandidates.filter(
(candidate) => semanticSignature(candidate) === topSemanticSignature,
);
if (semanticPeers.length !== topStructuralCandidates.length) {
return {
displayOrder,
best: null,
leadingCandidates: topStructuralCandidates,
status: "ambiguous",
tieType: "semantic_tie",
usedAlphabeticalOrdering: false,
reason:
"Leading candidates remain tied after score and structural checks, but differ in semantic contribution patterns.",
};
}
return {
displayOrder,
best: null,
leadingCandidates: topStructuralCandidates,
status: "ambiguous",
tieType: "complete_unresolved_tie",
usedAlphabeticalOrdering: false,
reason: "No justified distinction between leading unknowns.",
};
}
export function scoreUnknownCandidate(graph, node, resolvedNodeIds = []) {
const text = collectNodeText(node);
const matches = classifyUnknownPriority(text);
@@ -105,30 +339,15 @@ export function scoreUnknownCandidate(graph, node, resolvedNodeIds = []) {
resolvedNodeIds,
);
let score = downstreamCount * 4;
if (matches.objective) score += 12;
if (matches.actor) score += 10;
if (matches.criteria) score += 11;
if (matches.measure) score += 8;
if (matches.terminology) score += 7;
if (matches.constraint) score += 9;
if (matches.pricing) score -= 8;
if (matches.implementation) score -= 10;
if (matches.optimisation) score -= 9;
if (matches.speculative) score -= 12;
if (
matches.pricing &&
!matches.objective &&
!matches.criteria &&
!matches.actor
) {
score -= 6;
}
score -= unresolvedParentUnknownCount * 7;
const contributions = buildScoreContributions(
matches,
downstreamCount,
unresolvedParentUnknownCount,
);
const score = contributions.reduce(
(total, contribution) => total + contribution.delta,
0,
);
return {
nodeId: node.id,
@@ -137,6 +356,7 @@ export function scoreUnknownCandidate(graph, node, resolvedNodeIds = []) {
downstreamCount,
unresolvedParentUnknownCount,
matches,
contributions,
};
}
@@ -383,21 +603,36 @@ export function selectActiveUnknownCandidate(graph, resolvedNodeIds) {
...scoreUnknownCandidate(graph, node, resolvedNodeIds),
}));
scoredCandidates.sort((a, b) => {
if (b.score !== a.score) return b.score - a.score;
if (b.downstreamCount !== a.downstreamCount) {
return b.downstreamCount - a.downstreamCount;
}
if (a.unresolvedParentUnknownCount !== b.unresolvedParentUnknownCount) {
return a.unresolvedParentUnknownCount - b.unresolvedParentUnknownCount;
}
return a.node.label.localeCompare(b.node.label);
});
const selection = classifyCandidateOrdering(
scoredCandidates.map(({ node, ...candidate }) => ({
...candidate,
node,
})),
);
const best = scoredCandidates[0];
if (selection.status === "ambiguous") {
return {
selectedNode: null,
status: "ambiguous",
tieType: selection.tieType,
tiedCandidateIds: selection.leadingCandidates.map(
(candidate) => candidate.nodeId,
),
displayOrder: selection.displayOrder.map((candidate) => candidate.nodeId),
reason: selection.reason,
};
}
const best = selection.best;
if (!best) return null;
return {
selectedNode: {
nodeId: best.node.id,
label: best.node.label,
},
status: "selected",
tieType: selection.tieType,
nodeId: best.node.id,
label: best.node.label,
score: best.score,
@@ -406,6 +641,112 @@ export function selectActiveUnknownCandidate(graph, resolvedNodeIds) {
};
}
export function explainUnknownSelection(graph, resolvedNodeIds = []) {
const unresolved = graph.nodes.filter(
(n) => n.kind === "unknown" && !resolvedNodeIds.includes(n.id),
);
if (unresolved.length === 0) {
return {
selectedNodeId: null,
selectedNodeLabel: null,
status: "no_candidates",
tieType: "none",
resolvedNodeIds: [...resolvedNodeIds],
tiedCandidateIds: [],
candidates: [],
competitors: [],
tieBreakOrder: [
"score_desc",
"downstreamCount_desc",
"unresolvedParentUnknownCount_asc",
"label_asc",
],
summary: {
candidateCount: 0,
},
};
}
const candidates = unresolved.map((node) => ({
nodeId: node.id,
label: node.label,
...scoreUnknownCandidate(graph, node, resolvedNodeIds),
}));
const selection = classifyCandidateOrdering(candidates);
const orderedCandidates = selection.displayOrder;
const selected = selection.best;
const competitors = orderedCandidates
.filter((candidate) => candidate.nodeId !== selected?.nodeId)
.map((candidate) => ({
nodeId: candidate.nodeId,
label: candidate.label,
score: candidate.score,
downstreamCount: candidate.downstreamCount,
unresolvedParentUnknownCount: candidate.unresolvedParentUnknownCount,
matches: candidate.matches,
contributions: candidate.contributions,
outrankedBy: {
scoreDelta: (selected?.score ?? candidate.score) - candidate.score,
downstreamDelta:
(selected?.downstreamCount ?? candidate.downstreamCount) -
candidate.downstreamCount,
unresolvedPrerequisiteDelta:
candidate.unresolvedParentUnknownCount -
(selected?.unresolvedParentUnknownCount ??
candidate.unresolvedParentUnknownCount),
labelOrderWinner:
selected &&
selected.score === candidate.score &&
selected.downstreamCount === candidate.downstreamCount &&
selected.unresolvedParentUnknownCount ===
candidate.unresolvedParentUnknownCount
? selected.label.localeCompare(candidate.label) <= 0
? selected.label
: candidate.label
: null,
},
}));
return {
selectedNodeId: selected?.nodeId ?? null,
selectedNodeLabel: selected?.label ?? null,
status: selection.status,
tieType: selection.tieType,
resolvedNodeIds: [...resolvedNodeIds],
tiedCandidateIds: selection.leadingCandidates.map(
(candidate) => candidate.nodeId,
),
tieBreakOrder: [
"score_desc",
"downstreamCount_desc",
"unresolvedParentUnknownCount_asc",
"label_asc",
],
alphabeticalUsedAsReasoning: false,
candidates: orderedCandidates,
selected: selected
? {
nodeId: selected.nodeId,
label: selected.label,
score: selected.score,
downstreamCount: selected.downstreamCount,
unresolvedParentUnknownCount: selected.unresolvedParentUnknownCount,
matches: selected.matches,
contributions: selected.contributions,
}
: null,
competitors,
summary: {
candidateCount: orderedCandidates.length,
selectedReason: selected
? `highest_score=${selected.score}; downstream=${selected.downstreamCount}; unresolved_prerequisites=${selected.unresolvedParentUnknownCount}`
: selection.reason,
},
};
}
// ── Apply a graph update deterministically ──
export function applyGraphUpdate(graph, update) {
+198
View File
@@ -0,0 +1,198 @@
import { makeEdge, makeGraph, makeNode } from "@/lib/graph/schema.js";
function buildAmbiguityFixture({
key,
scenario,
summaryLabel,
contradictionLabel,
observationLabels,
unknownLabels,
disallowedQuestionTerms,
}) {
const summary = makeNode({
id: `${key}-summary`,
label: summaryLabel,
description: "Summary of the situation from the scenario text",
kind: "state",
status: "provisional",
confidence: "medium",
});
const contradiction = makeNode({
id: `${key}-contradiction`,
label: contradictionLabel,
description: contradictionLabel,
kind: "relationship",
status: "supported",
confidence: "medium",
});
const observations = observationLabels.map((label, index) =>
makeNode({
id: `${key}-obs-${index + 1}`,
label,
description: label,
kind: "observation",
status: "supported",
confidence: "high",
}),
);
const unknowns = unknownLabels.map((label, index) =>
makeNode({
id: `${key}-unknown-${index + 1}`,
label,
description: label,
kind: "unknown",
status: "unknown",
confidence: "high",
}),
);
const edges = [
...observations.map((node) =>
makeEdge({
id: `${node.id}-supports-summary`,
fromNodeId: node.id,
toNodeId: summary.id,
relationship: "supports",
description: `${node.label} supports the summary.`,
}),
),
...unknowns.map((node) =>
makeEdge({
id: `${node.id}-depends-summary`,
fromNodeId: node.id,
toNodeId: summary.id,
relationship: "depends_on",
description: `${node.label} is an unresolved factor for this situation.`,
}),
),
];
return {
key,
scenario,
disallowedQuestionTerms,
graph: makeGraph({
centralStatement: scenario,
nodes: [summary, contradiction, ...observations, ...unknowns],
edges,
activeUnknownNodeId: null,
resolvedNodeIds: [],
currentSummary: `Ambiguity fixture for ${key}`,
}),
};
}
export const ambiguityGeneralisationFixtures = [
buildAmbiguityFixture({
key: "revenue-cash",
scenario:
"Revenue increased by 18%, but cash in the bank fell over the same period.",
summaryLabel: "Revenue rose while cash fell",
contradictionLabel:
"Contradiction between revenue improvement and lower cash reserves.",
observationLabels: [
"Revenue increased by 18%.",
"Cash in the bank decreased over the same period.",
],
unknownLabels: [
"Possible explanation for the contradiction from one side of the situation.",
"Possible explanation for the contradiction from another side of the situation.",
],
disallowedQuestionTerms: [
"accounts receivable",
"capex",
"debt repayments",
"working capital",
],
}),
buildAmbiguityFixture({
key: "satisfaction-complaints",
scenario:
"Customer satisfaction scores increased, but complaints also increased.",
summaryLabel: "Satisfaction scores rose while complaints also rose",
contradictionLabel:
"Contradiction between higher satisfaction scores and higher complaint volume.",
observationLabels: [
"Customer satisfaction scores increased.",
"Complaints increased.",
],
unknownLabels: [
"Possible explanation for why the positive signal and negative signal moved together.",
"Another possible explanation for why the positive signal and negative signal moved together.",
],
disallowedQuestionTerms: [
"net promoter",
"ticket backlog",
"call deflection",
"support queue",
],
}),
buildAmbiguityFixture({
key: "delivery-cancellations",
scenario:
"Average delivery time decreased by 25%, but order cancellations increased.",
summaryLabel: "Delivery became faster while cancellations increased",
contradictionLabel:
"Contradiction between faster delivery and more order cancellations.",
observationLabels: [
"Average delivery time decreased by 25%.",
"Order cancellations increased.",
],
unknownLabels: [
"Possible explanation for why the faster result did not reduce the negative result.",
"Another possible explanation for why the faster result did not reduce the negative result.",
],
disallowedQuestionTerms: [
"fulfilment",
"last mile",
"warehouse",
"routing",
],
}),
buildAmbiguityFixture({
key: "traffic-sales",
scenario: "Website traffic doubled, but sales remained unchanged.",
summaryLabel: "Website traffic doubled while sales stayed flat",
contradictionLabel:
"Contradiction between much higher traffic and unchanged sales.",
observationLabels: [
"Website traffic doubled.",
"Sales remained unchanged.",
],
unknownLabels: [
"Possible explanation for why the stronger signal did not change the outcome.",
"Another possible explanation for why the stronger signal did not change the outcome.",
],
disallowedQuestionTerms: [
"conversion funnel",
"campaign attribution",
"landing page",
"checkout flow",
],
}),
buildAmbiguityFixture({
key: "output-defects",
scenario:
"Production output increased by 30%, but quality defects also increased.",
summaryLabel: "Production output rose while defects also rose",
contradictionLabel:
"Contradiction between higher output and more quality defects.",
observationLabels: [
"Production output increased by 30%.",
"Quality defects increased.",
],
unknownLabels: [
"Possible explanation for why the gain came with a worsening result.",
"Another possible explanation for why the gain came with a worsening result.",
],
disallowedQuestionTerms: [
"scrap rate",
"throughput",
"yield",
"root cause",
],
}),
];
+164
View File
@@ -0,0 +1,164 @@
import { makeEdge, makeGraph, makeNode } from "@/lib/graph/schema.js";
function buildComparabilityFixture({
key,
scenario,
observationLabels,
contradictionLabel,
expectedComparabilityStatus,
expectsComparisonQuestion,
}) {
const summary = makeNode({
id: `${key}-summary`,
label: scenario,
description: "Summary of the situation from the scenario text",
kind: "state",
status: "provisional",
confidence: "medium",
});
const observations = observationLabels.map((label, index) =>
makeNode({
id: `${key}-obs-${index + 1}`,
label,
description: label,
kind: "observation",
status: "supported",
confidence: "high",
}),
);
const contradiction = contradictionLabel
? [
makeNode({
id: `${key}-contradiction`,
label: contradictionLabel,
description: contradictionLabel,
kind: "relationship",
status: "supported",
confidence: "medium",
}),
]
: [];
const unknowns = [
makeNode({
id: `${key}-unknown-a`,
label: "Possible explanation from one side of the situation.",
description: "Possible explanation from one side of the situation.",
kind: "unknown",
status: "unknown",
confidence: "high",
}),
makeNode({
id: `${key}-unknown-b`,
label: "Possible explanation from another side of the situation.",
description: "Possible explanation from another side of the situation.",
kind: "unknown",
status: "unknown",
confidence: "high",
}),
];
const edges = [
...observations.map((node) =>
makeEdge({
id: `${node.id}-supports-summary`,
fromNodeId: node.id,
toNodeId: summary.id,
relationship: "supports",
description: `${node.label} supports the summary.`,
}),
),
...unknowns.map((node) =>
makeEdge({
id: `${node.id}-depends-summary`,
fromNodeId: node.id,
toNodeId: summary.id,
relationship: "depends_on",
description: `${node.label} is an unresolved factor for this situation.`,
}),
),
];
return {
key,
scenario,
expectedComparabilityStatus,
expectsComparisonQuestion,
graph: makeGraph({
centralStatement: scenario,
nodes: [summary, ...observations, ...contradiction, ...unknowns],
edges,
activeUnknownNodeId: null,
resolvedNodeIds: [],
currentSummary: `Comparability fixture for ${key}`,
}),
};
}
export const comparabilityAssessmentFixtures = [
buildComparabilityFixture({
key: "revenue-cash",
scenario:
"Revenue increased by 18%, but cash in the bank fell over the same period.",
observationLabels: [
"Revenue increased by 18%.",
"Cash in the bank decreased over the same period.",
],
contradictionLabel:
"Contradiction between revenue improvement and lower cash reserves.",
expectedComparabilityStatus: "uncertain",
expectsComparisonQuestion: true,
}),
buildComparabilityFixture({
key: "complaints-production",
scenario: "Complaints increased. Production increased.",
observationLabels: ["Complaints increased.", "Production increased."],
contradictionLabel:
