Confidence from other people lowered probability error after condition changes in eight response-aligned SEEG key readers with 32 training trials

We tested confidence from other people for key readers in public OpenNeuro ds004473 version 1.0.2. Each of eight people supplied 32 early training trials. Separate first-condition trials from the other seven people fitted one bounded confidence slope. The next condition supplied 979 test trials. Mean balanced Brier error fell from 0.35746 to 0.24982. A constant probability of 0.5 gives error of 0.25000. The donor change from this constant was -0.00018. Its 95 percent person bootstrap interval was -0.00082 to +0.00303. The bootstrap refitted the slope and excluded every copy of the test identity. Thus, the test did not establish a gain over the constant prediction. All eight readers had less error than the original reader. The fitted slopes were between zero and 0.02547. The later probabilities moved close to 0.5. The package provides code, complete feature arrays, split records, predictions, proofs, four figures, and pinned raw retrieval instructions. All source arrays matched a fresh rebuild from 2,032 raw SEEG trial ranges. A second analysis matched all 11 main numerical files exactly. The evidence is limited to this fixed reader, known response windows, eight people with epilepsy, and one condition order per person. This study concerns brain reading. It does not test brain stimulation.

stereoelectroencephalography · confidence scaling · brain-computer interface · probability error · identity exclusion · condition change