Methodology

ReasonRank measures comparative reasoning quality. Results indicate which side appears more rigorously argued under this framework. Results do not determine absolute truth.

Arena Scoring

• Users compare original and counter reasoning and vote on stronger reasoning structure.
• Vote totals are published transparently for each claim URL.
• Confidence level reflects vote stability and sample size.
• Confidence labels: low (<20 votes), medium (20-99), high (100+).
• Confidence is not a factual accuracy indicator.

Challenge Rubric

• Logic: 0-25
• Clarity: 0-25
• Counter-handling: 0-25
• Evidence use: 0-25
• Total score: 0-100

AI scoring is deterministic (temperature ≤ 0.2), schema-constrained, and audit logged.

AI Limitations

• AI can miss nuance, misread context, or overweight framing quality.
• Structured outputs can still contain judgment errors.
• Raw AI outputs are stored for audit and moderation review.
• Global AI evaluation can be paused by admin kill switch.

What ReasonRank Does Not Do

• No moderation of political position.
• No definitive truth adjudication.
• No personal advice.
• No social interactions or engagement mechanics.