"Possible contradiction between complaints and production movement.",
expectedComparabilityStatus: "uncertain",
expectsComparisonQuestion: true,
}),
buildComparabilityFixture({
key: "delivery-cancellations",
scenario:
"Average delivery time decreased by 25%, but order cancellations increased.",
observationLabels: [
"Average delivery time decreased by 25%.",
"Order cancellations increased.",
],
contradictionLabel:
"Contradiction between faster delivery and more cancellations.",
expectedComparabilityStatus: "uncertain",
expectsComparisonQuestion: true,
}),
buildComparabilityFixture({
key: "satisfaction-complaints",
scenario: "Customer satisfaction increased, but complaints increased.",
observationLabels: [
"Customer satisfaction increased.",
"Complaints increased.",
],
contradictionLabel:
"Contradiction between satisfaction improvement and more complaints.",
expectedComparabilityStatus: "uncertain",
expectsComparisonQuestion: true,
}),
buildComparabilityFixture({
key: "temperature-ice",
scenario: "Temperature increased. Ice melted.",
observationLabels: ["Temperature increased.", "Ice melted."],
contradictionLabel: null,
expectedComparabilityStatus: "confirmed",
expectsComparisonQuestion: false,
}),
buildComparabilityFixture({
key: "sales-same",
scenario: "Sales doubled. Sales doubled.",
observationLabels: ["Sales doubled.", "Sales doubled."],
contradictionLabel: null,
expectedComparabilityStatus: "confirmed",
expectsComparisonQuestion: false,
}),
];
+17 -5
View File
@@ -67,7 +67,11 @@ export const questionPriorityGeneralisationFixtures = [
"hire-bottleneck",
],
prohibitedFirstTopics: ["salary", "job advert", "programming language"],
acceptableQuestionStrategies: ["decision criterion", "constraint"],
acceptableQuestionStrategies: [
"decision_threshold",
"evidence_gathering",
"definition",
],
notes:
"The first question should establish whether more engineering capacity is justified before compensation or implementation details.",
graph: makeScenarioGraph({
@@ -143,7 +147,11 @@ export const questionPriorityGeneralisationFixtures = [
"paint colour",
"finance provider",
],
acceptableQuestionStrategies: ["decision criterion", "constraint"],
acceptableQuestionStrategies: [
"decision_threshold",
"evidence_gathering",
"definition",
],
notes:
"The first question should establish whether the fleet is failing a threshold that justifies replacement.",
graph: makeScenarioGraph({
@@ -219,7 +227,7 @@ export const questionPriorityGeneralisationFixtures = [
"office location",
"advertising channel",
],
acceptableQuestionStrategies: ["actor/customer", "decision criterion"],
acceptableQuestionStrategies: ["definition", "decision_threshold"],
notes:
"The first question should clarify the customer or value case for expansion before rollout logistics.",
graph: makeScenarioGraph({
@@ -291,7 +299,7 @@ export const questionPriorityGeneralisationFixtures = [
"project-remaining-benefit",
],
prohibitedFirstTopics: ["sunk cost", "project logo", "final launch date"],
acceptableQuestionStrategies: ["decision criterion", "objective"],
acceptableQuestionStrategies: ["decision_threshold", "definition"],
notes:
"The first question should establish remaining value or success threshold before sunk-cost framing or launch timing.",
graph: makeScenarioGraph({
@@ -367,7 +375,11 @@ export const questionPriorityGeneralisationFixtures = [
"payment provider",
"tier name",
],
acceptableQuestionStrategies: ["actor/customer", "decision criterion"],
acceptableQuestionStrategies: [
"definition",
"decision_threshold",
"baseline_reconstruction",
],
notes:
"The first question should establish who values paid support or what outcome would justify offering it before pricing details.",
graph: makeScenarioGraph({
@@ -0,0 +1,131 @@
import { describe, expect, it } from "vitest";
import {
formulateQuestion,
formulateTieResolutionQuestion,
} from "@/lib/graph/question-formulator.js";
import {
explainUnknownSelection,
selectActiveUnknownCandidate,
} from "@/lib/graph/utils.js";
import { ambiguityGeneralisationFixtures } from "@/tests/fixtures/ambiguity-generalisation.js";
function neutraliseUnknownLabels(graph) {
let counter = 0;
return {
...graph,
nodes: graph.nodes.map((node) => {
if (node.kind !== "unknown") return { ...node };
counter += 1;
return {
...node,
label: `Unknown ${String.fromCharCode(64 + counter)}`,
description: `Unknown factor ${counter}.`,
};
}),
};
}
function isSingleQuestion(question) {
return (question.match(/\?/g) || []).length === 1;
}
describe("ambiguity generalisation", () => {
it("preserves ambiguity across contradiction scenarios without favouring one explanation", () => {
const summary = ambiguityGeneralisationFixtures.map((fixture) => {
const explanation = explainUnknownSelection(fixture.graph, []);
const selection = selectActiveUnknownCandidate(fixture.graph, []);
const neutralExplanation = explainUnknownSelection(
neutraliseUnknownLabels(fixture.graph),
[],
);
const tieQuestion = formulateTieResolutionQuestion({
graph: fixture.graph,
});
const representativeUnknown = fixture.graph.nodes.find(
(node) => node.kind === "unknown",
);
const fallbackQuestion = formulateQuestion({
node: representativeUnknown,
graph: fixture.graph,
});
const lowerQuestion = tieQuestion.question.toLowerCase();
for (const term of fixture.disallowedQuestionTerms) {
expect(lowerQuestion).not.toContain(term.toLowerCase());
}
expect(explanation.status).toBe("ambiguous");
expect(selection.status).toBe("ambiguous");
expect(selection.selectedNode).toBeNull();
expect(explanation.selectedNodeId).toBeNull();
expect(explanation.candidates).toHaveLength(2);
expect(explanation.summary.selectedReason).toBe(
"No justified distinction between leading unknowns.",
);
expect(explanation.alphabeticalUsedAsReasoning).toBe(false);
expect(neutralExplanation.status).toBe("ambiguous");
expect(isSingleQuestion(tieQuestion.question)).toBe(true);
expect(tieQuestion.question.toLowerCase()).not.toContain(" or ");
return {
scenario: fixture.scenario,
candidateCount: explanation.candidates.length,
ambiguityStatus: explanation.status,
tieReason: explanation.summary.selectedReason,
investigationStrategy: tieQuestion.strategy,
question: tieQuestion.question,
explanationFavoured: explanation.selectedNodeId !== null,
};
});
expect(summary).toMatchInlineSnapshot(`
[
{
"ambiguityStatus": "ambiguous",
"candidateCount": 2,
"explanationFavoured": false,
"investigationStrategy": null,
"question": "Were these figures measured on the same basis and at the same scale?",
"scenario": "Revenue increased by 18%, but cash in the bank fell over the same period.",
"tieReason": "No justified distinction between leading unknowns.",
},
{
"ambiguityStatus": "ambiguous",
"candidateCount": 2,
"explanationFavoured": false,
"investigationStrategy": null,
"question": "Were these figures measured over the same period and at the same scale?",
"scenario": "Customer satisfaction scores increased, but complaints also increased.",
"tieReason": "No justified distinction between leading unknowns.",
},
{
"ambiguityStatus": "ambiguous",
"candidateCount": 2,
"explanationFavoured": false,
"investigationStrategy": null,
"question": "Were these figures measured over the same period and at the same scale?",
"scenario": "Average delivery time decreased by 25%, but order cancellations increased.",
"tieReason": "No justified distinction between leading unknowns.",
},
{
"ambiguityStatus": "ambiguous",
"candidateCount": 2,
"explanationFavoured": false,
"investigationStrategy": null,
"question": "Were these figures measured over the same period and at the same scale?",
"scenario": "Website traffic doubled, but sales remained unchanged.",
"tieReason": "No justified distinction between leading unknowns.",
},
{
"ambiguityStatus": "ambiguous",
"candidateCount": 2,
"explanationFavoured": false,
"investigationStrategy": null,
"question": "Were these figures measured over the same period and at the same scale?",
"scenario": "Production output increased by 30%, but quality defects also increased.",
"tieReason": "No justified distinction between leading unknowns.",
},
]
`);
});
});
+321
View File
@@ -3,6 +3,120 @@ import { applyValidatedProposal } from "@/lib/graph/apply-proposal.js";
import { makeEdge, makeGraph, makeNode } from "@/lib/graph/schema.js";
import { validateGraphReferences } from "@/lib/graph/utils.js";
function makeComparabilityUpdateFixture() {
const comparabilityUnknown = makeNode({
id: "n-comparability-unknown",
label: "Whether the figures are comparable",
description:
"Need to know whether the figures use the same period, basis, and scale before comparing them.",
kind: "unknown",
status: "unknown",
confidence: "high",
});
const revenueObservation = makeNode({
id: "n-revenue-observation",
label: "Revenue increased by 18%.",
description: "Revenue increased by 18%.",
kind: "observation",
status: "supported",
confidence: "high",
});
const cashObservation = makeNode({
id: "n-cash-observation",
label: "Cash in the bank decreased over the same period.",
description: "Cash in the bank decreased over the same period.",
kind: "observation",
status: "supported",
confidence: "high",
});
const unrelatedNode = makeNode({
id: "n-unrelated",
label: "Board update",
description: "A separate unchanged note.",
kind: "state",
status: "known",
confidence: "low",
});
const graph = makeGraph({
centralStatement:
"Revenue increased by 18%, but cash in the bank fell over the same period.",
nodes: [
comparabilityUnknown,
revenueObservation,
cashObservation,
unrelatedNode,
],
edges: [
makeEdge({
id: "e-revenue-comparability",
fromNodeId: revenueObservation.id,
toNodeId: comparabilityUnknown.id,
relationship: "supports",
confidence: "medium",
description: "Revenue observation requires comparability confirmation.",
}),
makeEdge({
id: "e-cash-comparability",
fromNodeId: cashObservation.id,
toNodeId: comparabilityUnknown.id,
relationship: "supports",
confidence: "medium",
description: "Cash observation requires comparability confirmation.",
}),
],
activeUnknownNodeId: comparabilityUnknown.id,
resolvedNodeIds: [],
currentSummary: "Initial comparability fixture",
reasoningState: {
comparabilityStatus: "uncertain",
comparabilityReason:
"Comparability between the observations is not yet established across period, scale, or measurement basis.",
comparabilityEvidence: [],
relationshipStatus: "insufficient_information",
relationshipReason:
"Relationship classification is deferred until comparability is established.",
relationshipAssessed: false,
contradictionReasoningAllowed: false,
reasoningStages: [
{
stage: "comparability",
status: "uncertain",
outcome:
"Comparability between the observations is not yet established across period, scale, or measurement basis.",
},
{
stage: "relationship",
status: "insufficient_information",
outcome: "not assessed until comparability is established",
},
],
},
});
const proposal = {
addedNodes: [],
updatedNodes: [
{
nodeId: comparabilityUnknown.id,
previousStatus: "unknown",
newStatus: "resolved",
previousValue: null,
newValue:
"Both figures cover the same accounting period and are taken from the same management accounts.",
reason: "The answer confirms the figures are comparable.",
},
],
addedEdges: [],
removedEdgeIds: [],
resolvedUnknownNodeIds: [comparabilityUnknown.id],
affectedNodeIds: [],
selectedQuestion: null,
};
return { graph, proposal, comparabilityUnknownId: comparabilityUnknown.id };
}
function makeApplicationFixture() {
const complaintRateUnknown = makeNode({
id: "n-complaint-rate-unknown",
@@ -997,4 +1111,211 @@ describe("applyValidatedProposal", () => {
"price",
);
});
it("resolves the existing comparability unknown and advances reasoning after the answer", () => {
const { graph, proposal, comparabilityUnknownId } =
makeComparabilityUpdateFixture();
const originalUnrelatedNode = JSON.stringify(
graph.nodes.find((node) => node.id === "n-unrelated"),
);
const result = applyValidatedProposal({
situationGraph: graph,
proposal,
previousQuestion:
"Were these figures measured on the same basis and at the same scale?",
answer:
"Yes. Both figures cover the same accounting period and are taken from the same management accounts.",
});
expect(result.success).toBe(true);
expect(result.resolvedUnknownNodeIds).toContain(comparabilityUnknownId);
expect(result.resolvedReasoningNodeIds).toEqual([
"reasoning:comparability",
]);
expect(result.emergentReasoningNodeCreated).toBe(true);
expect(result.emergentReasoningNodeId).toBeTruthy();
expect(result.emergentReasoningNodeReason).toContain("backed by the graph");
expect(result.previousReasoningState.comparabilityStatus).toBe("uncertain");
expect(result.reasoningState).toMatchObject({
comparabilityStatus: "confirmed",
relationshipStatus: "potentially_related",
relationshipAssessed: true,
});
expect(result.reasoningState.comparabilityEvidence).toEqual([
comparabilityUnknownId,
]);
expect(result.selectedQuestion?.nodeId).toBe(result.newActiveUnknownNodeId);
expect(result.selectedQuestion).toMatchObject({
nodeId: result.newActiveUnknownNodeId,
question:
"What evidence would clarify how the two observations were measured?",
});
expect(result.selectedQuestion?.question.toLowerCase()).not.toMatch(
/dso|debtor days|receivables turnover|working capital|receivables/,
);
expect(result.reasoningState.reasoningStages).toEqual([
{
stage: "comparability",
status: "confirmed",
outcome:
"Comparability was confirmed by the user answer covering the same period and source basis.",
},
{
stage: "relationship",
status: "potentially_related",
outcome:
"The observations concern connected business signals but do not establish a direct contradiction or cause.",
},
]);
const emergentNode = result.updatedSituationGraph.nodes.find(
(node) => node.id === result.emergentReasoningNodeId,
);
expect(emergentNode).toMatchObject({
kind: "unknown",
status: "unknown",
confidence: "medium",
});
expect(emergentNode.description.toLowerCase()).toContain("because");
expect(
result.updatedSituationGraph.edges.filter(
(edge) => edge.toNodeId === result.emergentReasoningNodeId,
),
).not.toEqual([]);
expect(
result.updatedSituationGraph.edges.some(
(edge) =>
