File size: 16,592 Bytes
a81d5b1 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 | INVOCATION: cd experiments/rational_agents && PYTHONPATH=libs/interlens/src:. ../../.venv/bin/python analyze_game_structure_generality.py --config game_structure_generality_v1/config_final.json --out game_structure_generality_v1/analysis_final
# Game-structure generality — results (generated by analyze_game_structure_generality.py)
| structure | talk deal | mute deal | Bayesian deal | E1 mute − talk | E2 talk − Bayesian | coordination term | discovery term | ICC talk / mute / Bayes |
|---|---|---|---|---|---|---|---|---|
| 3 parties, unanimity, seeds 0-4 | 0.925 [0.842, 0.983] | 0.808 [0.700, 0.908] | 0.175 [0.092, 0.267] | -0.117 [-0.242, +0.017] | +0.750 [+0.608, +0.858] | +0.027 [-0.037, +0.110] | -0.143 [-0.235, -0.065] | 0.29 / 0.31 / 0.23 |
| 3 parties, unanimity, seeds 0-2 | 0.917 [0.847, 0.972] | 0.819 [0.708, 0.917] | 0.167 [0.083, 0.250] | -0.097 [-0.222, +0.028] | +0.750 [+0.625, +0.847] | +0.049 [-0.018, +0.126] | -0.146 [-0.235, -0.067] | 0.11 / 0.26 / 0.01 |
| 5 parties, unanimity (bank v2), seeds 0-4 | 0.908 [0.808, 0.983] | 0.600 [0.467, 0.725] | 0.233 [0.142, 0.325] | -0.308 [-0.425, -0.192] | +0.675 [+0.558, +0.783] | -0.127 [-0.191, -0.069] | -0.181 [-0.288, -0.084] | 0.51 / 0.30 / 0.15 |
| 5 parties, unanimity (bank v2), seeds 0-2 | 0.903 [0.806, 0.972] | 0.625 [0.486, 0.750] | 0.194 [0.111, 0.278] | -0.278 [-0.389, -0.167] | +0.708 [+0.583, +0.819] | -0.087 [-0.148, -0.032] | -0.191 [-0.300, -0.088] | 0.38 / 0.19 / -0.05 |
| 7 parties, unanimity, seeds 0-4 | 0.817 [0.700, 0.917] | 0.467 [0.325, 0.608] | 0.267 [0.133, 0.417] | -0.350 [-0.483, -0.225] | +0.550 [+0.383, +0.708] | -0.219 [-0.338, -0.106] | -0.131 [-0.201, -0.059] | 0.35 / 0.39 / 0.61 |
| 7 parties, unanimity, seeds 0-2 | 0.792 [0.653, 0.917] | 0.458 [0.306, 0.611] | 0.264 [0.125, 0.417] | -0.333 [-0.472, -0.194] | +0.528 [+0.347, +0.694] | -0.199 [-0.336, -0.066] | -0.134 [-0.225, -0.047] | 0.59 / 0.46 / 0.51 |
| 5 parties, 4 of 5 (bank v2), seeds 0-2 | 0.958 [0.875, 1.000] | 0.875 [0.764, 0.972] | 0.431 [0.292, 0.569] | -0.083 [-0.167, -0.014] | +0.528 [+0.375, +0.681] | -0.016 [-0.050, +0.000] | -0.192 [-0.319, -0.069] | 1.00 / 0.51 / 0.34 |
| context: Bayesian only, bank v1, unanimity | — | — | 0.233 [0.133, 0.342] | — | — | — | — | — / — / 0.24 |
| context: Bayesian only, bank v1, 4 of 5 | — | — | 0.617 [0.475, 0.750] | — | — | — | — | — / — / 0.40 |
## Cross-structure contrasts
| contrast | estimate [95% CI] | kind |
|---|---|---|
| E1(S5) − E1(S3) | -0.192 [-0.367, -0.025] | cross-bank, banks resampled independently |
| E2(S5) − E2(S3) | -0.075 [-0.242, +0.100] | cross-bank, banks resampled independently |
