Confidence Prediction from EEG Recordings in a Multisensory Environment

This paper investigates the possibility of decoding decision confidence from electroencephalographic (EEG) brain activity of human subjects during a multisensory decision-making task. In recent research we have shown that decision confidence correlates could be extracted from EEG recordings during visual or auditory tasks. Here we extend these initial findings by (a) predicting the confidence in the decision from EEG recordings alone, and (b) investigating the impact of multisensory cues on decision-making behavioral data. Our results obtained from 12 participants recorded at two different sites show that the decision confidence could be predicted from EEG recordings on a single-trial basis with a mean absolute error of 0.226. Moreover, the presence of a multisensory cue did not improve the performance of the participants, but rather distracted them from the main task. Overall, these results may inform the development of cognitive systems that could monitor and alert users when they are not confident about their decisions.