edge.toNodeId === result.emergentReasoningNodeId &&
edge.relationship === "causes",
),
).toBe(false);
expect(
JSON.stringify(
result.updatedSituationGraph.nodes.find(
(node) => node.id === "n-unrelated",
),
),
).toBe(originalUnrelatedNode);
});
it("reuses an equivalent existing unresolved reasoning unknown instead of creating a duplicate", () => {
const { graph, proposal } = makeComparabilityUpdateFixture();
graph.nodes.push(
makeNode({
id: "n-existing-explanation",
label:
"Explanation for why Revenue increased by 18%, but cash in the bank fell over the same period",
description:
"Need to understand what change or event could explain why these observations differ, because that is needed to investigate their relationship.",
kind: "unknown",
status: "unknown",
confidence: "medium",
}),
);
const result = applyValidatedProposal({
situationGraph: graph,
proposal,
previousQuestion:
"Were these figures measured on the same basis and at the same scale?",
answer:
"Yes. Both figures cover the same accounting period and are taken from the same management accounts.",
});
expect(result.success).toBe(true);
expect(result.emergentReasoningNodeCreated).toBe(false);
expect(result.emergentReasoningNodeId).toBe("n-existing-explanation");
expect(result.newActiveUnknownNodeId).not.toBe("n-existing-explanation");
expect(result.selectedQuestion?.nodeId).not.toBe("n-existing-explanation");
expect(result.selectedQuestion?.question).toBe(
"What evidence would clarify how the two observations were measured?",
);
expect(
result.updatedSituationGraph.nodes.filter(
(node) => node.label === graph.nodes.at(-1).label,
),
).toHaveLength(1);
});
it("decomposes a composite selected unknown before asking the next question", () => {
const { graph, proposal } = makeComparabilityUpdateFixture();
const result = applyValidatedProposal({
situationGraph: graph,
proposal,
previousQuestion:
"Were these figures measured on the same basis and at the same scale?",
answer:
"Yes. Both figures cover the same accounting period and are taken from the same management accounts.",
});
expect(result.success).toBe(true);
expect(result.atomicityAssessment).toBe("composite");
expect(result.decompositionPerformed).toBe(true);
expect(result.childUnknownCount).toBe(5);
expect(result.childNodeIds).toHaveLength(5);
expect(result.atomicityReason).toBeTruthy();
expect(result.selectedQuestion?.nodeId).toBe(result.newActiveUnknownNodeId);
expect(result.selectedQuestion?.nodeId).not.toBe(
result.emergentReasoningNodeId,
);
expect(result.selectedQuestion?.question.toLowerCase()).not.toMatch(
/dso|working capital|receivables|capex/,
);
const parentNode = result.updatedSituationGraph.nodes.find(
(node) => node.id === result.emergentReasoningNodeId,
);
expect(parentNode?.status).toBe("unknown");
const childNodes = result.updatedSituationGraph.nodes.filter((node) =>
result.childNodeIds.includes(node.id),
);
expect(childNodes).toHaveLength(5);
expect(childNodes.every((node) => node.parentId === parentNode.id)).toBe(
true,
);
expect(
result.updatedSituationGraph.edges.filter(
(edge) =>
result.childNodeIds.includes(edge.fromNodeId) &&
edge.toNodeId === parentNode.id &&
edge.relationship === "depends_on",
),
).toHaveLength(5);
});
it("reuses existing decomposition children instead of duplicating them", () => {
const { graph, proposal } = makeComparabilityUpdateFixture();
const firstResult = applyValidatedProposal({
situationGraph: graph,
proposal,
previousQuestion:
"Were these figures measured on the same basis and at the same scale?",
answer:
"Yes. Both figures cover the same accounting period and are taken from the same management accounts.",
});
expect(firstResult.success).toBe(true);
const secondResult = applyValidatedProposal({
situationGraph: graph,
proposal,
previousQuestion:
"Were these figures measured on the same basis and at the same scale?",
answer:
"Yes. Both figures cover the same accounting period and are taken from the same management accounts.",
});
expect(secondResult.success).toBe(true);
expect(secondResult.atomicityAssessment).toBe("composite");
const uniqueChildIds = new Set(firstResult.childNodeIds);
expect(uniqueChildIds.size).toBe(firstResult.childNodeIds.length);
expect(
secondResult.updatedSituationGraph.nodes.filter((node) =>
firstResult.childNodeIds.includes(node.id),
),
).toHaveLength(firstResult.childNodeIds.length);
});
});
+118
View File
@@ -0,0 +1,118 @@
import { describe, expect, it } from "vitest";
import { assessUnknownAtomicity } from "@/lib/graph/question-formulator.js";
import { makeGraph, makeNode } from "@/lib/graph/schema.js";
function makeGraphWithUnknown(centralStatement, unknown, observations = []) {
return makeGraph({
centralStatement,
nodes: [unknown, ...observations],
edges: [],
activeUnknownNodeId: unknown.id,
resolvedNodeIds: [],
currentSummary: "Atomicity test graph",
});
}
describe("assessUnknownAtomicity", () => {
it("classifies denominator-style unknowns as atomic", () => {
const unknown = makeNode({
id: "n-denominator",
label: "Complaint rate denominator",
description:
"Need the denominator because it directly determines the complaint rate.",
kind: "unknown",
status: "unknown",
confidence: "high",
});
const result = assessUnknownAtomicity({
node: unknown,
graph: makeGraphWithUnknown(
"Production increased while complaints increased.",
unknown,
),
});
expect(result.atomicity).toBe("atomic");
expect(result.reason.toLowerCase()).toContain("directly");
});
it("classifies relationship explanation unknowns as composite", () => {
const unknown = makeNode({
id: "n-explanation",
label:
"Explanation for why revenue increased by 18%, but cash in the bank fell over the same period",
description:
"Need to understand what change or event could explain why these observations differ, because that is needed to investigate their relationship.",
kind: "unknown",
status: "unknown",
confidence: "medium",
});
const graph = makeGraphWithUnknown(
"Revenue increased by 18%, but cash in the bank fell over the same period.",
unknown,
[
makeNode({
id: "n-revenue",
label: "Revenue increased by 18%.",
description: "Revenue increased by 18%.",
kind: "observation",
status: "supported",
confidence: "high",
}),
makeNode({
id: "n-cash",
label: "Cash in the bank decreased over the same period.",
description: "Cash in the bank decreased over the same period.",
kind: "observation",
status: "supported",
confidence: "high",
}),
],
);
const result = assessUnknownAtomicity({ node: unknown, graph });
expect(result.atomicity).toBe("composite");
expect(result.decompositionKind).toBe("relationship_explanation");
});
it.each([
[
"Customer satisfaction rose, but complaints also rose.",
"Explanation for why customer satisfaction rose, but complaints also rose",
],
[
"Delivery time fell, but cancellations increased.",
"Possible causes of why delivery time fell, but cancellations increased",
],
[
"Traffic increased, but sales stayed flat.",
"Broad explanation for why traffic increased, but sales stayed flat",
],
[
"Production increased, but defects also increased.",
"Factors behind why production increased, but defects also increased",
],
])(
"classifies broad divergence unknowns as composite: %s",
(scenario, label) => {
const unknown = makeNode({
id: `n-${label.length}`,
label,
description: `${label} because the current unknown is too broad to ask directly.`,
kind: "unknown",
status: "unknown",
confidence: "medium",
});
const result = assessUnknownAtomicity({
node: unknown,
graph: makeGraphWithUnknown(scenario, unknown),
});
expect(result.atomicity).toBe("composite");
},
);
});
@@ -0,0 +1,180 @@
import { describe, expect, it } from "vitest";
import {
assessComparability,
classifyObservationRelationship,
formulateTieResolutionQuestion,
} from "@/lib/graph/question-formulator.js";
import { explainUnknownSelection } from "@/lib/graph/utils.js";
import { comparabilityAssessmentFixtures } from "@/tests/fixtures/comparability-assessment.js";
describe("comparability assessment", () => {
it("generates comparison or relationship questions only when warranted", () => {
const summary = comparabilityAssessmentFixtures.map((fixture) => {
const assessment = assessComparability(fixture.graph);
const relationship = classifyObservationRelationship(fixture.graph);
const question = formulateTieResolutionQuestion({ graph: fixture.graph });
const ambiguity = explainUnknownSelection(fixture.graph, []);
expect(assessment.comparabilityStatus).toBe(
fixture.expectedComparabilityStatus,
);
expect(question.comparabilityStatus).toBe(
fixture.expectedComparabilityStatus,
);
if (fixture.expectsComparisonQuestion) {
expect(question.question.toLowerCase()).toContain("same");
expect(question.contradictionReasoningAllowed).toBe(false);
} else {
expect(question.question?.toLowerCase() || "").not.toContain(
"same period and at the same scale",
);
}
if (fixture.key !== "sales-same") {
expect(ambiguity.status).toBe("ambiguous");
}
return {
scenario: fixture.scenario,
comparabilityStatus: assessment.comparabilityStatus,
relationshipStatus: relationship.relationshipStatus,
relationshipAssessed: relationship.relationshipAssessed,
contradictionReasoningAllowed: question.contradictionReasoningAllowed,
question: question.question,
};
});
expect(summary).toEqual([
{
scenario:
"Revenue increased by 18%, but cash in the bank fell over the same period.",
comparabilityStatus: "uncertain",
relationshipStatus: "insufficient_information",
relationshipAssessed: false,
contradictionReasoningAllowed: false,
question:
"Were these figures measured on the same basis and at the same scale?",
},
{
scenario: "Complaints increased. Production increased.",
comparabilityStatus: "uncertain",
relationshipStatus: "insufficient_information",
relationshipAssessed: false,
contradictionReasoningAllowed: false,
question:
"Were these figures measured over the same period and at the same scale?",
},
{
scenario:
"Average delivery time decreased by 25%, but order cancellations increased.",
comparabilityStatus: "uncertain",
relationshipStatus: "insufficient_information",
relationshipAssessed: false,
contradictionReasoningAllowed: false,
question:
"Were these figures measured over the same period and at the same scale?",
},
{
scenario: "Customer satisfaction increased, but complaints increased.",
comparabilityStatus: "uncertain",
relationshipStatus: "insufficient_information",
relationshipAssessed: false,
contradictionReasoningAllowed: false,
question:
"Were these figures measured over the same period and at the same scale?",
},
{
scenario: "Temperature increased. Ice melted.",
comparabilityStatus: "confirmed",
relationshipStatus: "compatible",
relationshipAssessed: true,
contradictionReasoningAllowed: false,
question: null,
},
{
scenario: "Sales doubled. Sales doubled.",
comparabilityStatus: "confirmed",
relationshipStatus: "duplicate",
relationshipAssessed: true,
contradictionReasoningAllowed: false,
question: null,
},
]);
});
it("defers relationship classification while comparability is uncertain", () => {
const fixture = comparabilityAssessmentFixtures[0];
const relationship = classifyObservationRelationship(fixture.graph);
expect(relationship).toMatchObject({
relationshipStatus: "insufficient_information",
relationshipAssessed: false,
contradictionReasoningAllowed: false,
questionRequired: true,
});
expect(relationship.reasoningStages).toEqual([
{
stage: "comparability",
status: "uncertain",
outcome:
"Comparability between the observations is not yet established across period, scale, or measurement basis.",
},
{
stage: "relationship",
status: "insufficient_information",
outcome: "not assessed until comparability is established",
},
]);
});
it("allows contradiction reasoning only for genuine contradictions", () => {
const serviceGraph = {
centralStatement:
"The service was reported as available throughout the hour and unavailable throughout the same hour.",
nodes: [
{
id: "service-available",
label: "The service was available throughout the hour.",
description: "The service was available throughout the hour.",
kind: "observation",
status: "supported",
confidence: "high",
value: null,
unit: null,
evidenceIds: [],
dependsOn: [],
affects: [],
parentId: null,
childIds: [],
},
{
id: "service-unavailable",
label: "The service was unavailable throughout the same hour.",
description: "The service was unavailable throughout the same hour.",
kind: "observation",
status: "supported",
confidence: "high",
value: null,
unit: null,
evidenceIds: [],
dependsOn: [],
affects: [],
parentId: null,
childIds: [],
},
],
edges: [],
activeUnknownNodeId: null,
resolvedNodeIds: [],
currentSummary: "Service contradiction fixture",
};
const relationship = classifyObservationRelationship(serviceGraph);
expect(relationship).toMatchObject({
relationshipStatus: "contradictory",
contradictionReasoningAllowed: true,
questionRequired: true,
});
});
});
+179
View File
@@ -0,0 +1,179 @@
import { describe, expect, it } from "vitest";
import { applyValidatedProposal } from "@/lib/graph/apply-proposal.js";
import { makeEdge, makeGraph, makeNode } from "@/lib/graph/schema.js";
function makeFixture() {
const parent = makeNode({
id: "n-parent",
label: "Explanation for why revenue increased while cash fell",
description:
"Need to understand what change or event could explain why these observations differ, because that is needed to investigate their relationship.",
kind: "unknown",
status: "unknown",
confidence: "medium",
});
const children = [
makeNode({
id: "n-child-1",
label: "How the two observations were measured",
description: "Need evidence about the measure used for each observation.",
kind: "unknown",
status: "unknown",
confidence: "medium",
parentId: parent.id,
}),
makeNode({
id: "n-child-2",
label: "Whether the two observations reflect different timing",
description:
"Need to know whether the two observations reflect different timing.",
kind: "unknown",
status: "unknown",
confidence: "medium",
parentId: parent.id,
}),
makeNode({
id: "n-child-3",
label: "Possible change mainly affecting revenue",
description:
"Need to know whether a possible change mainly affected revenue.",
kind: "unknown",
status: "unknown",
confidence: "medium",
parentId: parent.id,
}),
makeNode({
id: "n-child-4",
label: "Possible one-off event during the period",
description:
"Need to know whether a possible one-off event happened during the period.",
kind: "unknown",
status: "unknown",
confidence: "medium",
parentId: parent.id,
}),
];
return makeGraph({
centralStatement: "Revenue increased while cash fell.",
nodes: [parent, ...children],
edges: children.map((child, index) =>
makeEdge({
id: `e-${index + 1}`,
fromNodeId: child.id,
toNodeId: parent.id,
relationship: "depends_on",
description: `${child.label} feeds the parent.`,
}),
),
activeUnknownNodeId: "n-child-1",
resolvedNodeIds: [],
currentSummary: "confidence propagation fixture",
});
}
function makeProposal({ resolvedIds, contradictedIds = [] }) {
return {
addedNodes: [
makeNode({
id: "n-anchor",
label: "Update anchor",
description:
"Anchor state introduced by the answer because the update must contain a meaningful change.",
kind: "state",
status: "known",
confidence: "low",
}),
],
updatedNodes: [
...resolvedIds.map((id) => ({
nodeId: id,
previousStatus: "unknown",
newStatus: "resolved",
previousValue: null,
newValue: `answer:${id}`,
reason: "resolved child",
})),
...contradictedIds.map((id) => ({
nodeId: id,
previousStatus: "unknown",
newStatus: "contradicted",
previousValue: null,
newValue: `contradiction:${id}`,
reason: "contradictory child evidence",
})),
],
addedEdges: [],
removedEdgeIds: [],
resolvedUnknownNodeIds: resolvedIds,
affectedNodeIds: [],
selectedQuestion: null,
};
}
describe("confidence propagation", () => {
it("one of four children resolved does not yield high conclusion confidence", () => {
const result = applyValidatedProposal({
situationGraph: makeFixture(),
proposal: makeProposal({ resolvedIds: ["n-child-1"] }),
previousQuestion:
"What evidence would clarify how the two observations were measured?",
answer: "Same accounting period and same management accounts.",
});
const parent = result.updatedSituationGraph.nodes.find(
(n) => n.id === "n-parent",
);
expect(parent.status).toBe("provisional");
expect(parent.confidence).toBe("medium");
expect(parent.confidenceAssessment).toEqual({
evidenceConfidence: "medium",
completenessStatus: "partial",
conclusionConfidence: "medium",
});
expect(result.confidenceCapReason).toBe(
"unresolved_direct_children_cap_conclusion",
);
});
it("all children resolved with coherent evidence may yield high confidence", () => {
const result = applyValidatedProposal({
situationGraph: makeFixture(),
proposal: makeProposal({
resolvedIds: ["n-child-1", "n-child-2", "n-child-3", "n-child-4"],
}),
previousQuestion:
"What evidence would clarify how the two observations were measured?",
answer: "All direct child questions are answered.",
});
const parent = result.updatedSituationGraph.nodes.find(
(n) => n.id === "n-parent",
);
expect(parent.status).toBe("resolved");
expect(parent.confidenceAssessment).toEqual({
evidenceConfidence: "high",
completenessStatus: "complete",
conclusionConfidence: "high",
});
});
it("contradictory child evidence prevents high confidence", () => {
const result = applyValidatedProposal({
situationGraph: makeFixture(),
proposal: makeProposal({
resolvedIds: ["n-child-1"],
contradictedIds: ["n-child-2"],
}),
previousQuestion:
"What evidence would clarify how the two observations were measured?",
answer: "One child resolved, another contradicted.",
});
const parent = result.updatedSituationGraph.nodes.find(
(n) => n.id === "n-parent",
);
expect(parent.confidenceAssessment.conclusionConfidence).toBe("low");
expect(result.confidenceCapReason).toBe("contradictory_direct_children");
});
});
@@ -0,0 +1,306 @@
import { describe, expect, it } from "vitest";
import {
applyValidatedProposal,
evaluateBranchInteractions,
} from "@/lib/graph/apply-proposal.js";
import { makeEdge, makeGraph, makeNode } from "@/lib/graph/schema.js";
function makeParentWithBranches(children) {
const parent = makeNode({
id: "n-parent",
label: "Explanation for why revenue increased while cash fell",
description:
"Need to understand what change or event could explain why these observations differ, because that is needed to investigate their relationship.",
kind: "unknown",
status: "unknown",
confidence: "medium",
});
return makeGraph({
centralStatement: "Revenue increased while cash fell.",
nodes: [
parent,
...children.map((child) => ({ ...child, parentId: parent.id })),
],
edges: children.map((child, index) =>
makeEdge({
id: `e-${index + 1}`,
fromNodeId: child.id,
toNodeId: parent.id,
relationship: "depends_on",
description: `${child.label} feeds the parent.`,
}),
),
activeUnknownNodeId: children[0]?.id ?? null,
resolvedNodeIds: [],
currentSummary: "cross-branch corroboration fixture",
});
}
function makeResolvedChild(id, label, value, extra = {}) {
return makeNode({
id,
label,
description: label,
kind: "unknown",
status: "resolved",
confidence: "medium",
value,
evidenceIds: extra.evidenceIds ?? [],
});
}
function makeUnknownBranch(id, label, description, extra = {}) {
return makeNode({
id,
label,
description,
kind: "unknown",
status: extra.status ?? "unknown",
confidence: extra.confidence ?? "medium",
evidenceIds: extra.evidenceIds ?? [],
value: extra.value ?? null,
});
}
describe("evaluateBranchInteractions", () => {
it("detects corroborating independent branches", () => {
const graph = makeParentWithBranches([
makeResolvedChild("n-a", "Debtor balance increased", "bank-statement-a", {
evidenceIds: ["bank-statement-a"],
}),
makeResolvedChild(
"n-b",
"Cash receipts were delayed",
"receipts-ledger-b",
{ evidenceIds: ["receipts-ledger-b"] },
),
]);
const parentNode = graph.nodes.find((node) => node.id === "n-parent");
const result = evaluateBranchInteractions({ parentNode, graph });
expect(result.interactionSummary.corroboratingBranchCount).toBe(1);
expect(result.interactionSummary.duplicateEvidenceCount).toBe(0);
expect(result.interactionSummary.conflictingBranchCount).toBe(0);
});
it("detects duplicate evidence instead of corroboration", () => {
const graph = makeParentWithBranches([
makeResolvedChild(
"n-a",
"Bank statement shows increased debtor balance",
"same-bank",
{
evidenceIds: ["same-bank"],
},
),
makeResolvedChild(
"n-b",
"Delayed receipts also cite the bank statement",
"same-bank",
{
evidenceIds: ["same-bank"],
},
),
]);
const parentNode = graph.nodes.find((node) => node.id === "n-parent");
const result = evaluateBranchInteractions({ parentNode, graph });
expect(result.interactionSummary.duplicateEvidenceCount).toBe(1);
expect(result.interactionSummary.corroboratingBranchCount).toBe(0);
});
it("detects conflicting branches", () => {
const graph = makeParentWithBranches([
makeResolvedChild("n-a", "Revenue recognised correctly", "correctly"),
makeNode({
id: "n-b",
label: "Revenue recognised incorrectly",
description: "Revenue recognised incorrectly",
kind: "unknown",
status: "contradicted",
confidence: "medium",
value: "incorrectly",
}),
]);
const parentNode = graph.nodes.find((node) => node.id === "n-parent");
const result = evaluateBranchInteractions({ parentNode, graph });
expect(result.interactionSummary.conflictingBranchCount).toBe(1);
});
});
describe("cross-branch corroboration effects", () => {
function applyToGraph(children, resolvedIds, contradictedIds = []) {
const graph = makeParentWithBranches(children);
return applyValidatedProposal({
situationGraph: graph,
proposal: {
addedNodes: [
makeNode({
id: "n-anchor",
label: "Update anchor",
description:
"Anchor state introduced by the answer because the update must contain a meaningful change.",
kind: "state",
status: "known",
confidence: "low",
}),
],
updatedNodes: [
...resolvedIds.map((id) => ({
nodeId: id,
previousStatus: "unknown",
newStatus: "resolved",
previousValue: null,
newValue: `answer:${id}`,
reason: "resolved child",
})),
...contradictedIds.map((id) => ({
nodeId: id,
previousStatus: "unknown",
newStatus: "contradicted",
previousValue: null,
newValue: `contradiction:${id}`,
reason: "contradicted child",
})),
],
addedEdges: [],
removedEdgeIds: [],
resolvedUnknownNodeIds: resolvedIds,
affectedNodeIds: [],
selectedQuestion: null,
},
previousQuestion: "What evidence would clarify this branch?",
answer: "deterministic branch update",
});
}
it("independent corroboration increases justified confidence without reaching high on incomplete parent", () => {
const result = applyToGraph(
[
makeUnknownBranch(
"n-a",
"Debtor balance increased",
"Debtor balance increased",
),
makeUnknownBranch(
"n-b",
"Cash receipts delayed",
"Cash receipts delayed",
),
makeUnknownBranch(
"n-c",
"Possible one-off event during the period",
"Possible one-off event during the period",
),
makeUnknownBranch(
"n-d",
"Whether the two observations reflect different timing",
"Whether the two observations reflect different timing",
),
],
["n-a", "n-b"],
);
expect(result.success).toBe(true);
expect(result.interactionSummary?.corroboratingBranchCount).toBeGreaterThan(
0,
);
expect(result.interactionSummary?.duplicateEvidenceCount).toBe(0);
expect(result.parentConfidenceAfter).toBe("medium");
expect(result.confidenceCapReason).toBe(
"independent_corroboration_with_incomplete_parent",
);
});
it("duplicate evidence does not increase confidence", () => {
const result = applyToGraph(
[
makeUnknownBranch(
"n-a",
"Bank statement shows increased debtor balance",
"Bank statement shows increased debtor balance",
{ evidenceIds: ["same-bank"] },
),
makeUnknownBranch(
"n-b",
"Delayed receipts also cite the bank statement",
"Delayed receipts also cite the bank statement",
{ evidenceIds: ["same-bank"] },
),
makeUnknownBranch(
"n-c",
"Possible one-off event during the period",
"Possible one-off event during the period",
),
],
["n-a", "n-b"],
);
expect(result.success).toBe(true);
expect(result.interactionSummary?.duplicateEvidenceCount).toBeGreaterThan(
0,
);
expect(result.interactionSummary?.corroboratingBranchCount).toBe(0);
expect(result.confidenceCapReason).toBe(
"duplicate_evidence_no_extra_confidence",
);
});
it("conflicting evidence caps confidence", () => {
const result = applyToGraph(
[
makeUnknownBranch(
"n-a",
"Revenue recognised correctly",
"Revenue recognised correctly",
),
makeUnknownBranch(
"n-b",
"Revenue recognised incorrectly",
"Revenue recognised incorrectly",
),
makeUnknownBranch(
"n-c",
"Possible one-off event during the period",
"Possible one-off event during the period",
),
],
["n-a"],
["n-b"],
);
expect(result.success).toBe(true);
expect(result.interactionSummary?.conflictingBranchCount).toBeGreaterThan(
0,
);
expect(result.conclusionConfidenceAfter).toBe("low");
});
it("independent branches stay interaction-neutral", () => {
const result = applyToGraph(
[
makeUnknownBranch(
"n-a",
"Marketing campaign changed traffic",
"Marketing campaign changed traffic",
),
makeUnknownBranch(
"n-b",
"Equipment maintenance occurred",
"Equipment maintenance occurred",
),
],
["n-a"],
);
expect(result.success).toBe(true);
expect(result.interactionSummary?.independentBranchCount).toBeGreaterThan(
0,
);
});
});
+278
View File
@@ -0,0 +1,278 @@
import { describe, expect, it } from "vitest";
import {
assessChildUnknownQuality,
applyValidatedProposal,
MAX_DECOMPOSITION_DEPTH,
} from "@/lib/graph/apply-proposal.js";
import { makeGraph, makeNode } from "@/lib/graph/schema.js";
function makeParentGraph({
centralStatement,
parentLabel,
parentDescription,
observations = [],
}) {
const parent = makeNode({
id: "n-parent",
label: parentLabel,
description: parentDescription,
kind: "unknown",
status: "unknown",
confidence: "medium",
});
return {
parent,
graph: makeGraph({
centralStatement,
nodes: [parent, ...observations],
edges: [],
activeUnknownNodeId: parent.id,
resolvedNodeIds: [],
currentSummary: "Decomposition quality graph",
}),
};
}
describe("assessChildUnknownQuality", () => {
it("rejects 'Timing or measurement basis' as compound", () => {
const { parent, graph } = makeParentGraph({
centralStatement:
"Revenue increased by 18%, but cash in the bank fell over the same period.",
parentLabel:
"Explanation for why revenue increased by 18%, but cash in the bank fell over the same period",
parentDescription:
"Need to understand what change or event could explain why these observations differ, because that is needed to investigate their relationship.",
});
const child = makeNode({
id: "n-child",
label: "Timing or measurement basis",
description:
"Need evidence about whether a timing or measurement-basis difference could explain the observations, because that would change how they should be interpreted.",
kind: "unknown",
status: "unknown",
confidence: "medium",
parentId: parent.id,
});
const result = assessChildUnknownQuality({
parentNode: parent,
childNode: child,
siblingNodes: [child],
graph,
});
expect(result.valid).toBe(false);
expect(result.compoundSignals).toContain("timing_or_measurement_basis");
expect(result.reasons).toContain("compound_child");
});
it("accepts a child with one directly answerable uncertainty", () => {
const { parent, graph } = makeParentGraph({
centralStatement: "Traffic increased, but sales stayed flat.",
parentLabel:
"What explains why more website traffic did not produce more sales?",
parentDescription:
"Need an explanation because the observations moved differently.",
});
const child = makeNode({
id: "n-child",
label: "Different measurement basis between the two observations",
description:
"Need evidence about whether the two observations use different measurement bases, because that could help explain the difference.",
kind: "unknown",
status: "unknown",
confidence: "medium",
parentId: parent.id,
});
const result = assessChildUnknownQuality({
parentNode: parent,
childNode: child,
siblingNodes: [child],
graph,
});
expect(result.valid).toBe(true);
expect(result.atomic).toBe(true);
expect(result.directlyAnswerable).toBe(true);
expect(result.narrowerThanParent).toBe(true);
});
it("rejects sibling duplicates", () => {
const { parent, graph } = makeParentGraph({
centralStatement: "Production increased, but defects also increased.",
parentLabel: "What explains why output and defects both increased?",
parentDescription:
