Heuristics and Public Policy: Decision-making Under Bounded Rationality

How do human beings make decisions when, as the evidence indicates, the assumptions of the Bayesian rationality approach in economics do not hold? Do human beings optimize, or can they? Several decades of research have shown that people possess a toolkit of heuristics to make decisions under certainty, risk, subjective uncertainty, and true uncertainty (or Knightian uncertainty). We outline recent advances in knowledge about the use of heuristics and departures from Bayesian rationality, with particular emphasis on growing formalization of those departures, which add necessary precision. We also explore the relationship between bounded rationality and libertarian paternalism, or nudges, and show that some recent objections, founded on psychological work on the usefulness of certain heuristics, are based on serious misunderstandings. JEL classifications: D01, D04, D81, D9

Dynamic Random Utility

We provide an axiomatic analysis of dynamic random utility, characterizing the stochastic choice behavior of agents who solve dynamic decision problems by maximizing some stochastic process ( U t ) of utilities. We show first that even when ( U t ) is arbitrary, dynamic random utility imposes new testable across‐period restrictions on behavior, over and above period‐by‐period analogs of the static random utility axioms. An important feature of dynamic random utility is that behavior may appear history‐dependent, because period‐ t choices reveal information about U t , which may be serially correlated; however, our key new axioms highlight that the model entails specific limits on the form of history dependence that can arise. Second, we show that imposing natural Bayesian rationality axioms restricts the form of randomness that ( U t ) can display. By contrast, a specification of utility shocks that is widely used in empirical work violates these restrictions, leading to behavior that may display a negative option value and can produce biased parameter estimates. Finally, dynamic stochastic choice data allow us to characterize important special cases of random utility—in particular, learning and taste persistence—that on static domains are indistinguishable from the general model.

People Make the Same Bayesian Judgment They Criticize in Others

When two individuals from different social groups exhibit identical behavior, egalitarian codes of conduct call for equal judgments of both individuals. However, this moral imperative is at odds with the statistical imperative to consider priors based on group membership. Insofar as these priors differ, Bayesian rationality calls for unequal judgments of both individuals. We show that participants criticized the morality and intellect of someone else who made a Bayesian judgment, shared less money with this person, and incurred financial costs to punish this person. However, participants made unequal judgments as a Bayesian statistician would, thereby rendering the same judgment that they found repugnant when offered by someone else. This inconsistency, which can be reconciled by differences in which base rate is attended to, suggests that participants use group membership in a way that reflects the savvy of a Bayesian and the disrepute of someone they consider to be a bigot.

Rationality without optimality: Bounded and ecological rationality from a Marrian perspective

Ecological rationality provides an alternative to the view that rational responses toenvironmental uncertainty are optimal probabilistic responses. Focusing on the ecological rationality of simple heuristics, critics have enlisted Marr's levels of analysis and the distinction between function and mechanism to argue that the study of ecological rationality addresses the question of how organisms make decisions, but not the question of what constitutes a rational decision and why. The claim is that the insights of ecological rationality are, after the fact, reducible to instances of optimal Bayesian inference and require principles of Bayesian rationality to explain. Here, I respond to these critiques by clarifying that ecological rationality is more than a set of algorithmic conjectures. It is also driven by statistical commitments governing the treatment of unquantifiable uncertainty. This statistical perspective establishes why ecological rationality is distinct from Bayesian optimality, is incompatible with Marr's levels of analysis, and undermines a strict separation of function and mechanism. This argument finds support in Marr's broader but largely overlooked views on information processing systems and Savage's stance on the limits on Bayesian decision theory. Rationality principles make assumptions, and ecological rationality assumes that environmental uncertainty can render optimal probabilistic responses indeterminable.

The Evolution–Rationality Connection

There are two related dimensions to the evolution–rationality connection. The first is the evolution of rationality itself, thought of as an actual phenotypic attribute of some organisms; the second is the use of rationality-inspired concepts to describe evolved organisms, as in agential thinking. Rationality may be understood either as consistency of choice or as having good reasons for beliefs/actions; these notions have distinct domains of application. The adaptive significance of rationality over arationality is clear; what is less clear is whether evolution would always favour rationality over irrationality. In a simple model, an evolutionary basis for the norms of Bayesian rationality emerges; however, the model relies on restrictive assumptions. The possibility of an evolutionary naturalization of traditional rationality norms, though philosophically coherent, appears empirically unlikely.

Original Position Models, Trade-offs and Continuity

John Harsanyi has offered an argument grounded in Bayesian decision theory that purports to show that John Rawls's original position analysis leads directly to utilitarian conclusions. After explaining why a prominent Rawlsian line of response to Harsanyi's argument fails, I argue that a seemingly innocuous Bayesian rationality assumption, the continuity axiom, is at the heart of a fundamental disagreement between Harsanyi and Rawls. The most natural way for a Rawlsian to respond to Harsanyi's line of analysis, I argue, is to reject continuity. I then argue that this Rawlsian response fails as a defence of the difference principle, and I raise some concerns about whether it makes sense to posit the discontinuities needed to support the other elements of Rawls's view, although I suggest that Rawls may be able to invoke discontinuity to vindicate part of his first principle of justice.

Bayesian Rationality and Social Epistemology

This chapter uses epistemic game theory to expand on the notion of social norms as choreographer of a correlated equilibrium, and to elucidate the socio-psychological prerequisites for the notion that social norms implement correlated equilibria. The correlated equilibrium is a much more natural equilibrium criterion than the Nash equilibrium, because of a famous theorem of Aumann (1987), who showed that Bayesian rational agents in an epistemic game G with a common subjective prior play a correlated equilibrium of G. Thus, while rationality and common priors do not imply Nash equilibrium, these assumptions do imply correlated equilibrium and social norms act not only as choreographer, but also supply the epistemic conditions for common priors.