Bayesian belief updating provides a normative framework for how a rational agent should revise probabilities in light of new evidence. However, empirical research in psychology and behavioral economics has consistently found systematic biases in human belief updating, such as conservatism (under-reaction to new data), base-rate neglect (under-weighting prior information), and confirmation bias (over-weighting confirmatory evidence). Existing theories either derive Bayesian updating axiomatically or interpret it through information-theoretic principles, but they do not explain these biases in a unified way. In this paper, we introduce a generalized Bayesian updating model that includes explicit cognitive constraints and information-processing costs. By adding a parametric distortion of the log Bayes factor and an explicit information cost term, we develop an update rule that reduces to classical Bayes’ rule as a special case. This formulation explains well-known cognitive biases as rational outcomes under bounded rationality. Our model connects normative Bayesian theory with descriptive cognitive phenomena and has implications for cognitive science, artificial intelligence, and decision support systems.

• Bayesian Updating
• Bounded Rationality
• Information Theory
• Cognitive Bias
• Kullback–Leibler Divergence
• Belief Revision
• Conservatism
• Confirmation Bias

[1] A. M. Jakobsen, Coarse Bayesian Updating, Oxford University Press, 2025. https://amjakobsen.com/wp-content/uploads/2025/07/cbu_2025_07_31.pdf.

[2] B. Kopp, Cognitive biases as Bayesian probability weighting in context, Frontiers in Psychology, 16, 2025, 1572168. https://doi.org/10.3389/fpsyg.2025.1572168

[3] P. A. Ortega and D. A. Braun, Thermodynamics as a theory of decision-making with information-processing costs, Proc. Royal Society A 469, 2012. 20120683. https://doi.org/10.48550/arXiv.1204.6481.

[4] J. Hyland and D. Albarracín, On the variational costs of changing our minds, 6th International Workshop on Active Inference, 2025. https://doi.org/10.48550/arXiv.2509.17957.

[5] V. Seckarová, Dynamic Parameter Estimation Based on Minimum Cross-Entropy Method for Combining Information Sources, Pliska Studia Mathematica Bulgarica, Vol. 24, No 1, 2015, 181p-188p. http://hdl.handle.net/10525/3524.

[6] C. Ba, J. A. Bohren, and A. Imas, Over- and Underreaction to Information, SSRN Working Paper, 2024. https://dx.doi.org/10.2139/ssrn.4274617.