"Need an explanation because both observations increased.",
});
const childA = makeNode({
id: "n-child-a",
label: "Different timing between the two observations",
description:
"Need evidence about whether the two observations reflect different timing, because that could help explain the difference.",
kind: "unknown",
status: "unknown",
confidence: "medium",
parentId: parent.id,
});
const childB = makeNode({
id: "n-child-b",
label: "Different timing between the two observations",
description:
"Need evidence about whether the two observations reflect different timing, because that could help explain the difference.",
kind: "unknown",
status: "unknown",
confidence: "medium",
parentId: parent.id,
});
const result = assessChildUnknownQuality({
parentNode: parent,
childNode: childA,
siblingNodes: [childA, childB],
graph,
});
expect(result.valid).toBe(false);
expect(result.duplicateSiblingIds).toContain("n-child-b");
});
it("rejects parent paraphrases", () => {
const { parent, graph } = makeParentGraph({
centralStatement:
"Customer satisfaction scores increased, but complaints also increased.",
parentLabel:
"What explains why satisfaction and complaints both increased?",
parentDescription:
"Need a broad explanation because the observations moved differently.",
});
const child = makeNode({
id: "n-child",
label: "What explains why satisfaction and complaints both increased?",
description:
"Need a broad explanation because the observations moved differently.",
kind: "unknown",
status: "unknown",
confidence: "medium",
parentId: parent.id,
});
const result = assessChildUnknownQuality({
parentNode: parent,
childNode: child,
siblingNodes: [child],
graph,
});
expect(result.valid).toBe(false);
expect(result.reasons).toContain("not_narrower_than_parent");
});
});
describe("decomposition stopping conditions", () => {
function makeMeaningfulNoOpProposal() {
return {
addedNodes: [
makeNode({
id: "n-anchor",
label: "Update anchor",
description:
"Anchor state introduced by the answer because the update must contain a meaningful change.",
kind: "state",
status: "known",
confidence: "low",
}),
],
updatedNodes: [],
addedEdges: [],
removedEdgeIds: [],
resolvedUnknownNodeIds: [],
affectedNodeIds: [],
selectedQuestion: null,
};
}
it("does not decompose an atomic selected unknown", () => {
const atomic = makeNode({
id: "n-atomic",
label: "Were both figures measured over the same accounting period?",
description:
"Need to know whether both figures cover the same accounting period because that determines whether they are directly comparable.",
kind: "unknown",
status: "unknown",
confidence: "high",
});
const graph = makeGraph({
centralStatement:
"Revenue increased by 18%, but cash in the bank fell over the same period.",
nodes: [atomic],
edges: [],
activeUnknownNodeId: atomic.id,
resolvedNodeIds: [],
currentSummary: "Atomic selected node graph",
});
const result = applyValidatedProposal({
situationGraph: graph,
proposal: makeMeaningfulNoOpProposal(),
});
expect(result.success).toBe(true);
expect(result.decompositionAttempted).toBe(false);
expect(result.decompositionStoppedReason).toBe(
"Selected unknown is already atomic.",
);
});
it("stops once a directly answerable child is selected", () => {
const { parent, graph } = makeParentGraph({
centralStatement: "Traffic increased, but sales stayed flat.",
parentLabel:
"What explains why more website traffic did not produce more sales?",
parentDescription:
"Need an explanation because the observations moved differently.",
observations: [
makeNode({
id: "n-traffic",
label: "Website traffic increased.",
description: "Website traffic increased.",
kind: "observation",
status: "supported",
confidence: "high",
}),
makeNode({
id: "n-sales",
label: "Sales stayed flat.",
description: "Sales stayed flat.",
kind: "observation",
status: "supported",
confidence: "high",
}),
],
});
const result = applyValidatedProposal({
situationGraph: graph,
proposal: makeMeaningfulNoOpProposal(),
});
expect(result.success).toBe(true);
expect(result.decompositionAttempted).toBe(true);
expect(result.decompositionAccepted).toBe(true);
expect(result.selectedQuestion).toMatchObject({
nodeId: expect.any(String),
question:
"What evidence would clarify how the two observations were measured?",
});
expect(result.selectedChildNodeId).toBe(result.selectedQuestion?.nodeId);
expect(result.decompositionStoppedReason).toBe(
"Selected child is atomic and directly answerable.",
);
});
it("exposes the configured maximum decomposition depth", () => {
expect(MAX_DECOMPOSITION_DEPTH).toBeGreaterThanOrEqual(2);
expect(MAX_DECOMPOSITION_DEPTH).toBeLessThanOrEqual(3);
});
});
+316
View File
@@ -118,6 +118,90 @@ function makeProposal(overrides = {}) {
};
}
function makeComparabilityScenarioGraph() {
return makeGraph({
centralStatement:
"Revenue increased by 18%, but cash in the bank fell over the same period.",
nodes: [
makeNode({
id: "n-comparability-unknown",
label: "Whether the figures are comparable",
description:
"Need to know whether the figures use the same period, basis, and scale before comparing them.",
kind: "unknown",
status: "unknown",
confidence: "high",
}),
makeNode({
id: "n-revenue-observation",
label: "Revenue increased by 18%.",
description: "Revenue increased by 18%.",
kind: "observation",
status: "supported",
confidence: "high",
}),
makeNode({
id: "n-cash-observation",
label: "Cash in the bank decreased over the same period.",
description: "Cash in the bank decreased over the same period.",
kind: "observation",
status: "supported",
confidence: "high",
}),
],
edges: [],
activeUnknownNodeId: "n-comparability-unknown",
resolvedNodeIds: [],
currentSummary: "Comparability scenario",
reasoningState: {
comparabilityStatus: "uncertain",
comparabilityReason:
"Comparability between the observations is not yet established across period, scale, or measurement basis.",
comparabilityEvidence: [],
relationshipStatus: "insufficient_information",
relationshipReason:
"Relationship classification is deferred until comparability is established.",
relationshipAssessed: false,
contradictionReasoningAllowed: false,
reasoningStages: [
{
stage: "comparability",
status: "uncertain",
outcome:
"Comparability between the observations is not yet established across period, scale, or measurement basis.",
},
{
stage: "relationship",
status: "insufficient_information",
outcome: "not assessed until comparability is established",
},
],
},
});
}
function makeComparabilityProposal() {
return {
addedNodes: [],
updatedNodes: [
{
nodeId: "n-comparability-unknown",
previousStatus: "unknown",
newStatus: "resolved",
previousValue: null,
newValue:
"Both figures cover the same accounting period and are taken from the same management accounts.",
reason: "The answer confirms comparability.",
},
],
addedEdges: [],
removedEdgeIds: [],
resolvedUnknownNodeIds: ["n-comparability-unknown"],
affectedNodeIds: [],
selectedQuestion: null,
};
}
describe("lib/graph/orchestrator startCase", () => {
beforeEach(() => {
vi.resetModules();
@@ -178,6 +262,96 @@ describe("lib/graph/orchestrator startCase", () => {
expect(result.success).toBe(true);
expect(result.situationGraph.activeUnknownNodeId).toBeTruthy();
expect(result.diagnostics.unknownSelectionExplanation?.selectedNodeId).toBe(
result.situationGraph.activeUnknownNodeId,
);
});
it("returns an ambiguous tie result instead of choosing by label order", async () => {
mockAnalyseScenario.mockResolvedValue(
makeAnalysisResult({
reconstruction: {
summary: "Revenue up while cash falls",
actors: [],
systemsOrObjects: [],
expectedStates: [],
observedStates: [
{
id: "obs-1",
label: "Revenue increased by 18%.",
description: "Revenue increased by 18%.",
confidence: "high",
},
{
id: "obs-2",
label: "Cash in the bank decreased over the same period.",
description: "Cash in the bank decreased over the same period.",
confidence: "high",
},
],
differences: [],
knownTransitions: [],
unexplainedTransitions: [],
contradictions: [
{
id: "c-1",
label:
"Apparent contradiction between profitability/revenue expansion and liquidity reduction.",
description:
"Apparent contradiction between profitability/revenue expansion and liquidity reduction.",
confidence: "medium",
},
],
importantUnknowns: [
{
id: "unk-1",
label:
"Whether revenue recognition timing differs from cash collection timing.",
description:
"Whether revenue recognition timing differs from cash collection timing.",
confidence: "high",
},
{
id: "unk-2",
label:
"Magnitude and nature of cash outflows (operating expenses, debt repayments, capex, or working capital shifts).",
description:
"Magnitude and nature of cash outflows (operating expenses, debt repayments, capex, or working capital shifts).",
confidence: "high",
},
],
plausibleInterpretations: [],
},
}),
);
const { startCase } = await import("@/lib/graph/orchestrator.js");
const result = await startCase({
scenario:
"Revenue increased by 18%, but cash in the bank fell over the same period.",
});
expect(result.success).toBe(true);
expect(result.situationGraph.activeUnknownNodeId).toBeNull();
expect(result.selectedQuestion).toMatchObject({
id: "q_tie_resolution",
selectionStatus: "ambiguous",
question:
"Were these figures measured on the same basis and at the same scale?",
tiedCandidateIds: expect.arrayContaining([expect.any(String)]),
comparabilityStatus: "uncertain",
relationshipStatus: "insufficient_information",
relationshipAssessed: false,
contradictionReasoningAllowed: false,
});
expect(result.diagnostics.unknownSelectionExplanation).toMatchObject({
status: "ambiguous",
tieType: "complete_unresolved_tie",
selectedNodeId: null,
alphabeticalUsedAsReasoning: false,
tieResolutionQuestion:
"Were these figures measured on the same basis and at the same scale?",
});
});
it("returns structured failure when graph reference validation fails", async () => {
@@ -262,6 +436,69 @@ describe("lib/graph/orchestrator startCase", () => {
expect(result.diagnostics.compatibilityChanges).toHaveLength(1);
});
it("preserves selected question bytes while adding selection explanation diagnostics", async () => {
const { updateCase } = await import("@/lib/graph/orchestrator.js");
const provider = {
generateReconstruction: vi.fn().mockResolvedValue(
makeProposal({
addedNodes: [
makeNode({
id: "n-build-decision",
label: "Build Confidence Engine decision",
description: "Decision introduced by the answer.",
kind: "state",
status: "supported",
confidence: "medium",
}),
makeNode({
id: "n-commercial-value",
label: "Commercial value definition",
description:
"Need a concrete definition because the decision depends on it.",
kind: "unknown",
status: "unknown",
confidence: "high",
}),
],
addedEdges: [
{
id: "e-build-commercial-value",
fromNodeId: "n-build-decision",
toNodeId: "n-commercial-value",
relationship: "depends_on",
confidence: "medium",
description:
"The decision depends on commercial value definition.",
},
],
selectedQuestion: {
nodeId: "n-commercial-value",
question:
"How should commercial value be defined for this decision?",
reason: "Consequential unresolved uncertainty remains.",
},
}),
),
};
const first = await updateCase(makeUpdateRequest(), {
provider,
config: MOCK_CONFIG,
applyProposal: true,
});
const second = await updateCase(makeUpdateRequest(), {
provider,
config: MOCK_CONFIG,
applyProposal: true,
});
expect(first.selectedQuestion.question).toBe(
second.selectedQuestion.question,
);
expect(first.selectedQuestion.reason).toBe(second.selectedQuestion.reason);
expect(first.diagnostics.unknownSelectionExplanation).toBeTruthy();
});
it("produces a validated update proposal for a valid request", async () => {
const { updateCase } = await import("@/lib/graph/orchestrator.js");
const provider = {
@@ -777,6 +1014,85 @@ describe("lib/graph/orchestrator startCase", () => {
});
});
it("advances reasoning after comparability is resolved by the update answer", async () => {
const { updateCase } = await import("@/lib/graph/orchestrator.js");
const provider = {
generateReconstruction: vi
.fn()
.mockResolvedValue(makeComparabilityProposal()),
};
const result = await updateCase(
{
situationGraph: makeComparabilityScenarioGraph(),
previousQuestion:
"Were these figures measured on the same basis and at the same scale?",
answer:
"Yes. Both figures cover the same accounting period and are taken from the same management accounts.",
promptVersion: "v0.4",
},
{
provider,
config: MOCK_CONFIG,
applyProposal: true,
},
);
expect(result.success).toBe(true);
expect(result.resolvedUnknownNodeIds).toEqual(["n-comparability-unknown"]);
expect(result.diagnostics).toMatchObject({
previousComparabilityStatus: "uncertain",
comparabilityStatus: "confirmed",
relationshipStatus: "potentially_related",
relationshipAssessed: true,
resolvedReasoningNodeIds: ["reasoning:comparability"],
emergentReasoningNodeCreated: true,
atomicityAssessment: "composite",
decompositionPerformed: true,
childUnknownCount: 5,
});
expect(result.diagnostics.emergentReasoningNodeId).toBeTruthy();
expect(result.diagnostics.childNodeIds).toHaveLength(5);
expect(result.diagnostics.atomicityReason).toBeTruthy();
expect(result.diagnostics.emergentReasoningNodeReason).toContain(
"backed by the graph",
);
expect(result.diagnostics.reasoningStagesBefore).toEqual([
{
stage: "comparability",
status: "uncertain",
outcome:
"Comparability between the observations is not yet established across period, scale, or measurement basis.",
},
{
stage: "relationship",
status: "insufficient_information",
outcome: "not assessed until comparability is established",
},
]);
expect(result.diagnostics.reasoningStagesAfter).toEqual([
{
stage: "comparability",
status: "confirmed",
outcome:
"Comparability was confirmed by the user answer covering the same period and source basis.",
},
{
stage: "relationship",
status: "potentially_related",
outcome:
"The observations concern connected business signals but do not establish a direct contradiction or cause.",
},
]);
expect(result.selectedQuestion?.nodeId).toBe(result.newActiveUnknownNodeId);
expect(result.selectedQuestion?.question).toBe(
"What evidence would clarify how the two observations were measured?",
);
expect(result.selectedQuestion?.question.toLowerCase()).not.toMatch(
/same basis|dso|receivables|debtor days|working capital/,
);
});
it("startCase behaviour remains unchanged", async () => {
mockAnalyseScenario.mockResolvedValue(makeAnalysisResult());
const { startCase } = await import("@/lib/graph/orchestrator.js");
+283 -18
View File
@@ -1,5 +1,10 @@
import { describe, expect, it } from "vitest";
import { formulateQuestion } from "@/lib/graph/question-formulator.js";
import {
assessUnknownAtomicity,
formulateQuestion,
formulateTieResolutionQuestion,
selectInvestigationStrategy,
} from "@/lib/graph/question-formulator.js";
import { makeEdge, makeGraph, makeNode } from "@/lib/graph/schema.js";
function makeGraphFor(node, extra = {}) {
@@ -14,7 +19,64 @@ function makeGraphFor(node, extra = {}) {
}
describe("formulateQuestion", () => {
it("commercial viability plus build decision produces a decision-criterion question", () => {
it("atomicity assessment leaves focused unknowns direct and marks broad explanation unknowns composite", () => {
const atomicUnknown = makeNode({
id: "n-atomic",
label: "Complaint rate denominator",
description:
"Need the denominator because it directly determines the complaint rate.",
kind: "unknown",
status: "unknown",
confidence: "high",
});
const compositeUnknown = makeNode({
id: "n-composite",
label:
"Explanation for why revenue increased by 18%, but cash in the bank fell over the same period",
description:
"Need to understand what change or event could explain why these observations differ, because that is needed to investigate their relationship.",
kind: "unknown",
status: "unknown",
confidence: "medium",
});
const compositeGraph = makeGraphFor(compositeUnknown, {
centralStatement:
"Revenue increased by 18%, but cash in the bank fell over the same period.",
nodes: [
makeNode({
id: "n-revenue-observation",
label: "Revenue increased by 18%.",
description: "Revenue increased by 18%.",
kind: "observation",
status: "supported",
confidence: "high",
}),
makeNode({
id: "n-cash-observation",
label: "Cash in the bank decreased over the same period.",
description: "Cash in the bank decreased over the same period.",
kind: "observation",
status: "supported",
confidence: "high",
}),
],
});
expect(
assessUnknownAtomicity({
node: atomicUnknown,
graph: makeGraphFor(atomicUnknown),
}).atomicity,
).toBe("atomic");
expect(
assessUnknownAtomicity({
node: compositeUnknown,
graph: compositeGraph,
}).atomicity,
).toBe("composite");
});
it("commercial viability plus build decision produces a decision-threshold question", () => {
const unknown = makeNode({
id: "n-commercial",
label: "Uncertainty regarding the commercial value of the product",
@@ -42,7 +104,7 @@ describe("formulateQuestion", () => {
const result = formulateQuestion({ node: unknown, graph });
expect(result.strategy).toBe("decision criterion");
expect(result.strategy).toBe("decision_threshold");
expect(result.question).toContain("What outcome");
expect(result.question.toLowerCase()).toContain("justify");
});
@@ -100,7 +162,7 @@ describe("formulateQuestion", () => {
graph: makeGraphFor(unknown),
});
expect(result.strategy).toBe("evidence");
expect(result.strategy).toBe("evidence_gathering");
expect(result.question).toContain("What evidence");
});
@@ -120,33 +182,41 @@ describe("formulateQuestion", () => {
graph: makeGraphFor(unknown),
});
expect(result.strategy).toBe("baseline");
expect(result.strategy).toBe("baseline_reconstruction");
expect(result.question).toContain("What was the comparable state before");
});
it("unknown customer produces an actor/customer question", () => {
it("conflicting claim produces a contradiction-resolution question", () => {
const unknown = makeNode({
id: "n-customer",
label: "Target customer",
id: "n-conflict",
label: "Conflicting churn claim",
description:
"Need to know the customer because value depends on who receives it.",
"Need to resolve the inconsistency because the current figures contradict each other.",
kind: "unknown",
status: "unknown",
confidence: "high",
confidence: "medium",
});
const contradiction = makeNode({
id: "n-contradiction",
label: "Contradicted report",
description: "Two sources disagree about churn.",
kind: "conclusion",
status: "contradicted",
confidence: "low",
childIds: [unknown.id],
});
const result = formulateQuestion({
node: unknown,
graph: makeGraphFor(unknown),
graph: makeGraphFor(unknown, { nodes: [contradiction] }),
});
expect(result.strategy).toBe("actor/customer");
expect(result.question).toContain(
"Who experiences the problem or receives the value",
);
expect(result.strategy).toBe("contradiction_resolution");
expect(result.question).toContain("resolve the contradiction");
});
it("constraint unknown produces a constraint question", () => {
it("constraint unknown uses evidence-gathering within the fixed strategy set", () => {
const unknown = makeNode({
id: "n-constraint",
label: "Budget constraint",
@@ -162,8 +232,95 @@ describe("formulateQuestion", () => {
graph: makeGraphFor(unknown),
});
expect(result.strategy).toBe("constraint");
expect(result.question).toContain("What constraint most limits");
expect(result.strategy).toBe("evidence_gathering");
expect(result.question).toContain("What evidence");
});
it("the same unknown can produce different questions when paired with different strategies", () => {
const thresholdUnknown = makeNode({
id: "n-threshold-unknown",
label: "Value threshold",
description: "Need to resolve the value threshold.",
kind: "unknown",
status: "unknown",
confidence: "high",
});
const definitionUnknown = makeNode({
id: "n-definition-unknown",
label: "Value term",
description: "Need to resolve what value term refers to in this context.",
kind: "unknown",
status: "unknown",
confidence: "high",
});
const decisionGraph = makeGraphFor(thresholdUnknown, {
centralStatement: "We are deciding whether to launch this product.",
nodes: [
makeNode({
id: "n-decision",
label: "Launch decision",
description: "Decision depends on the value threshold.",
kind: "state",
status: "known",
confidence: "medium",
childIds: [thresholdUnknown.id],
value: "Deciding whether to launch the product",
}),
],
});
const definitionGraph = makeGraphFor(definitionUnknown, {
centralStatement:
"The team uses the term value threshold inconsistently.",
nodes: [
makeNode({
id: "n-definition",
label: "Definition disagreement about value threshold",
description:
"Need a definition of value threshold because the term is used inconsistently before comparing options.",
kind: "state",
status: "known",
confidence: "medium",
childIds: [definitionUnknown.id],
}),
],
});
const decisionResult = formulateQuestion({
node: thresholdUnknown,
graph: decisionGraph,
});
const definitionResult = formulateQuestion({
node: definitionUnknown,
graph: definitionGraph,
});
expect(decisionResult.strategy).toBe("decision_threshold");
expect(definitionResult.strategy).toBe("definition");
expect(decisionResult.question).not.toBe(definitionResult.question);
});
it("strategy selection is deterministic and explainable", () => {
const unknown = makeNode({
id: "n-threshold",
label: "Success threshold",
description:
"Need the success threshold because the decision depends on it.",
kind: "unknown",
status: "unknown",
confidence: "high",
});
const graph = makeGraphFor(unknown, {
centralStatement: "We need to decide whether to continue investing.",
});
const first = selectInvestigationStrategy({ node: unknown, graph });
const second = selectInvestigationStrategy({ node: unknown, graph });
expect(first).toEqual(second);
expect(first.key).toBe("decision_threshold");
expect(first.reason).toContain("threshold");
});
it("question is singular and answerable", () => {
@@ -206,4 +363,112 @@ describe("formulateQuestion", () => {
"What would resolve uncertainty regarding",
);
});
it("ambiguous contradiction produces a broad distinguishing question without accounting jargon", () => {
const unknown = makeNode({
id: "n-cause-a",
label: "Cash outflow cause",
description: "Unclear explanation for the contradiction.",
kind: "unknown",
status: "unknown",
confidence: "high",
});
const contradiction = makeNode({
id: "n-contradiction",
label: "Divergent movement between revenue and cash",
description: "Two signals moved in opposite directions.",
kind: "relationship",
status: "supported",
confidence: "medium",
});
const revenueObservation = makeNode({
id: "n-revenue-observation",
label: "Revenue increased by 18%.",
description: "Revenue increased by 18%.",
kind: "observation",
status: "supported",
confidence: "high",
});
const cashObservation = makeNode({
id: "n-cash-observation",
label: "Cash in the bank decreased over the same period.",
description: "Cash in the bank decreased over the same period.",
kind: "observation",
status: "supported",
confidence: "high",
});
const graph = makeGraphFor(unknown, {
centralStatement:
"Revenue increased by 18%, but cash in the bank fell over the same period.",
nodes: [contradiction, revenueObservation, cashObservation],
});
const result = formulateTieResolutionQuestion({ graph });
expect(result.question).toBe(
"Were these figures measured on the same basis and at the same scale?",
);
expect(result.comparabilityStatus).toBe("uncertain");
expect(result.question.toLowerCase()).not.toMatch(
/accounts receivable|capex|debt repayments|working capital/,
);
});
it("definition is selected only for genuine definition unknowns", () => {
const unknown = makeNode({
id: "n-definition-only",
label: "Definition of success criteria",
description: "The term is used inconsistently and needs a definition.",
kind: "unknown",
status: "unknown",
confidence: "medium",
});
const result = formulateQuestion({
node: unknown,
graph: makeGraphFor(unknown),
});
expect(result.strategy).toBe("definition");
});
it("an unknown about possible causes does not become a definition question", () => {
const unknown = makeNode({
id: "n-causes",
label: "Possible causes of the divergence",
description: "Several causes may explain the divergence.",
kind: "unknown",
status: "unknown",
confidence: "medium",
});
const result = formulateQuestion({
node: unknown,
graph: makeGraphFor(unknown),
});
expect(result.strategy).toBeNull();
expect(result.question).toBe(
"What would clarify possible causes of the divergence in this situation?",
);
});
it("malformed punctuation is rejected", () => {
const unknown = makeNode({
id: "n-punct",
label: "Magnitude and nature of cash outflows (operating expenses).",
description:
"Magnitude and nature of cash outflows (operating expenses).",
kind: "unknown",
status: "unknown",
confidence: "medium",
});
const result = formulateQuestion({
node: unknown,
graph: makeGraphFor(unknown),
});
expect(result.question).not.toContain("). is true?");
expect(result.question).toBe(
"What would clarify magnitude and nature of cash outflows (operating expenses) in this situation?",
);
});
});
@@ -127,27 +127,27 @@ describe("question priority generalisation", () => {
{
"nodeId": "hire-success-criteria",
"scenario": "Should we hire another engineer?",
"strategy": "decision criterion",
"strategy": "decision_threshold",
},
{
"nodeId": "van-reliability-threshold",
"scenario": "Should we replace the delivery vans?",
"strategy": "decision criterion",
"strategy": "decision_threshold",
},
{
"nodeId": "country-value-threshold",
"scenario": "Should we launch in another country?",
"strategy": "actor/customer",
"strategy": "decision_threshold",
},
{
"nodeId": "project-benefit-threshold",
"scenario": "Should we continue a project that is over budget?",
"strategy": "decision criterion",
"strategy": "decision_threshold",
},
{
"nodeId": "support-value-threshold",
"scenario": "Should we introduce a paid support tier?",
"strategy": "actor/customer",
"strategy": "baseline_reconstruction",
},
]
`);
@@ -0,0 +1,303 @@
import { describe, expect, it } from "vitest";
import {
formulateQuestion,
formulateTieResolutionQuestion,
} from "@/lib/graph/question-formulator.js";
import { makeEdge, makeGraph, makeNode } from "@/lib/graph/schema.js";
import {
explainUnknownSelection,
selectActiveUnknownCandidate,
} from "@/lib/graph/utils.js";
function buildLiveShapedGraph() {
const summary = makeNode({
id: "nnvog0y",
label:
"Revenue grew by 18% while corporate cash reserves declined over an identical time frame.",
description: "Summary of the situation from the scenario text",
kind: "state",
status: "provisional",
confidence: "medium",
});
const revenueObservation = makeNode({
id: "nri36w9",
label: "Revenue increased by 18%.",
description: "Revenue increased by 18%.",
kind: "observation",
status: "supported",
confidence: "high",
evidenceIds: ["obs_rev"],
});
const cashObservation = makeNode({
id: "nnfc48j",
label: "Cash in the bank decreased over the same period.",
description: "Cash in the bank decreased over the same period.",
kind: "observation",
status: "supported",
confidence: "high",
evidenceIds: ["obs_cash"],
});
const revenueMetric = makeNode({
id: "nhsd6d5",
label: "Revenue metric (typically accrual-based income statement figure)",
description:
"Revenue metric (typically accrual-based income statement figure)",
kind: "metric",
status: "known",
confidence: "high",
});
const cashMetric = makeNode({
id: "neh5m6m",
label:
"Cash balance (liquidity measure on the balance sheet or cash flow statement)",
description:
"Cash balance (liquidity measure on the balance sheet or cash flow statement)",
kind: "metric",
status: "known",
confidence: "high",
});
const directionalRelationship = makeNode({
id: "nwo6070",
label:
"Divergent directional movement between top-line revenue growth and net cash position contraction.",
description:
"Divergent directional movement between top-line revenue growth and net cash position contraction.",
kind: "relationship",
status: "supported",
confidence: "high",
});
const contradictionRelationship = makeNode({
id: "nuiab02",
label:
"Apparent contradiction between profitability/revenue expansion and liquidity reduction.",
description:
"Apparent contradiction between profitability/revenue expansion and liquidity reduction.",
kind: "relationship",
status: "supported",
confidence: "medium",
