Evidence Conflict Detection

Context Block

Page Type: Evidence System Layer
Function: Conflict Identification Engine
Position: After Evidence Classification and Scoring
Role: Detects inconsistencies across evidence sets

This layer ensures the system does not treat all evidence as coherent. It explicitly surfaces disagreement between sources before aggregation or ranking.

Core Objective

• Detect contradictions across evidence sources
• Identify semantic inconsistencies within datasets
• Flag conflicting truth claims early in pipeline
• Prevent silent merging of incompatible evidence
• Prepare input for conflict resolution layer

Conflict Detection Pipeline

1. Semantic Comparison
Compares meaning across evidence units for alignment or divergence.

2. Entity-Level Conflict Scan
Detects contradictions on shared entities (people, systems, metrics).

3. Numerical Discrepancy Analysis
Identifies mismatches in quantitative data points.

4. Temporal Consistency Check
Validates whether time-based claims align or conflict.

5. Conflict Classification Output
Labels type and severity of detected conflicts.

Conflict Types

• Hard Conflict — direct contradiction of facts
• Soft Conflict — interpretational differences
• Temporal Conflict — outdated vs updated information
• Contextual Conflict — same data, different domain framing

Detection Signals

• Opposing factual statements
• Mismatch in entity attributes
• Statistical deviation beyond threshold
• Ontology-level inconsistency
• Cross-source disagreement patterns

Example Conflict

Evidence A: SEO update improves ranking stability

Evidence B: SEO update causes ranking volatility

Detected: Hard Conflict (causal contradiction)

Integration in GEO Pipeline

• Evidence Scoring
• Evidence Ranking
• Evidence Conflict Resolution
• Evidence Validation

Evidence Conflict Detection is the prerequisite layer for structured reconciliation of inconsistent truth signals.

Failure Modes

• Missed subtle semantic contradictions
• False conflict detection in compatible evidence
• Over-triggering conflict flags on minor variations
• Ignoring context-dependent truth differences

Structured Output Model

Each detection cycle produces:

• Conflict Map (between evidence units)
• Conflict Type Classification
• Severity Score
• Affected Entities List
• Pre-resolution Dependency Graph

Relationship Block

Parent Layer: /evidence/
Upstream: Evidence Scoring, Evidence Classification
Downstream: Evidence Conflict Resolution, Evidence Validation
Connected Systems: Ontology Layer, Retrieval Engine, Answer System

Structured Summary

Evidence Conflict Detection is the inconsistency discovery layer of the Evidence system. It identifies contradictions across evidence sources and prepares structured conflict maps for downstream resolution.

This layer ensures that conflicting truth signals are explicitly surfaced instead of silently merged into incorrect conclusions.