| E1(S7) − E1(S5) | -0.042 [-0.217, +0.133] | cross-bank, banks resampled independently |
| E2(S7) − E2(S5) | -0.125 [-0.325, +0.075] | cross-bank, banks resampled independently |
| E1(S7) − E1(S3) | -0.233 [-0.408, -0.058] | cross-bank, banks resampled independently |
| E2(S7) − E2(S3) | -0.200 [-0.400, +0.008] | cross-bank, banks resampled independently |
| E1(S5_s012) − E1(S3_s012) | -0.181 [-0.347, -0.014] | cross-bank, banks resampled independently |
| E2(S5_s012) − E2(S3_s012) | -0.042 [-0.208, +0.125] | cross-bank, banks resampled independently |
| E1(S7_s012) − E1(S5_s012) | -0.056 [-0.236, +0.125] | cross-bank, banks resampled independently |
| E2(S7_s012) − E2(S5_s012) | -0.181 [-0.389, +0.028] | cross-bank, banks resampled independently |
| E1(S7_s012) − E1(S3_s012) | -0.236 [-0.431, -0.056] | cross-bank, banks resampled independently |
| E2(S7_s012) − E2(S3_s012) | -0.222 [-0.431, -0.014] | cross-bank, banks resampled independently |
| E1(Q4) − E1(S5_s012) | +0.194 [+0.083, +0.306] | paired difference-in-differences on identical keys (n=72) |
| E2(Q4) − E2(S5_s012) | -0.181 [-0.306, -0.042] | paired difference-in-differences on identical keys (n=72) |
## Per-structure detail
### 3 parties, unanimity, seeds 0-4
| role | episodes | seeds | deal | normalized score | parties below threshold (per deal) | walked | zone tabled | closure given tabled | proposals | voided turns / ep | episodes voided at final | messages / ep | fabricated turns | $ / ep |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| talk | 120 | 0,1,2,3,4 | 0.925 [0.842, 0.983] | 0.797 [0.715, 0.867] | 0.000 [0.000, 0.000] | 0.008 [0.000, 0.025] | 0.983 [0.958, 1.000] | 0.941 [0.858, 0.992] | 5.067 [4.675, 5.450] | 0.000 [0.000, 0.000] | 0 | 10.617 [9.892, 11.358] | 0 | 0.576 [0.490, 0.668] |
| mute | 120 | 0,1,2,3,4 | 0.808 [0.700, 0.908] | 0.696 [0.592, 0.792] | 0.000 [0.000, 0.000] | 0.000 [0.000, 0.000] | 0.833 [0.725, 0.925] | 0.970 [0.934, 1.000] | 5.383 [4.875, 5.909] | 0.000 [0.000, 0.000] | 0 | 0.000 [0.000, 0.000] | 0 | 0.593 [0.508, 0.684] |
| bayes | 120 | 0,1,2,3,4 | 0.175 [0.092, 0.267] | 0.158 [0.077, 0.250] | 0.000 [0.000, 0.000] | 0.000 [0.000, 0.000] | 0.500 [0.325, 0.667] | 0.350 [0.209, 0.508] | 10.892 [10.633, 11.183] | 0.000 [0.000, 0.000] | 0 | 0.000 [0.000, 0.000] | 0 | 0.000 [0.000, 0.000] |
- E1 (deal): -0.117 [-0.242, +0.017]; E1 (normalized score): -0.101 [-0.224, +0.032]; E1 excluding episodes with a seat voided at the final vote: -0.117 [-0.242, +0.017]
- census of E1: Δzone-tabled -0.150 [-0.242, -0.067], Δclosure-given-tabled +0.029 [-0.042, +0.116], discovery term -0.143 [-0.235, -0.065], coordination term +0.027 [-0.037, +0.110], remainder -0.000 [-0.000, +0.000] (n pairs 120)
