Dataset Information

Confirmation bias in human reinforcement learning: Evidence from counterfactual feedback processing.

ABSTRACT: Previous studies suggest that factual learning, that is, learning from obtained outcomes, is biased, such that participants preferentially take into account positive, as compared to negative, prediction errors. However, whether or not the prediction error valence also affects counterfactual learning, that is, learning from forgone outcomes, is unknown. To address this question, we analysed the performance of two groups of participants on reinforcement learning tasks using a computational model that was adapted to test if prediction error valence influences learning. We carried out two experiments: in the factual learning experiment, participants learned from partial feedback (i.e., the outcome of the chosen option only); in the counterfactual learning experiment, participants learned from complete feedback information (i.e., the outcomes of both the chosen and unchosen option were displayed). In the factual learning experiment, we replicated previous findings of a valence-induced bias, whereby participants learned preferentially from positive, relative to negative, prediction errors. In contrast, for counterfactual learning, we found the opposite valence-induced bias: negative prediction errors were preferentially taken into account, relative to positive ones. When considering valence-induced bias in the context of both factual and counterfactual learning, it appears that people tend to preferentially take into account information that confirms their current choice.

SUBMITTER: Palminteri S

PROVIDER: S-EPMC5568446 | biostudies-literature | 2017 Aug

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Confirmation bias in human reinforcement learning: Evidence from counterfactual feedback processing.

Palminteri Stefano S Lefebvre Germain G Kilford Emma J EJ Blakemore Sarah-Jayne SJ

PLoS computational biology 20170811 8

Previous studies suggest that factual learning, that is, learning from obtained outcomes, is biased, such that participants preferentially take into account positive, as compared to negative, prediction errors. However, whether or not the prediction error valence also affects counterfactual learning, that is, learning from forgone outcomes, is unknown. To address this question, we analysed the performance of two groups of participants on reinforcement learning tasks using a computational model t ...[more]

PMID: 28800597

Similar Datasets

Project description:Counterfactual conditionals are frequently used in language to express potentially valid reasoning from factually false suppositions. Counterfactuals provide two pieces of information: their literal meaning expresses a suppositional dependency between an antecedent (If the dice had been rigged…) and a consequent (… then the game would have been unfair). Their second, backgrounded meaning refers to the opposite state of affairs and suggests that, in fact, the dice were not rigged and the game was fair. Counterfactual antecedents are particularly intriguing because they set up a counterfactual world which is known to be false, but which is nevertheless kept to when evaluating the conditional's consequent. In the last years several event-related potential (ERP) studies have targeted the processing of counterfactual consequents, yet counterfactual antecedents have remained unstudied. We present an EEG/ERP investigation which employed German conditionals to compare subjunctive mood (which marks counterfactuality) with indicative mood at the critical point of mood disambiguation via auxiliary introduction in the conditional's antecedent. Conditional sentences were presented visually one word at a time. Participants completed an acceptability judgment and probe detection task which was not related to the critical manipulation of linguistic mood. ERPs at the point of mood disambiguation in the antecedent were compared between indicative and subjunctive. Our main finding is a transient negative deflection in frontal regions for subjunctive compared to indicative mood in a time-window of 450-600 ms. We discuss this novel finding in respect to working memory requirements for rule application and increased referential processing demands for the representation of counterfactuals' dual meaning. Our result suggests that the counterfactually implied dual meaning is processed without any delay at the earliest point where counterfactuality is marked by subjunctive mood.

Dataset Information

Confirmation bias in human reinforcement learning: Evidence from counterfactual feedback processing.

Publications

Confirmation bias in human reinforcement learning: Evidence from counterfactual feedback processing.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets