Abstract for

"The Economics of Cognition. I. Algorithmic Information-Theoretic Explanation of Cognitive Biases and Fallacies"

This paper attempts to explain well-established, incorrigible cognitive biases and fallacies in lay reasoning as outcomes of motivated but possibly unconscious cognitive choices over a set of models. The paper begins by modeling the agent as an information processor in the classic information-theoretic of Claude Shannon, with a small but significant change: the Shannon decoder (Maximum Likelihood or Minimum Mean Squared Error) is replaced with a generalized computational device with finite or costly memory and computational resources. The problem of cognitive choice that each agent can be understood as attempting to solve when formulating a judgment about an unknown quantity is that of selecting a working model (WM) that either minimizes the use of computational resources (computational load) for convergence to a judgment subject to working memory limitations, or one that tries to minimize working memory requirements subject to hard limits on the use of computational resources. This framework allows us to study computation-bound and memory-bound cognitive choice scenarios and to adduce cognitively rationalizable explanations for well-known characteristics of human reasoning processes, such as cognitive dissonance aversion, representativeness heuristics, conjunction biases and disjunctive biases.