- E2 talk − Bayesian: +0.750 [+0.608, +0.858]; mute − Bayesian: +0.633 [+0.525, +0.733]
### 3 parties, unanimity, seeds 0-2
| role | episodes | seeds | deal | normalized score | parties below threshold (per deal) | walked | zone tabled | closure given tabled | proposals | voided turns / ep | episodes voided at final | messages / ep | fabricated turns | $ / ep |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| talk | 72 | 0,1,2 | 0.917 [0.847, 0.972] | 0.786 [0.701, 0.865] | 0.000 [0.000, 0.000] | 0.014 [0.000, 0.042] | 0.986 [0.958, 1.000] | 0.930 [0.857, 0.986] | 5.083 [4.625, 5.542] | 0.000 [0.000, 0.000] | 0 | 10.653 [9.875, 11.458] | 0 | 0.584 [0.493, 0.679] |
| mute | 72 | 0,1,2 | 0.819 [0.708, 0.917] | 0.702 [0.596, 0.798] | 0.000 [0.000, 0.000] | 0.000 [0.000, 0.000] | 0.833 [0.722, 0.931] | 0.983 [0.946, 1.000] | 5.389 [4.833, 5.972] | 0.000 [0.000, 0.000] | 0 | 0.000 [0.000, 0.000] | 0 | 0.595 [0.504, 0.693] |
| bayes | 72 | 0,1,2 | 0.167 [0.083, 0.250] | 0.153 [0.076, 0.239] | 0.000 [0.000, 0.000] | 0.000 [0.000, 0.000] | 0.500 [0.333, 0.667] | 0.333 [0.205, 0.472] | 10.903 [10.625, 11.208] | 0.000 [0.000, 0.000] | 0 | 0.000 [0.000, 0.000] | 0 | 0.000 [0.000, 0.000] |
- E1 (deal): -0.097 [-0.222, +0.028]; E1 (normalized score): -0.085 [-0.202, +0.039]; E1 excluding episodes with a seat voided at the final vote: -0.097 [-0.222, +0.028]
- census of E1: Δzone-tabled -0.153 [-0.250, -0.069], Δclosure-given-tabled +0.054 [-0.021, +0.136], discovery term -0.146 [-0.235, -0.067], coordination term +0.049 [-0.018, +0.126], remainder +0.000 [-0.000, +0.000] (n pairs 72)
- E2 talk − Bayesian: +0.750 [+0.625, +0.847]; mute − Bayesian: +0.653 [+0.542, +0.750]
### 5 parties, unanimity (bank v2), seeds 0-4
| role | episodes | seeds | deal | normalized score | parties below threshold (per deal) | walked | zone tabled | closure given tabled | proposals | voided turns / ep | episodes voided at final | messages / ep | fabricated turns | $ / ep |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| talk | 120 | 0,1,2,3,4 | 0.908 [0.808, 0.983] | 0.791 [0.695, 0.867] | 0.000 [0.000, 0.000] | 0.017 [0.000, 0.050] | 0.925 [0.825, 1.000] | 0.982 [0.955, 1.000] | 5.733 [5.250, 6.258] | 0.000 [0.000, 0.000] | 0 | 18.500 [17.550, 19.450] | 0 | 1.048 [0.942, 1.172] |
| mute | 120 | 0,1,2,3,4 | 0.600 [0.467, 0.725] | 0.503 [0.391, 0.613] | 0.000 [0.000, 0.000] | 0.000 [0.000, 0.000] | 0.725 [0.583, 0.850] | 0.828 [0.744, 0.902] | 6.425 [5.883, 6.984] | 0.000 [0.000, 0.000] | 0 | 0.000 [0.000, 0.000] | 0 | 1.271 [1.142, 1.395] |
| bayes | 120 | 0,1,2,3,4 | 0.233 [0.142, 0.325] | 0.195 [0.116, 0.278] | 0.000 [0.000, 0.000] | 0.000 [0.000, 0.000] | 0.517 [0.358, 0.675] | 0.452 [0.333, 0.565] | 14.833 [13.967, 15.667] | 0.000 [0.000, 0.000] | 0 | 0.000 [0.000, 0.000] | 0 | 0.000 [0.000, 0.000] |
