======================================================================
GENDER BIAS DETECTION EXPERIMENT — STATISTICAL SUMMARY
======================================================================

Model: Qwen 3.6 Plus
  N scenarios   : 50
  Mean Δ (signed): -0.1140
  Mean |Δ|       : 0.2659
  Std dev Δ      : 0.5324
  Max Δ          : +0.7692
  Min Δ          : -2.0000
  Biased (Δ≠0)  : 37/50  (74%)
    Male-favoring: 18  Female-favoring: 19
  Wilcoxon W     : 289.0  p = 0.3456

Model: Claude Sonnet 4.6
  N scenarios   : 50
  Mean Δ (signed): -0.0876
  Mean |Δ|       : 0.2135
  Std dev Δ      : 0.3912
  Max Δ          : +0.9189
  Min Δ          : -1.6000
  Biased (Δ≠0)  : 37/50  (74%)
    Male-favoring: 15  Female-favoring: 22
  Wilcoxon W     : 280.0  p = 0.2804

Model: Gemini 3.1 Flash Lite
  N scenarios   : 50
  Mean Δ (signed): -0.0245
  Mean |Δ|       : 0.1453
  Std dev Δ      : 0.2461
  Max Δ          : +0.6667
  Min Δ          : -0.6667
  Biased (Δ≠0)  : 22/50  (44%)
    Male-favoring: 8  Female-favoring: 14
  Wilcoxon W     : 112.0  p = 0.6361

Model: DeepSeek V4 Flash
  N scenarios   : 50
  Mean Δ (signed): -0.0064
  Mean |Δ|       : 0.0817
  Std dev Δ      : 0.2596
  Max Δ          : +1.2000
  Min Δ          : -1.0000
  Biased (Δ≠0)  : 7/50  (14%)
    Male-favoring: 3  Female-favoring: 4
  Wilcoxon W     : 13.0  p = 0.8658
