stemedb/uat/consumer-health/disagreement-dashboard.md

UAT: Disagreement Dashboard (Resolved/Active/Emerging)

Date: YYYY-MM-DD Feature: Conflict Status Classification Status: [ ] PASS / [ ] FAIL / [ ] BLOCKED

Scenario

A "Living Review" dashboard needs to categorize assertions by conflict status:

1. Resolved - Had conflict, now resolved (one claim dominates or epoch superseded)
2. Active Disagreement - Ongoing contested claims from authoritative sources
3. Emerging Signal - New anecdotal cluster that may indicate unreported effect

This enables triage: researchers focus on Active Disagreement, regulators monitor Emerging Signals.

Acceptance Criteria

Criterion	Expected	Met?
Resolved items identified	Status = "resolved"	[ ]
Active disagreement identified	Status = "active_disagreement"	[ ]
Emerging signal identified	Status = "emerging_signal"	[ ]
Dashboard summary	Counts per category	[ ]
Drill-down available	Full claims for each	[ ]

Test Matrix

Step	Action	Expected	Actual	Status
1	Create resolved conflict	Consensus reached		[ ]
2	Create active disagreement	Clinical studies conflict		[ ]
3	Create emerging signal	Anecdotal cluster, no authority		[ ]
4	Query dashboard summary	3 categories populated		[ ]
5	Drill into active disagreement	Full claim details		[ ]

Conflict Status Definitions

Status	Criteria
Resolved	conflict_score < 0.1 OR single-tier unanimous OR epoch-superseded
Active Disagreement	conflict_score >= 0.4 AND Tier 0-2 sources present on both sides
Emerging Signal	Tier 5 cluster >= 100 AND no Tier 0-2 coverage
Low Priority	Everything else (minor disagreements in low-tier sources)

Setup Commands

# Start StemeDB
cargo run --bin stemedb-api &
sleep 2

Test Commands

Step 1: Create Resolved Conflict (Dose Adjustment History)

# Old studies said 1mg was max
curl -X POST http://localhost:18180/v1/assertions \
  -H "Content-Type: application/json" \
  -d '{
    "subject": "Semaglutide",
    "predicate": "recommended_max_dose",
    "object": {"Number": 1.0},
    "confidence": 0.9,
    "source_class": "Clinical",
    "epoch": "0000000000000000000000000000000000000000000000000000000000000001"
  }'

# New FDA guidance supersedes (2.4mg approved)
curl -X POST http://localhost:18180/v1/assertions \
  -H "Content-Type: application/json" \
  -d '{
    "subject": "Semaglutide",
    "predicate": "recommended_max_dose",
    "object": {"Number": 2.4},
    "confidence": 1.0,
    "source_class": "Regulatory",
    "epoch": "0000000000000000000000000000000000000000000000000000000000000002"
  }'

Expected: Conflict resolved by epoch supersession Actual: Status: [ ]

Step 2: Create Active Disagreement (Muscle Loss Debate)

# Study A: Significant muscle loss
curl -X POST http://localhost:18180/v1/assertions \
  -H "Content-Type: application/json" \
  -d '{
    "subject": "GLP1:MuscleEffect",
    "predicate": "lean_mass_impact",
    "object": {"Text": "Significant reduction observed"},
    "confidence": 0.85,
    "source_class": "Clinical"
  }'

# Study B: Minimal muscle impact
curl -X POST http://localhost:18180/v1/assertions \
  -H "Content-Type: application/json" \
  -d '{
    "subject": "GLP1:MuscleEffect",
    "predicate": "lean_mass_impact",
    "object": {"Text": "Minimal reduction with exercise"},
    "confidence": 0.82,
    "source_class": "Clinical"
  }'

Expected: Active disagreement - two clinical studies conflict Actual: Status: [ ]

Step 3: Create Emerging Signal (Hair Loss Reports)

# 150 anecdotal reports, no clinical data
for i in $(seq 1 150); do
  HASH=$(printf '%064d' $((5000 + i)))
  curl -s -X POST http://localhost:18180/v1/assertions \
    -H "Content-Type: application/json" \
    -d "{
      \"subject\": \"Semaglutide\",
      \"predicate\": \"hair_thinning_reported\",
      \"object\": {\"Boolean\": true},
      \"confidence\": 0.70,
      \"source_class\": \"Anecdotal\",
      \"source_hash\": \"$HASH\"
    }" > /dev/null
done
echo "Created 150 hair loss reports"

Expected: Emerging signal - anecdotal cluster without authoritative coverage Actual: Status: [ ]

Step 4: Query Dashboard Summary

curl "http://localhost:18180/v1/dashboard/conflicts"

Expected:

{
  "summary": {
    "resolved": 1,
    "active_disagreement": 1,
    "emerging_signal": 1,
    "low_priority": 0
  },
  "items": {
    "resolved": [
      {"subject": "Semaglutide", "predicate": "recommended_max_dose", "resolution": "epoch_superseded"}
    ],
    "active_disagreement": [
      {"subject": "GLP1:MuscleEffect", "predicate": "lean_mass_impact", "conflict_score": 0.88}
    ],
    "emerging_signal": [
      {"subject": "Semaglutide", "predicate": "hair_thinning_reported", "cluster_count": 150}
    ]
  }
}

Actual: Status: [ ]

Step 5: Drill Into Active Disagreement

curl "http://localhost:18180/v1/skeptic?subject=GLP1:MuscleEffect&predicate=lean_mass_impact"

Expected: Full claim breakdown with:

• Both clinical studies listed
• Supporting evidence for each
• Conflict score >= 0.4
• Status = "Contested"

Actual: Status: [ ]

Sign-Off Checklist

• Resolved conflicts identified correctly
• Active disagreements surfaced
• Emerging signals detected
• Dashboard provides summary counts
• Drill-down returns full details

Notes

Dashboard categories are computed at query time, not stored. This ensures freshness but may have performance implications for large datasets.

---

Tester: Date: Result: