B5
Heuristics & BiasesRecency Bias
Recent events should not disproportionately influence advice
4
Models Tested
4
Confirmatory
-1.327
Mean Effect
1.611
Max Effect
Theoretical Context
Theoretical Anchor
Benartzi (2001)
Normative Violation
Recent events should not disproportionately influence advice
Cross-Model Comparison
Effect sizes for Recency Bias across all tested models
OpenAI
GPT-5.2
The Steady Traditionalist
h = -1.611Confirmatory
Anthropic
Claude Sonnet 4.6
The Cautious Contrarian
h = -1.587Confirmatory
OpenAI
GPT-5.3 Instant
The Directive Optimist
h = -1.196Confirmatory
OpenAI
GPT-5.4 Thinking
The Deliberative Calibrator
h = -0.914Confirmatory
Statistical Details
Full results with confidence intervals and sample sizes
| Model | n (A) | n (B) | Cohen's h | 95% CI | Status |
|---|---|---|---|---|---|
| GPT-5.2 | 50 | 50 | -1.6111 | — | Confirmatory |
| Claude Sonnet 4.6 | 50 | 50 | -1.5868 | — | Confirmatory |
| GPT-5.3 Instant | 50 | 50 | -1.1957 | — | Confirmatory |
| GPT-5.4 Thinking | 50 | 50 | -0.9138 | — | Confirmatory |