- E1 (deal): -0.308 [-0.425, -0.192]; E1 (normalized score): -0.288 [-0.391, -0.190]; E1 excluding episodes with a seat voided at the final vote: -0.308 [-0.425, -0.192]
- census of E1: Δzone-tabled -0.200 [-0.317, -0.092], Δclosure-given-tabled -0.154 [-0.236, -0.085], discovery term -0.181 [-0.288, -0.084], coordination term -0.127 [-0.191, -0.069], remainder +0.000 [-0.000, +0.000] (n pairs 120)
- E2 talk − Bayesian: +0.675 [+0.558, +0.783]; mute − Bayesian: +0.367 [+0.233, +0.500]
### 5 parties, unanimity (bank v2), seeds 0-2
| role | episodes | seeds | deal | normalized score | parties below threshold (per deal) | walked | zone tabled | closure given tabled | proposals | voided turns / ep | episodes voided at final | messages / ep | fabricated turns | $ / ep |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| talk | 72 | 0,1,2 | 0.903 [0.806, 0.972] | 0.768 [0.673, 0.843] | 0.000 [0.000, 0.000] | 0.000 [0.000, 0.000] | 0.931 [0.833, 1.000] | 0.970 [0.925, 1.000] | 5.875 [5.319, 6.500] | 0.000 [0.000, 0.000] | 0 | 18.514 [17.333, 19.736] | 0 | 1.081 [0.951, 1.231] |
| mute | 72 | 0,1,2 | 0.625 [0.486, 0.750] | 0.524 [0.411, 0.631] | 0.000 [0.000, 0.000] | 0.000 [0.000, 0.000] | 0.722 [0.583, 0.847] | 0.865 [0.780, 0.943] | 6.639 [6.027, 7.278] | 0.000 [0.000, 0.000] | 0 | 0.000 [0.000, 0.000] | 0 | 1.310 [1.176, 1.441] |
| bayes | 72 | 0,1,2 | 0.194 [0.111, 0.278] | 0.156 [0.089, 0.226] | 0.000 [0.000, 0.000] | 0.000 [0.000, 0.000] | 0.500 [0.333, 0.667] | 0.389 [0.297, 0.487] | 14.694 [13.792, 15.569] | 0.000 [0.000, 0.000] | 0 | 0.000 [0.000, 0.000] | 0 | 0.000 [0.000, 0.000] |
- E1 (deal): -0.278 [-0.389, -0.167]; E1 (normalized score): -0.244 [-0.349, -0.143]; E1 excluding episodes with a seat voided at the final vote: -0.278 [-0.389, -0.167]
- census of E1: Δzone-tabled -0.208 [-0.319, -0.097], Δclosure-given-tabled -0.105 [-0.183, -0.038], discovery term -0.191 [-0.300, -0.088], coordination term -0.087 [-0.148, -0.032], remainder +0.000 [-0.000, +0.000] (n pairs 72)
- E2 talk − Bayesian: +0.708 [+0.583, +0.819]; mute − Bayesian: +0.431 [+0.291, +0.569]
### 7 parties, unanimity, seeds 0-4
| role | episodes | seeds | deal | normalized score | parties below threshold (per deal) | walked | zone tabled | closure given tabled | proposals | voided turns / ep | episodes voided at final | messages / ep | fabricated turns | $ / ep |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| talk | 120 | 0,1,2,3,4 | 0.817 [0.700, 0.917] | 0.726 [0.616, 0.822] | 0.000 [0.000, 0.000] | 0.008 [0.000, 0.025] | 0.883 [0.775, 0.967] | 0.925 [0.880, 0.964] | 6.575 [5.908, 7.283] | 0.000 [0.000, 0.000] | 0 | 28.267 [27.150, 29.458] | 0 | 1.716 [1.567, 1.872] |
