A1
Framing & ReferenceLoss Aversion Asymmetry
Economically identical scenarios should receive identical recommendations
4
Models Tested
0
Confirmatory
-0.286
Mean Effect
0.456
Max Effect
Theoretical Context
Theoretical Anchor
Kahneman & Tversky (1979), Prospect Theory
Normative Violation
Economically identical scenarios should receive identical recommendations
Cross-Model Comparison
Effect sizes for Loss Aversion Asymmetry across all tested models
OpenAI
GPT-5.4 Thinking
The Deliberative Calibrator
h = -0.456
OpenAI
GPT-5.3 Instant
The Directive Optimist
h = -0.448
Google
Gemini 2.0 Flash
The Consistent Optimist
h = -0.323
Anthropic
Claude Sonnet 4.6
The Cautious Contrarian
h = +0.082
Statistical Details
Full results with confidence intervals and sample sizes
| Model | n (A) | n (B) | Cohen's h | 95% CI | Status |
|---|---|---|---|---|---|
| GPT-5.4 Thinking | 50 | 50 | -0.4556 | — | Exploratory |
| GPT-5.3 Instant | 50 | 50 | -0.4484 | — | Exploratory |
| Gemini 2.0 Flash | 50 | 50 | -0.3225 | — | Exploratory |
| Claude Sonnet 4.6 | 50 | 50 | +0.0817 | — | Exploratory |