Karl Wärneryd, economics, Stockholm School of Economics

2009-2013

The purpose of this project has been to theoretically investigate the evolutionary stability of beliefs not founded on evidence, such as religious beliefs. During the course of the project, attention has also come to be paid to issues of rational behavior more generally.

In economics, and today also largely in philosophy, rationality means having preferences and beliefs and acting so as to make oneself as well off as possible given those preferences and beliefs. We might call this instrumental notion of rationality act-rationality. But there is also another sense of rationality, one that, arguably, used to be the standard one: That of rationality as the use of reason--drawing logically correct conclusions from the evidence at hand. For simplicity, we might refer to this type of rationality as belief-rationality.

Recent work on experimental and behavioral economics has to a large extent focused on examining whether act-rationality is a reasonable theoretical assumption, and, to a lesser extent, on whether departures from act-rationality have important implications for the predictions of economic models. Relatively little attention has been paid to the issue of belief-rationality.

Act-rationality does not, of course, preclude belief-irrationality. There is nothing to stop us from applying the standard tools of economic theory to behavior that may seem to us irrational in the broader sense. Although it has seldom come up in economics--presumably because not much work has been done on applying economic theory to the content of religious belief--the question of whether one should try to give rationalistic explanations of superstitious practices is an old one in social anthropology. One part of the present project has focused on whether irrational beliefs can serve strategic purposes in social interaction.

A major result concerns the use of oracles, or conversing with the supernatural more generally, as a commitment device. In ancient times, a king might consult an oracle in order to decide whether to take his nation to war or not. The Greek historian Herodotus, himself a firm believer in the usefulness of oracles, in his Histories reports many such instances--including the tale of Croesus of Lydia, whose interpretation of a pronouncement by the oracle at Delphi as an encouragement to go to war had disastrous consequences. The classical oracles were notorious for the opacity and ambiguity of their advice. And somehow that seems to have been the point.

A central insight of game theory is that commitment to a strategy can be advantageous in a situation of strategic interaction. In the paper "Religion, Ritual, and Randomization" (under review) I investigate the implications of viewing the consultation of oracles, or the supernatural more generally, as a commitment to a randomized strategy in a situation of potential conflict with asymmetric information. Consider the owner of some resource who is attacked by a bandit who demands a ransom-- a situation that would not have been uncommon in the setting in which human behaviors were selected for and evolved. Suppose whether the owner's resource is of positive value or not is the owner's private information. If a refusal to pay the ransom is followed by outright conflict, the bandit's beliefs about the value of the resource will determine how hard he fights. In a game model of this situation
it holds that if the owner has the option of delegation his decision of whether to pay the ransom or not to a randomization device, he will want to delegate. This allows him to optimally manipulate the bandit's beliefs, and will lower the equilibrium probability of outright conflict. One way of achieving such delegation is through adopting a religious belief that can be publicly observed--and actually believing in its contents.

A second function of religion is as a cognitive scheme for perceiving reality. One view, associated with Stewart Guthrie and others, is that religion is essentially anthropomorphism, the assigning of human-like intentionality to natural phenomena. The idea is that human perception is necessarily of limited complexity. In a sophisticated variation on Pascal's wager, Guthrie suggests that our perceptive schemes fail to distinguish between the acts of intentional agents and natural phenomena in contexts where a failure to detect intentionality would be most costly.

It is instructive to consider an example from the natural world. The male robin becomes aggressive when it sees the red breast of another male of the same species. This makes evolutionary sense if there is competition among the males over, e.g., mates. But the male robin also goes into the aggressive stance when confronted with any other red object of similar size, such as a red mitten waved in front of it. Given that the robin's perceptual scheme cannot be of infinite complexity, it is less costly to mistake irrelevant red objects for other males than to fail to react to an actual competitor.

In the present project I have studied the evolutionary stability of perceptual schemes (or information partitions, in the language of decision theory) given that there are fitness costs associated with complexity (i.e., finer information partitions are more costly in evolutionary terms). Surprisingly little theoretical work has been done on this issue. The paper "Optimal Pareidolia" (work in progress; joint with Axel Bernergård) is intended to be a first step in this direction.

A related paper, "Observable Strategies" (under review), considers the implications of coarse information about opponents' strategies being available to players in a game. I show that for any noncooperative 2-player game in normal form, it is possible to set up such an information structure so that any feasible and individually rational payoff profile of the original game is attainable at a Nash equilibrium of the extended game.

All papers from this project are expected to eventually appear in leading economics journals.

Publications

“Religion, Ritual, and Randomization,” working paper

“Optimal Pareidolia” (tillsammans med Axel Bernergård), working paper

“Observable Strategies,” working paper