Table 3.
Performance on the RAT and similarity patterns in the response search.
| Analysis | Humans | Model |
|---|---|---|
| Average problem accuracy | 42% | 43% |
| -Correlation with human data (r) | 0.49* | |
| Shortest response sequence | 1 | 1 |
| Longest response sequence | 49 | 33 |
| Average number of responses per trial | 7.78 | 8.44 |
| -Correlation with human data (r) | 0.93*** | |
| AVERAGE RESPONSE SIMILARITY | ||
| -Within vs. across cue clusters | 0.189 vs. 0.041 | 0.141 vs. 0.054 |
| CI: [0.134, 0.162] | CI: [0.079, 0.095] | |
| -Permutation test | 0.182 vs. 0.040 | 0.142 vs. 0.054 |
| CI: [0.124, 0.160] | CI: [0.077, 0.100] | |
| -Within vs. across cue clusters (cleaned responses) | 0.180 vs. 0.039 | 0.141 vs. 0.054 |
| CI: [0.128, 0.154] | CI: [0.079, 0.095] | |
| Baseline vs. actual percentage of response pairs with the same primary cue (two-sided exact binomial test) | 33.3 vs. 37.1%*** | 34.2 vs. 54.4%*** |
| AVERAGE SIMILARITY BETWEEN ADJACENT AND NON-ADJACENT | ||
| RESPONSES | ||
| -With different primary cues (across cluster) | 0.041 vs. 0.016 | 0.054 vs. 0.011 |
| CI: [0.063, 0.098] | CI: [0.038, 0.047] | |
| -With same primary cues (within cluster) | 0.189 vs. 0.108 | 0.141 vs. 0.076 |
| CI: [0.063, 0.098] | CI: [0.057, 0.072] | |
Stated 95% confidence intervals are computed on the difference of reported mean values. Values significant at p < 0.05 are marked with *, significant at p < 0.001 with ***.