skylv-metacognition-engine

AI Agent metacognition engine. "Thinking about thinking" — detect cognitive biases, enable self-correction.

Problem

AI agents can have cognitive biases just like humans:

• Confirmation bias (seeking only supporting evidence)
• Overconfidence (being too certain)
• Anchoring (relying too much on first information)
• Sunk cost fallacy (continuing wrong decisions due to past investment)

Solution

Metacognition Engine — agents analyze their own reasoning process, detect biases, and self-correct.

Usage

# Analyze reasoning file
node metacognition_engine.js analyze reasoning.txt

# Reflect on a decision
node metacognition_engine.js reflect "I'm definitely sure this must be correct"

# List bias types
node metacognition_engine.js biases

Cognitive Biases Detected (6)

Bias	Description	Detection
Confirmation Bias	Seeking only supporting evidence	High frequency of "only", "just", "exactly"
Overconfidence	Overestimating accuracy	High frequency of "definitely", "must"
Anchoring Bias	Over-reliance on first info	High frequency of "first", "initial"
Sunk Cost Fallacy	Continuing due to past investment	High frequency of "already", "spent"
Availability Heuristic	Based on easily recalled info	High frequency of "recently", "remember"
Framing Effect	Influenced by how question is framed	Presence of alternative perspectives

Output

• Bias detection with confidence scores
• Self-correction suggestions
• Alternative thinking approaches

Market Data

Blue ocean:

metacognition-skill

scores 0.605 — almost no competition.

---

"The first step to wisdom is knowing what you don't know."