Response-based outcome predictions and confidence regulate feedback processing and learning

Abstract

Influential theories emphasize the importance of predictions in learning: we learn from feedback to the extent that it is surprising, and thus conveys new information. Here, we explore the hypothesis that surprise depends not only on comparing current events to past experience, but also on online evaluation of performance via internal monitoring. Specifically, we propose that people leverage insights from response-based performance monitoring – outcome predictions and confidence – to control learning from feedback. In line with predictions from a Bayesian inference model, we find that people who are better at calibrating their confidence to the precision of their outcome predictions learn more quickly. Further in line with our proposal, EEG signatures of feedback processing are sensitive to the accuracy of, and confidence in, post-response outcome predictions. Taken together, our results suggest that online predictions and confidence serve to calibrate neural error signals to improve the efficiency of learning.

Feedback-related EEG dynamics separately reflect decision parameters, biases, and future choices

“…Jocham et al (2014)) converge on a common integrative process reflected in a centroparietal positivity. In line with previous research (Chase et al, 2011;Fromer et al, 2021;Nassar et al, 2019;Ullsperger, Fischer, et al, 2014;Wessel & Aron, 2017) we suggest, that this EEG correlate reflects a final pathway of adaptation that sets the stage for future behaviour. This view is supported by our decoding analyses suggesting that behaviour switches are best predicted by an EEG activity with a topography and latency matching the P3b.…”

Learning when effort matters: Neural dynamics underlying updating and adaptation to changes in performance efficacy

“…Our efficacy-updating results are a reminder that many aspects of feedback are reflected in prediction error signals (Langdon et al 2018; Frömer et al 2021b). In the present study, we intentionally separated feedback about reward and efficacy to isolate the cognitive and neural learning mechanisms associated with each.…”

“…To visualize the topographies of the relevant ERP effects, we fitted the relevant models to all 64 channels and then plotted the posterior estimates of the effects of interest at each electrode (cf. Frömer et al, 2021b).…”

Event-related potential correlates of learning to produce novel foreign phonemes

“…The authors suggest that confidence approximates the amount of noise in the efference copy signals (i.e., their precision), which together with the efference copy and feedback, allow the individuals to make inferences about the success of their action. Their results showed that participants with more accurately "calibrated" confidence (i.e., more accurate estimate of noise in efference copies) learned better (Frömer et al, 2021). This suggests that LSW could reflect higher-order monitoring mechanisms: Whereas SIS in the N1 time-window may reflect automatic "online" corrections based on the efference copy (Hickok et al, 2011;Houde & Nagarajan, 2011;Tourville & Guenther, 2011), the LSW could reflect monitoring processes that integrate multiple sources of information (e.g., perceived auditory feedback on one's pronunciation, predictions based on the efference copy, and the individual's confidence in the accuracy of these predictions).…”