Last modified by Robert Schaub on 2026/02/08 08:31

Show last authors
1 {{info}}
2 **Current Implementation (v2.6.33)** - Scenarios have been replaced by **KeyFactors** - optional decomposition questions discovered during the understanding phase.
3 {{/info}}
4
5 = Claim Analysis Workflow =
6
7 {{mermaid}}
8
9 graph TB
10 Start[User Submission]
11
12 subgraph Step1[Step 1 Understand]
13 Extract{understandClaim LLM Analysis}
14 Gate1{Gate 1 Claim Validation}
15 DetectType[Detect Input Type]
16 DetectScopes[Detect Scopes]
17 KeyFactors[Discover KeyFactors]
18 end
19
20 subgraph Step2[Step 2 Research]
21 Decide[decideNextResearch]
22 Search[Web Search]
23 Fetch[Fetch Sources]
24 Facts[extractFacts]
25 end
26
27 subgraph Step3[Step 3 Verdict]
28 Verdict[generateVerdicts]
29 Gate4{Gate 4 Confidence Check}
30 end
31
32 subgraph Output[Output]
33 Publish[Publish Result]
34 LowConf[Low Confidence Flag]
35 end
36
37 Start --> Extract
38 Extract --> Gate1
39 Gate1 -->|Pass Factual| DetectType
40 Gate1 -->|Fail Opinion| Exclude[Exclude from analysis]
41 DetectType --> DetectScopes
42 DetectScopes --> KeyFactors
43 KeyFactors --> Decide
44 Decide --> Search
45 Search --> Fetch
46 Fetch --> Facts
47 Facts -->|More research needed| Decide
48 Facts -->|Sufficient evidence| Verdict
49 Verdict --> Gate4
50 Gate4 -->|High or Medium confidence| Publish
51 Gate4 -->|Low or Insufficient| LowConf
52
53 {{/mermaid}}
54
55 == Quality Gates (Implemented) ==
56
57 |= Gate |= Name |= Purpose |= Pass Criteria
58 | **Gate 1** | Claim Validation | Filter non-factual claims | Factual, opinion score 0.3 or less, specificity 0.3 or more
59 | **Gate 4** | Verdict Confidence | Ensure sufficient evidence | 2 or more sources, avg quality 0.6 or more, agreement 60% or more
60
61 //Gates 2 (Contradiction Search) and 3 (Uncertainty Quantification) are not yet implemented.//
62
63 == KeyFactors (Replaces Scenarios) ==
64
65 **KeyFactors** are optional decomposition questions discovered during the understanding phase:
66 * Not stored as separate entities
67 * Help break down complex claims into checkable sub-questions
68 * See ##Docs/ARCHITECTURE/KeyFactors_Design.md## for design rationale
69
70 == 7-Point Verdict Scale ==
71
72 * **TRUE** (86-100%) - Claim is well-supported by evidence
73 * **MOSTLY-TRUE** (72-85%) - Largely accurate with minor caveats
74 * **LEANING-TRUE** (58-71%) - More evidence supports than contradicts
75 * **MIXED** (43-57%, high confidence) - Roughly equal evidence both ways
76 * **UNVERIFIED** (43-57%, low confidence) - Insufficient evidence to determine
77 * **LEANING-FALSE** (29-42%) - More evidence contradicts than supports
78 * **MOSTLY-FALSE** (15-28%) - Largely inaccurate
79 * **FALSE** (0-14%) - Claim is refuted by evidence