});
const cashTiming = makeNode({
id: "niewza",
label:
"Whether revenue recognition timing differs from cash collection timing.",
description:
"Whether revenue recognition timing differs from cash collection timing.",
kind: "unknown",
status: "unknown",
confidence: "high",
});
const cashOutflows = makeNode({
id: "nqdzobz",
label:
"Magnitude and nature of cash outflows (operating expenses, debt repayments, capex, or working capital shifts).",
description:
"Magnitude and nature of cash outflows (operating expenses, debt repayments, capex, or working capital shifts).",
kind: "unknown",
status: "unknown",
confidence: "high",
});
const edges = [
makeEdge({
id: "e-revenue-summary",
fromNodeId: revenueObservation.id,
toNodeId: summary.id,
relationship: "supports",
description: "Revenue increase supports the scenario summary.",
}),
makeEdge({
id: "e-cash-summary",
fromNodeId: cashObservation.id,
toNodeId: summary.id,
relationship: "supports",
description: "Cash decline supports the scenario summary.",
}),
makeEdge({
id: "e-unk-niewza",
fromNodeId: cashTiming.id,
toNodeId: summary.id,
relationship: "depends_on",
description:
"Whether revenue recognition timing differs from cash collection timing. is an unresolved factor for this situation",
}),
makeEdge({
id: "e-unk-nqdzobz",
fromNodeId: cashOutflows.id,
toNodeId: summary.id,
relationship: "depends_on",
description:
"Magnitude and nature of cash outflows (operating expenses, debt repayments, capex, or working capital shifts). is an unresolved factor for this situation",
}),
];
return makeGraph({
centralStatement:
"Revenue increased by 18%, but cash in the bank fell over the same period.",
nodes: [
summary,
revenueObservation,
cashObservation,
revenueMetric,
cashMetric,
directionalRelationship,
contradictionRelationship,
cashTiming,
cashOutflows,
],
edges,
activeUnknownNodeId: null,
resolvedNodeIds: [],
currentSummary: "Diagnostic selection influence fixture",
});
}
function orderCandidates(explanation) {
return explanation.candidates.map((candidate) => ({
nodeId: candidate.nodeId,
label: candidate.label,
score: candidate.score,
downstreamCount: candidate.downstreamCount,
unresolvedParentUnknownCount: candidate.unresolvedParentUnknownCount,
}));
}
function removeDependencyLinks(graph) {
const nodes = graph.nodes.map((node) => ({
...node,
dependsOn: [],
affects: [],
parentId: null,
childIds: [],
}));
const edges = (graph.edges || []).filter(
(edge) => edge.relationship !== "depends_on",
);
return makeGraph({ ...graph, nodes, edges, activeUnknownNodeId: null });
}
function neutraliseUnknownWording(graph) {
let counter = 0;
const nodes = graph.nodes.map((node) => {
if (node.kind !== "unknown") return { ...node };
counter += 1;
return {
...node,
label: `Unknown ${String.fromCharCode(64 + counter)}`,
description: `Unknown factor ${counter} relevant to the scenario.`,
};
});
return makeGraph({ ...graph, nodes, activeUnknownNodeId: null });
}
describe("selection influence diagnostic", () => {
it("records ambiguous ordering changes for live-shaped, structure-only, and wording-neutralised fixtures", () => {
const liveGraph = buildLiveShapedGraph();
const liveExplanation = explainUnknownSelection(liveGraph, []);
const liveSelection = selectActiveUnknownCandidate(liveGraph, []);
const tieQuestion = formulateTieResolutionQuestion({ graph: liveGraph });
const noLinksExplanation = explainUnknownSelection(
removeDependencyLinks(liveGraph),
[],
);
const noLinksSelection = selectActiveUnknownCandidate(
removeDependencyLinks(liveGraph),
[],
);
const neutralWordingExplanation = explainUnknownSelection(
neutraliseUnknownWording(liveGraph),
[],
);
const neutralSelection = selectActiveUnknownCandidate(
neutraliseUnknownWording(liveGraph),
[],
);
const fallbackQuestion = formulateQuestion({
node: liveGraph.nodes.find((node) => node.id === "nqdzobz"),
graph: liveGraph,
});
const diagnosticRecord = {
liveStatus: liveExplanation.status,
liveShapedCandidateOrdering: orderCandidates(liveExplanation),
liveTiedCandidateIds: liveExplanation.tiedCandidateIds,
noLinksCandidateOrdering: orderCandidates(noLinksExplanation),
noLinksStatus: noLinksExplanation.status,
neutralWordingCandidateOrdering: orderCandidates(
neutralWordingExplanation,
),
neutralStatus: neutralWordingExplanation.status,
selectedExplanationContributions: liveExplanation.selected?.contributions,
tieQuestion: tieQuestion.question,
liveSelection,
noLinksSelection,
neutralSelection,
fallbackQuestion,
};
expect(diagnosticRecord.liveStatus).toBe("ambiguous");
expect(diagnosticRecord.liveTiedCandidateIds).toEqual([
"nqdzobz",
"niewza",
]);
expect(diagnosticRecord.liveShapedCandidateOrdering).toEqual([
{
nodeId: "nqdzobz",
label:
"Magnitude and nature of cash outflows (operating expenses, debt repayments, capex, or working capital shifts).",
score: 0,
downstreamCount: 0,
unresolvedParentUnknownCount: 0,
},
{
nodeId: "niewza",
label:
"Whether revenue recognition timing differs from cash collection timing.",
score: 0,
downstreamCount: 0,
unresolvedParentUnknownCount: 0,
},
]);
expect(diagnosticRecord.liveSelection).toMatchObject({
selectedNode: null,
status: "ambiguous",
tieType: "complete_unresolved_tie",
tiedCandidateIds: ["nqdzobz", "niewza"],
});
expect(diagnosticRecord.noLinksCandidateOrdering).toEqual(
diagnosticRecord.liveShapedCandidateOrdering,
);
expect(diagnosticRecord.noLinksStatus).toBe("ambiguous");
expect(diagnosticRecord.noLinksSelection.status).toBe("ambiguous");
expect(diagnosticRecord.neutralWordingCandidateOrdering).toEqual([
{
nodeId: "niewza",
label: "Unknown A",
score: 0,
downstreamCount: 0,
unresolvedParentUnknownCount: 0,
},
{
nodeId: "nqdzobz",
label: "Unknown B",
score: 0,
downstreamCount: 0,
unresolvedParentUnknownCount: 0,
},
]);
expect(diagnosticRecord.neutralStatus).toBe("ambiguous");
expect(diagnosticRecord.neutralSelection.status).toBe("ambiguous");
expect(diagnosticRecord.selectedExplanationContributions).toBeUndefined();
expect(diagnosticRecord.tieQuestion).toBe(
"Were these figures measured on the same basis and at the same scale?",
);
expect(diagnosticRecord.tieQuestion.toLowerCase()).not.toMatch(
/accounts receivable|capex|debt repayments|working capital/,
);
expect(diagnosticRecord.fallbackQuestion.strategy).toBeNull();
expect(diagnosticRecord.fallbackQuestion.question).toBe(
"What would clarify magnitude and nature of cash outflows (operating expenses, debt repayments, capex, or working capital shifts) in this situation?",
);
});
});
+419
View File
@@ -0,0 +1,419 @@
import { describe, expect, it } from "vitest";
import { applyValidatedProposal } from "@/lib/graph/apply-proposal.js";
import { makeEdge, makeGraph, makeNode } from "@/lib/graph/schema.js";
function makePropagationFixture({
key,
centralStatement,
firstObservationLabel,
secondObservationLabel,
}) {
const parent = makeNode({
id: `${key}-parent`,
label: `Explanation for why ${centralStatement}`,
description:
"Need to understand what change or event could explain why these observations differ, because that is needed to investigate their relationship.",
kind: "unknown",
status: "unknown",
confidence: "medium",
});
const measurementChild = makeNode({
id: `${key}-child-measurement`,
label: "How the two observations were measured",
description: `Need evidence about the measure used for each observation, because that could help explain ${centralStatement}.`,
kind: "unknown",
status: "unknown",
confidence: "medium",
parentId: parent.id,
});
const timingChild = makeNode({
id: `${key}-child-timing`,
label: "Whether the two observations reflect different timing",
description: `Need to know whether the two observations reflect different timing, because that could help explain ${centralStatement}.`,
kind: "unknown",
status: "unknown",
confidence: "medium",
parentId: parent.id,
});
const cashMovementChild = makeNode({
id: `${key}-child-cash-movement`,
label: `Possible change mainly affecting ${secondObservationLabel}`,
description: `Need to know whether a possible change mainly affected ${secondObservationLabel}, because that could help explain ${centralStatement}.`,
kind: "unknown",
status: "unknown",
confidence: "medium",
parentId: parent.id,
});
const oneOffChild = makeNode({
id: `${key}-child-one-off`,
label: "Possible one-off event during the period",
description: `Need to know whether a possible one-off event happened during the period, because that could help explain ${centralStatement}.`,
kind: "unknown",
status: "unknown",
confidence: "medium",
parentId: parent.id,
});
const ancestor = makeNode({
id: `${key}-ancestor`,
label: `Reasoning for ${centralStatement}`,
description:
"Higher-level reasoning node depending on the parent explanation.",
kind: "unknown",
status: "unknown",
confidence: "medium",
childIds: [parent.id],
});
const unrelated = makeNode({
id: `${key}-unrelated`,
label: "Unrelated branch",
description: "Should remain unchanged.",
kind: "unknown",
status: "unknown",
confidence: "low",
});
const firstObservation = makeNode({
id: `${key}-obs-1`,
label: firstObservationLabel,
description: firstObservationLabel,
kind: "observation",
status: "supported",
confidence: "high",
});
const secondObservation = makeNode({
id: `${key}-obs-2`,
label: secondObservationLabel,
description: secondObservationLabel,
kind: "observation",
status: "supported",
confidence: "high",
});
const graph = makeGraph({
centralStatement,
nodes: [
ancestor,
parent,
measurementChild,
timingChild,
cashMovementChild,
oneOffChild,
unrelated,
firstObservation,
secondObservation,
],
edges: [
makeEdge({
id: `${key}-e-parent-ancestor`,
fromNodeId: parent.id,
toNodeId: ancestor.id,
relationship: "depends_on",
description: "Ancestor depends on the parent explanation.",
}),
makeEdge({
id: `${key}-e-child-measurement-parent`,
fromNodeId: measurementChild.id,
toNodeId: parent.id,
relationship: "depends_on",
description: "Measurement child depends into the parent explanation.",
}),
makeEdge({
id: `${key}-e-child-timing-parent`,
fromNodeId: timingChild.id,
toNodeId: parent.id,
relationship: "depends_on",
description: "Timing child depends into the parent explanation.",
}),
makeEdge({
id: `${key}-e-child-cash-parent`,
fromNodeId: cashMovementChild.id,
toNodeId: parent.id,
relationship: "depends_on",
description: "Cash-movement child depends into the parent explanation.",
}),
makeEdge({
id: `${key}-e-child-one-off-parent`,
fromNodeId: oneOffChild.id,
toNodeId: parent.id,
relationship: "depends_on",
description: "One-off child depends into the parent explanation.",
}),
],
activeUnknownNodeId: measurementChild.id,
resolvedNodeIds: [],
currentSummary: `Propagation fixture for ${key}`,
});
return {
graph,
ids: {
ancestor: ancestor.id,
parent: parent.id,
measurementChild: measurementChild.id,
timingChild: timingChild.id,
cashMovementChild: cashMovementChild.id,
oneOffChild: oneOffChild.id,
unrelated: unrelated.id,
},
};
}
const scenarios = [
{
key: "revenue-cash",
centralStatement: "revenue increased while cash fell",
firstObservationLabel: "Revenue increased by 18%.",
secondObservationLabel: "Cash in the bank fell over the same period.",
},
{
key: "satisfaction-complaints",
centralStatement:
"customer satisfaction increased while complaints increased",
firstObservationLabel: "Customer satisfaction increased.",
secondObservationLabel: "Complaints increased.",
},
{
key: "traffic-sales",
centralStatement: "traffic increased while sales stayed flat",
firstObservationLabel: "Website traffic increased.",
secondObservationLabel: "Sales stayed flat.",
},
{
key: "delivery-cancellations",
centralStatement: "delivery time fell while cancellations increased",
firstObservationLabel: "Average delivery time decreased.",
secondObservationLabel: "Cancellations increased.",
},
{
key: "production-defects",
centralStatement: "production increased while defects increased",
firstObservationLabel: "Production increased.",
secondObservationLabel: "Defects increased.",
},
];
describe("upward propagation", () => {
it.each(scenarios)(
"propagates resolved measurement child upward for $key",
({
key,
centralStatement,
firstObservationLabel,
secondObservationLabel,
}) => {
const { graph, ids } = makePropagationFixture({
key,
centralStatement,
firstObservationLabel,
secondObservationLabel,
});
const unrelatedBefore = JSON.stringify(
graph.nodes.find((node) => node.id === ids.unrelated),
);
const result = applyValidatedProposal({
situationGraph: graph,
proposal: {
addedNodes: [
makeNode({
id: `${key}-anchor`,
label: "Update anchor",
description:
"Anchor state introduced by the answer because the update must contain a meaningful change.",
kind: "state",
status: "known",
confidence: "low",
}),
],
updatedNodes: [
{
nodeId: ids.measurementChild,
previousStatus: "unknown",
newStatus: "resolved",
previousValue: null,
newValue:
"The figures were measured over the same accounting period using the same management accounts.",
reason: "The answer resolves the measurement child.",
},
],
addedEdges: [],
removedEdgeIds: [],
resolvedUnknownNodeIds: [ids.measurementChild],
affectedNodeIds: [],
selectedQuestion: null,
},
previousQuestion:
"What evidence would clarify how the two observations were measured?",
answer:
"The figures were measured over the same accounting period using the same management accounts.",
});
expect(result.success).toBe(true);
expect(result.resolvedUnknownNodeIds).toContain(ids.measurementChild);
expect(result.propagationPerformed).toBe(true);
expect(result.resolvedChildNodeId).toBe(ids.measurementChild);
expect(result.parentNodeId).toBe(ids.parent);
expect(result.parentStatusBefore).toBe("unknown");
expect(result.parentStatusAfter).toBe("provisional");
expect(result.parentConfidenceBefore).toBe("medium");
expect(result.parentConfidenceAfter).toBe("medium");
expect(result.evidenceConfidenceBefore).toBe("medium");
expect(result.evidenceConfidenceAfter).toBe("medium");
expect(result.completenessBefore).toBe("empty");
expect(result.completenessAfter).toBe("partial");
expect(result.conclusionConfidenceBefore).toBe("low");
expect(result.conclusionConfidenceAfter).toBe("medium");
expect(result.confidenceCapReason).toBe(
"unresolved_direct_children_cap_conclusion",
);
expect(result.parentResolved).toBe(false);
expect(result.affectedAncestorIds).toContain(ids.parent);
expect(result.affectedAncestorIds).toContain(ids.ancestor);
expect(result.nextSelectedSibling).toBe(result.newActiveUnknownNodeId);
expect(result.nextSelectedSibling).toBe(result.selectedQuestion?.nodeId);
expect(result.nextSelectedSibling).not.toBe(ids.measurementChild);
expect([
ids.timingChild,
ids.cashMovementChild,
ids.oneOffChild,
]).toContain(result.nextSelectedSibling);
expect(result.selectedQuestion?.question.toLowerCase()).not.toContain(
"measured",
);
const parentNode = result.updatedSituationGraph.nodes.find(
(node) => node.id === ids.parent,
);
expect(parentNode).toMatchObject({
status: "provisional",
confidence: "medium",
confidenceAssessment: {
evidenceConfidence: "medium",
completenessStatus: "partial",
conclusionConfidence: "medium",
},
});
const ancestorNode = result.updatedSituationGraph.nodes.find(
(node) => node.id === ids.ancestor,
);
expect(ancestorNode).toMatchObject({
status: "provisional",
confidence: "low",
confidenceAssessment: {
evidenceConfidence: "low",
completenessStatus: "empty",
conclusionConfidence: "low",
},
});
const resolvedChild = result.updatedSituationGraph.nodes.find(
(node) => node.id === ids.measurementChild,
);
expect(resolvedChild.status).toBe("resolved");
expect(resolvedChild.evidenceIds).toContain(
`answer:${ids.measurementChild}`,
);
expect(
result.updatedSituationGraph.nodes.filter(
(node) => node.id === ids.measurementChild,
),
).toHaveLength(1);
expect(
JSON.stringify(
result.updatedSituationGraph.nodes.find(
(node) => node.id === ids.unrelated,
),
),
).toBe(unrelatedBefore);
},
);
it("resolves the parent only after all direct children are resolved", () => {
const { graph, ids } = makePropagationFixture({
key: "completion-rule",
centralStatement: "revenue increased while cash fell",
firstObservationLabel: "Revenue increased by 18%.",
secondObservationLabel: "Cash in the bank fell over the same period.",
});
const result = applyValidatedProposal({
situationGraph: graph,
proposal: {
addedNodes: [
makeNode({
id: "completion-rule-anchor",
label: "Update anchor",
description:
"Anchor state introduced by the answer because the update must contain a meaningful change.",
kind: "state",
status: "known",
confidence: "low",
}),
],
updatedNodes: [
{
nodeId: ids.measurementChild,
previousStatus: "unknown",
newStatus: "resolved",
previousValue: null,
newValue: "same management accounts",
reason: "resolved measurement child",
},
{
nodeId: ids.timingChild,
previousStatus: "unknown",
newStatus: "resolved",
previousValue: null,
newValue: "timing aligned",
reason: "resolved timing child",
},
{
nodeId: ids.cashMovementChild,
previousStatus: "unknown",
newStatus: "resolved",
previousValue: null,
newValue: "cash left through operations",
reason: "resolved movement child",
},
{
nodeId: ids.oneOffChild,
previousStatus: "unknown",
newStatus: "resolved",
previousValue: null,
newValue: "no exceptional movement",
reason: "resolved one-off child",
},
],
addedEdges: [],
removedEdgeIds: [],
resolvedUnknownNodeIds: [
ids.measurementChild,
ids.timingChild,
ids.cashMovementChild,
ids.oneOffChild,
],
affectedNodeIds: [],
selectedQuestion: null,
},
previousQuestion:
"What evidence would clarify how the two observations were measured?",
answer: "All direct child questions are now answered.",
});
expect(result.success).toBe(true);
expect(result.parentResolved).toBe(true);
expect(result.resolvedUnknownNodeIds).toContain(ids.parent);
expect(
result.updatedSituationGraph.nodes.find((node) => node.id === ids.parent),
).toMatchObject({
status: "resolved",
confidence: "high",
confidenceAssessment: {
evidenceConfidence: "high",
completenessStatus: "complete",
conclusionConfidence: "high",
},
});
});
});
+67
View File
@@ -456,6 +456,7 @@ describe("selectActiveUnknownCandidate", () => {
const result = selectActiveUnknownCandidate(graph, []);
expect(result.nodeId).toBe("unknown-a"); // Has more dependents (score 2 vs 0)
expect(result.status).toBe("selected");
});
it("returns one candidate (not array)", () => {
@@ -619,6 +620,72 @@ describe("selectActiveUnknownCandidate", () => {
const childScore = scoreUnknownCandidate(graph, childUnknown, []);
expect(parentScore.score).toBeGreaterThan(childScore.score);
});
it("returns ambiguous for a complete unresolved tie instead of label-based winner", () => {
const unknownA = makeNode({
id: "tie-a",
label: "Magnitude and nature of cash outflows",
description: "Magnitude and nature of cash outflows.",
kind: "unknown",
status: "unknown",
confidence: "high",
});
const unknownB = makeNode({
id: "tie-b",
label:
"Whether revenue recognition timing differs from cash collection timing",
description:
"Whether revenue recognition timing differs from cash collection timing.",
kind: "unknown",
status: "unknown",
confidence: "high",
});
const graph = makeGraph({
centralStatement:
"Revenue increased by 18%, but cash in the bank fell over the same period.",
nodes: [unknownA, unknownB],
edges: [],
activeUnknownNodeId: null,
resolvedNodeIds: [],
currentSummary: "Tie case",
});
const result = selectActiveUnknownCandidate(graph, []);
expect(result).toMatchObject({
selectedNode: null,
status: "ambiguous",
tieType: "complete_unresolved_tie",
tiedCandidateIds: ["tie-a", "tie-b"],
});
expect(result.nodeId).toBeUndefined();
});
it("alphabetical renaming does not resolve a complete tie", () => {
const unknownA = makeNode({
id: "tie-a",
label: "Unknown B",
description: "Unknown factor one.",
kind: "unknown",
});
const unknownB = makeNode({
id: "tie-b",
label: "Unknown A",
description: "Unknown factor two.",
kind: "unknown",
});
const graph = makeGraph({
centralStatement: "Two conflicting signals remain unresolved.",
nodes: [unknownA, unknownB],
edges: [],
activeUnknownNodeId: null,
resolvedNodeIds: [],
currentSummary: "Tie case",
});
const result = selectActiveUnknownCandidate(graph, []);
expect(result.status).toBe("ambiguous");
expect(result.tiedCandidateIds.sort()).toEqual(["tie-a", "tie-b"]);
});
});
describe("applyGraphUpdate", () => {
+156 -21
View File
@@ -80,7 +80,7 @@ function makeUpdateSuccess(overrides = {}) {
updatedSituationGraph: {
centralStatement: "Complaints increased while production increased.",
currentSummary: "Updated summary",
activeUnknownNodeId: "n-next-unknown",
activeUnknownNodeId: "n-child-1",
resolvedNodeIds: ["n-unknown"],
nodes: [
{
@@ -115,32 +115,83 @@ function makeUpdateSuccess(overrides = {}) {
},
{
id: "n-next-unknown",
label: "Commercial value definition",
description: "Need a definition because the decision depends on it.",
label:
"Explanation for why revenue increased by 18%, but cash in the bank fell over the same period",
description:
"Need to understand what change or event could explain why these observations differ, because that is needed to investigate their relationship.",
kind: "unknown",
status: "unknown",
confidence: "high",
confidence: "medium",
value: null,
unit: null,
},
{
id: "n-child-1",
label: "How the two observations were measured",
description:
"Need evidence about the measure used for each observation, because that could help explain revenue increased by 18%, but cash in the bank fell over the same period.",
kind: "unknown",
status: "unknown",
confidence: "medium",
value: null,
unit: null,
parentId: "n-next-unknown",
},
],
edges: [
{
id: "e-rel-next",
fromNodeId: "n-conclusion",
toNodeId: "n-next-unknown",
relationship: "depends_on",
confidence: "medium",
description:
"This unresolved explanation arises from the now-assessed relationship between the observations.",
},
{
id: "e-child-next",
fromNodeId: "n-child-1",
toNodeId: "n-next-unknown",
relationship: "depends_on",
confidence: "medium",
description:
"This child unknown must be investigated before the broader parent explanation can be resolved.",
},
],
edges: [],
},
proposal: {
addedNodes: [
{
id: "n-next-unknown",
label: "Commercial value definition",
description: "Need a definition because the decision depends on it.",
label:
"Explanation for why revenue increased by 18%, but cash in the bank fell over the same period",
description:
"Need to understand what change or event could explain why these observations differ, because that is needed to investigate their relationship.",
kind: "unknown",
status: "unknown",
confidence: "high",
confidence: "medium",
value: null,
unit: null,
evidenceIds: [],
dependsOn: ["n-conclusion"],
affects: [],
parentId: "n-conclusion",
childIds: [],
},
{
id: "n-child-1",
label: "How the two observations were measured",
description:
"Need evidence about the measure used for each observation, because that could help explain revenue increased by 18%, but cash in the bank fell over the same period.",
kind: "unknown",
status: "unknown",
confidence: "medium",
value: null,
unit: null,
evidenceIds: [],
dependsOn: [],
affects: [],
parentId: null,
parentId: "n-next-unknown",
childIds: [],
},
],
@@ -152,20 +203,68 @@ function makeUpdateSuccess(overrides = {}) {
resolvedUnknownNodeIds: ["n-unknown"],
affectedNodeIds: ["n-conclusion"],
selectedQuestion: {
nodeId: "n-next-unknown",
question: "How should commercial value be defined for this decision?",
reason: "A narrower consequential uncertainty remains.",
nodeId: "n-child-1",
question:
"What evidence would clarify how the two observations were measured?",
reason:
"Formulated from graph context using the evidence_gathering investigation strategy.",
},
},
selectedQuestion: {
nodeId: "n-next-unknown",
question: "How should commercial value be defined for this decision?",
reason: "A narrower consequential uncertainty remains.",
nodeId: "n-child-1",
question:
"What evidence would clarify how the two observations were measured?",
reason:
"Formulated from graph context using the evidence_gathering investigation strategy.",
},
affectedNodeIds: ["n-conclusion"],
resolvedUnknownNodeIds: ["n-unknown"],
previousActiveUnknownNodeId: "n-unknown",
newActiveUnknownNodeId: "n-next-unknown",
newActiveUnknownNodeId: "n-child-1",
emergentReasoningNodeCreated: true,
emergentReasoningNodeId: "n-next-unknown",
emergentReasoningNodeReason:
"Created a new unresolved reasoning unknown so the next justified question is backed by the graph.",
atomicityAssessment: "composite",
decompositionPerformed: true,
childUnknownCount: 1,
childNodeIds: ["n-child-1"],
atomicityReason:
"Decomposed a composite unknown into smaller broad candidate dimensions before asking the next question.",
previousReasoningState: {
comparabilityStatus: "uncertain",
reasoningStages: [
{
stage: "comparability",
status: "uncertain",
outcome:
"Comparability between the observations is not yet established across period, scale, or measurement basis.",
},
{
stage: "relationship",
status: "insufficient_information",
outcome: "not assessed until comparability is established",
},
],
},
reasoningState: {
comparabilityStatus: "confirmed",
relationshipStatus: "insufficient_information",
reasoningStages: [
{
stage: "comparability",
status: "confirmed",
outcome:
"Comparability was confirmed by the user answer covering the same period and source basis.",
},
{
stage: "relationship",
status: "insufficient_information",
outcome:
"There is not enough structure to classify the relationship safely.",
},
],
},
changesApplied: {
updatedNodeCount: 2,
resolvedUnknownCount: 1,
@@ -370,7 +469,9 @@ describe("graph-backed UI rendering", () => {
);
expect(html).toContain("Newly surfaced unknowns");
expect(html).toContain("Commercial value definition");
expect(html).toContain(
"Explanation for why revenue increased by 18%, but cash in the bank fell over the same period",
);
});
it("affected nodes render", () => {
@@ -398,7 +499,7 @@ describe("graph-backed UI rendering", () => {
);
expect(html).toContain(
"How should commercial value be defined for this decision?",
"What evidence would clarify how the two observations were measured?",
);
});
@@ -435,7 +536,9 @@ describe("graph-backed UI rendering", () => {
expect(html).toContain("Previous active unknown");
expect(html).toContain("Complaint rate denominator");
expect(html).toContain("New active unknown");
expect(html).toContain("Commercial value definition");
expect(html).toContain(
"Explanation for why revenue increased by 18%, but cash in the bank fell over the same period",
);
});
it("successful update renders prior and new state together", () => {
@@ -454,10 +557,40 @@ describe("graph-backed UI rendering", () => {
expect(html).toContain("New active unknown");
expect(html).toContain("Next question");
expect(html).toContain(
"How should commercial value be defined for this decision?",
"What evidence would clarify how the two observations were measured?",
);
});
it("update view shows comparability progression without raw ids in the normal view", () => {
const html = renderToStaticMarkup(
<GraphUpdateView
updateResult={{
...makeUpdateSuccess({
selectedQuestion: {
nodeId: "n-next-unknown",
question:
"What evidence would clarify timing or measurement basis?",
reason: "A broad follow-up is now justified.",
},
}),
previousSituationGraph: makeGraphResult().situationGraph,
}}
/>,
);
expect(html).toContain("Comparability:");
expect(html).toContain("uncertain → confirmed");
expect(html).toContain("Relationship status:");
expect(html).toContain("insufficient_information");
expect(html).toContain("Reasoning stages:");
expect(html).toContain("comparability: confirmed");
expect(html).toContain("relationship: insufficient_information");
expect(html).toContain(
"What evidence would clarify how the two observations were measured?",
);
expect(html).not.toContain("reasoning:comparability");
});
it("situation graph marks newly surfaced and active unknowns", () => {
const html = renderToStaticMarkup(
<SituationGraphView
@@ -479,7 +612,9 @@ describe("graph-backed UI rendering", () => {
/>,
);
expect(html).toContain("How should commercial value be defined for this decision?");
expect(html).toContain(
"What evidence would clarify how the two observations were measured?",
);
});
it("raw ids remain only in collapsed proposal details", () => {