A Comment on "Bayesianism and Wishful Thinking are Compatible"

Melnikoff and Strohminger (2024) report that affective prediction errors drive wishful belief updating, i.e., the tendency to adjust beliefs in the direction of one's desires. Historically, this phenomenon has posed a challenge to Bayesian accounts of reasoning, which assume that beliefs are updated relative to the available evidence and prior beliefs. However, the authors propose that wishful belief updating can align with Bayesian principles when affective prediction errors as "hidden information signals" are taken into account. Across five experiments, the authors provide compelling evidence for this account and show that affective prediction errors systematically influence belief updates in the direction of desires, and they formalise this in a Bayesian model. We were partially able to computationally reproduce the findings from Experiments 1, 2, and 3 using the provided code and data. Reported results from experiments 1 and 2 could be fully reproduced by inferring the statistical models based on reported results and rewriting part of the analysis code, as the provided code appeared incomplete. To assess robustness, we reanalyzed Experiments 1, 2, and 3 by including participants previously excluded for failing attention checks. The findings remained robust under these alternative specifications. For Experiment 4, data and code were not available on OSF. The data in the folder labeled "Experiment 4" instead appeared to provide data and code for Experiment 5. We used this to partially reproduce the findings from Experiment 5, but for more complex analyses, information on models and code was incomplete, making it impossible for us to fully reproduce the results. We provide the additional code we used to reproduce the findings.