Oaksford and Chater redefine rationality as reasoning about uncertainty, not logical deduction

In Bayesian Rationality: The Probabilistic Approach to Human Reasoning (2007), and in its précis for Behavioral and Brain Sciences (2009), Mike Oaksford and Nick Chater argue that the standard against which human reasoning had been judged was the wrong one. On their account, "rationality is defined instead by the ability to reason about uncertainty" — not by conformity to formal logic. People are "typically poor at numerical reasoning about probability," yet "human thought is sensitive to subtle patterns of qualitative Bayesian, probabilistic reasoning."

The move displaces the older, broadly Piagetian view in which logical deduction is the developmental end-point and the yardstick of rational thought. Where classical experiments read deviations from logic as human error, the probabilistic approach treats those same behaviours as well-calibrated inference under uncertainty — the kind of reasoning an agent actually faces outside the artificial closure of a logic puzzle.

This is one of two twentieth/twenty-first-century rehabilitations of human reasoning that share a single rhetorical shape — "humans aren't irrational, the yardstick was wrong" — while disagreeing on mechanism. Gigerenzer's ecological rationality rehabilitates reasoning via simple heuristics matched to environmental structure (see the ecological framework summarised in rationality-continuity-problem); Oaksford and Chater rehabilitate it via optimal Bayesian inference. The two lineages are treated as distinct-but-adjacent foundational frameworks in recent LLM-meta-learning work, though whether they are genuinely compatible or in tension is unsettled (question-verify-ecological-rationality-vs-rational-analysis-compatibility).

The concrete payoff of the reframing is the reanalysis of specific reasoning tasks — most sharply the Wason selection task (claim-oaksford-chater-recast-wason-task-as-optimal-data-selection). The "rationality is bounded" tradition of Simon reframed optimality as good-enough search (claim-simon-satisficing-and-heuristic-search-same-move); Oaksford and Chater instead keep optimality but change its currency from logical validity to Bayesian information gain.