Learning Bayesian Statistics in Adulthood

Abstract

Probabilities are a pervasive aspect of human life and probabilistic thinking is part of (sociological, psychological) models of rational (social) actors. Yet research documents poor understanding of probability in the general public and suggest that people do not update their estimates about future events when given additional information. This may point to the difficult nature of Bayesian thinking, the framework that would allow rational decision makers to update their estimates about future probabilistic events. Drawing on first-person methods for the study of cognition, I articulate invariants of learning an advanced statistical topic: Bayesian statistics. The first case study focuses on the learning of some fundamentals (such as those that one may find as content of a Wikipedia page); the second case study presents and analyzes a learning episode in the case of quantitative social science research that takes into account prior studies for establishing prior probabilities required for calculating posterior probabilities given the information collected in the study. The analyses show—consistent with pragmatic theories of language—that the essential dimension of learning is what equations, terms, and formulae require to be done rather than their (elusive) “meaning.”