Belief updating across information ecologies: The trade-off between information gain and noise filtering

This paper investigates the ecology of belief updating in environments with varying reliability structures, focusing on how individuals navigate the trade-off between information utility and noise filtering. Combining computational simulations with behavioral experiments, we examined how belief-updating strategies ranging from discrete gating to continuous discounting performed across environments differing in source credibility and heterogeneity, and how frequently participants' behavior was best captured by each strategy. A simulation study mapped the performance landscape of these heuristics and showed that threshold-based strategies exhibit a relative advantage in long-sequence environments with low-to-moderate credibility, whereas performance differences across strategies are attenuated in high-credibility settings. Two behavioral experiments using a sequential estimation task revealed substantial between-participant heterogeneity: rather than a single strategy characterizing all participants, threshold-based, discounting, and Bayesian strategies each best described a subset of participants, with the prevalence of each strategy differing across informational environments in a manner broadly consistent with the simulations. Among participants best fit by the threshold model, fitted threshold values were also higher under conditions of greater mean source credibility, offering preliminary, population-level evidence that the decision criterion underlying threshold gating may be sensitive to the average reliability of the environment. Together, these findings highlight the importance of environmental context in shaping belief updating under uncertain source reliability.