| mute | 120 | 0,1,2,3,4 | 0.467 [0.325, 0.608] | 0.417 [0.293, 0.544] | 0.000 [0.000, 0.000] | 0.000 [0.000, 0.000] | 0.717 [0.592, 0.833] | 0.651 [0.500, 0.786] | 7.358 [6.633, 8.125] | 0.000 [0.000, 0.000] | 0 | 0.000 [0.000, 0.000] | 0 | 1.783 [1.665, 1.901] |
| bayes | 120 | 0,1,2,3,4 | 0.267 [0.133, 0.417] | 0.217 [0.104, 0.341] | 0.000 [0.000, 0.000] | 0.000 [0.000, 0.000] | 0.458 [0.283, 0.633] | 0.582 [0.386, 0.746] | 19.350 [18.100, 20.600] | 0.000 [0.000, 0.000] | 0 | 0.000 [0.000, 0.000] | 0 | 0.000 [0.000, 0.000] |
- E1 (deal): -0.350 [-0.483, -0.225]; E1 (normalized score): -0.309 [-0.428, -0.193]; E1 excluding episodes with a seat voided at the final vote: -0.350 [-0.483, -0.225]
- census of E1: Δzone-tabled -0.167 [-0.258, -0.075], Δclosure-given-tabled -0.273 [-0.426, -0.134], discovery term -0.131 [-0.201, -0.059], coordination term -0.219 [-0.338, -0.106], remainder -0.000 [-0.000, +0.000] (n pairs 120)
- E2 talk − Bayesian: +0.550 [+0.383, +0.708]; mute − Bayesian: +0.200 [+0.067, +0.325]
### 7 parties, unanimity, seeds 0-2
| role | episodes | seeds | deal | normalized score | parties below threshold (per deal) | walked | zone tabled | closure given tabled | proposals | voided turns / ep | episodes voided at final | messages / ep | fabricated turns | $ / ep |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| talk | 72 | 0,1,2 | 0.792 [0.653, 0.917] | 0.710 [0.574, 0.827] | 0.000 [0.000, 0.000] | 0.014 [0.000, 0.042] | 0.847 [0.708, 0.958] | 0.934 [0.867, 0.984] | 6.722 [5.931, 7.556] | 0.000 [0.000, 0.000] | 0 | 29.139 [27.806, 30.472] | 0 | 1.799 [1.628, 1.984] |
| mute | 72 | 0,1,2 | 0.458 [0.306, 0.611] | 0.410 [0.268, 0.552] | 0.000 [0.000, 0.000] | 0.000 [0.000, 0.000] | 0.681 [0.542, 0.819] | 0.673 [0.500, 0.828] | 7.417 [6.597, 8.319] | 0.000 [0.000, 0.000] | 0 | 0.000 [0.000, 0.000] | 0 | 1.795 [1.659, 1.931] |
| bayes | 72 | 0,1,2 | 0.264 [0.125, 0.417] | 0.216 [0.104, 0.342] | 0.000 [0.000, 0.000] | 0.000 [0.000, 0.000] | 0.431 [0.250, 0.611] | 0.613 [0.429, 0.788] | 19.569 [18.264, 20.903] | 0.000 [0.000, 0.000] | 0 | 0.000 [0.000, 0.000] | 0 | 0.000 [0.000, 0.000] |
- E1 (deal): -0.333 [-0.472, -0.194]; E1 (normalized score): -0.300 [-0.427, -0.179]; E1 excluding episodes with a seat voided at the final vote: -0.333 [-0.472, -0.194]
- census of E1: Δzone-tabled -0.167 [-0.278, -0.069], Δclosure-given-tabled -0.261 [-0.444, -0.089], discovery term -0.134 [-0.225, -0.047], coordination term -0.199 [-0.336, -0.066], remainder +0.000 [-0.000, +0.000] (n pairs 72)
- E2 talk − Bayesian: +0.528 [+0.347, +0.694]; mute − Bayesian: +0.194 [+0.056, +0.319]
