01
Detection
At Cross-Path Synthesis (Stage 5), the system compares branch outputs and identifies points where branches reach different conclusions about the same question.
ACIE Protocol
Contradiction Resolution
When parallel reasoning branches reach different conclusions, the ACIE detects the disagreement, classifies its nature, and triggers evidence augmentation. Contradictions are resolved through source priority, not suppressed through averaging.
Interactive demonstration
Watch the ACIE detect and resolve a contradiction.
Parallel Reasoning
Contradiction Engine
Dispatching Branches
Decision Query
Evaluate grid stability risk for Q3 deployment
Branch Alpha
Grid capacity sufficient - 12% headroom at peak load
SCADA telemetry, 90-day rolling average0.87
Dispatching isolated reasoning branches...
The resolution process: four stages.
02
Classification
Each contradiction is classified by type: factual disagreement (branches cite different evidence), inferential disagreement (branches draw different conclusions from the same evidence), or scope disagreement (branches address different aspects of the question).
03
Evidence augmentation
The system retrieves additional evidence from the Evidence Kernel to adjudicate between conflicting branch conclusions. This is not a re-run of the original query - it is a targeted evidence retrieval focused on the specific point of disagreement.
04
Resolution or preservation
If the additional evidence resolves the contradiction, the system produces a unified conclusion with the resolution path documented. If the contradiction cannot be resolved, both positions are preserved in the output with their respective evidence bases and confidence levels.
Why contradiction surfacing is a governance requirement.
In regulated environments, the suppression of contradictory evidence is a compliance failure. If a decision system produces a recommendation based on one interpretation of the evidence while an equally valid interpretation exists and was not surfaced, the decision-maker has been given an incomplete picture. In healthcare, finance, energy, and defense, this is not merely a quality issue - it is a liability issue.
The platform treats contradiction surfacing as a first-class architectural requirement, not as a debugging feature. The ACIE operates at Stage 6 of the processing pipeline, after parallel branches have converged but before quality assessment and output compilation. This ensures that contradictions are resolved or preserved before the output enters the governance layer.
Contradiction types
Factual
Branches cite different evidence sources that support different conclusions. Resolution: retrieve the primary source and determine which evidence is authoritative.
Inferential
Branches draw different conclusions from the same evidence. Resolution: examine the reasoning chain for logical gaps or unstated assumptions.
Scope
Branches address different aspects of the question, producing conclusions that appear contradictory but are actually complementary. Resolution: reframe the output to capture both aspects.
See how the platform compares.
Competitive positioning against existing AI decision systems across governance depth and reasoning capability.