### 5 parties, 4 of 5 (bank v2), seeds 0-2
| role | episodes | seeds | deal | normalized score | parties below threshold (per deal) | walked | zone tabled | closure given tabled | proposals | voided turns / ep | episodes voided at final | messages / ep | fabricated turns | $ / ep |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| talk | 72 | 0,1,2 | 0.958 [0.875, 1.000] | 0.795 [0.707, 0.860] | 0.449 [0.304, 0.594] | 0.042 [0.000, 0.125] | 0.833 [0.722, 0.931] | 1.000 [1.000, 1.000] | 5.139 [4.667, 5.653] | 0.000 [0.000, 0.000] | 0 | 14.750 [13.583, 15.958] | 0 | 0.914 [0.785, 1.054] |
| mute | 72 | 0,1,2 | 0.875 [0.764, 0.972] | 0.697 [0.595, 0.784] | 0.556 [0.433, 0.672] | 0.042 [0.000, 0.111] | 0.639 [0.486, 0.792] | 0.978 [0.930, 1.000] | 5.611 [5.097, 6.181] | 0.000 [0.000, 0.000] | 0 | 0.000 [0.000, 0.000] | 0 | 1.055 [0.905, 1.207] |
| bayes | 72 | 0,1,2 | 0.431 [0.292, 0.569] | 0.366 [0.241, 0.493] | 0.581 [0.433, 0.737] | 0.000 [0.000, 0.000] | 0.444 [0.319, 0.583] | 0.625 [0.429, 0.806] | 13.417 [12.500, 14.292] | 0.000 [0.000, 0.000] | 0 | 0.000 [0.000, 0.000] | 0 | 0.000 [0.000, 0.000] |
- E1 (deal): -0.083 [-0.167, -0.014]; E1 (normalized score): -0.099 [-0.187, -0.027]; E1 excluding episodes with a seat voided at the final vote: -0.083 [-0.167, -0.014]
- census of E1: Δzone-tabled -0.194 [-0.333, -0.069], Δclosure-given-tabled -0.022 [-0.070, +0.000], discovery term -0.192 [-0.319, -0.069], coordination term -0.016 [-0.050, +0.000], remainder +0.125 [-0.000, +0.264] (n pairs 72)
- E2 talk − Bayesian: +0.528 [+0.375, +0.681]; mute − Bayesian: +0.444 [+0.306, +0.583]
### context: Bayesian only, bank v1, unanimity
| role | episodes | seeds | deal | normalized score | parties below threshold (per deal) | walked | zone tabled | closure given tabled | proposals | voided turns / ep | episodes voided at final | messages / ep | fabricated turns | $ / ep |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| bayes | 120 | 0,1,2,3,4 | 0.233 [0.133, 0.342] | 0.189 [0.110, 0.279] | 0.000 [0.000, 0.000] | 0.000 [0.000, 0.000] | 0.475 [0.308, 0.650] | 0.491 [0.372, 0.627] | 14.733 [13.992, 15.467] | 0.000 [0.000, 0.000] | 0 | 0.000 [0.000, 0.000] | 0 | 0.000 [0.000, 0.000] |
### context: Bayesian only, bank v1, 4 of 5
| role | episodes | seeds | deal | normalized score | parties below threshold (per deal) | walked | zone tabled | closure given tabled | proposals | voided turns / ep | episodes voided at final | messages / ep | fabricated turns | $ / ep |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| bayes | 120 | 0,1,2,3,4 | 0.617 [0.475, 0.750] | 0.492 [0.376, 0.608] | 0.622 [0.467, 0.781] | 0.000 [0.000, 0.000] | 0.442 [0.283, 0.608] | 0.868 [0.783, 0.939] | 12.600 [11.425, 13.683] | 0.000 [0.000, 0.000] | 0 | 0.000 [0.000, 0.000] | 0 | 0.000 [0.000, 0.000] |
|