Diagnostic Utility

What it is

Prelec and Bodner start from a simple observation. Economists usually model a choice by what it produces: the outcome, discounted by cost and risk. They add that a choice also tells the chooser something. Taking the action is evidence about a trait, such as generosity, self-control or health, that the person cannot observe directly.

On their account a choice therefore carries two kinds of reward. Outcome utility comes from the action's direct consequences. Diagnostic utility is the change in self-esteem from updating one's self-image after the choice. A person who gives to charity gets the outcome, which is a donation received, and also the news that they are the kind of person who gives.

The authors separate three motives for caring about that news. It can be intrinsic, when a person cares about the trait itself. It can be instrumental, when a person cares about what the trait predicts. And it can be magical, when a person believes that their action signals how other people will behave, which is the logic of the voter who feels that turning up shows what like-minded people will do.

They also cross two questions: whether diagnostic motives affect choice, and whether people discount their own actions for those motives. Their rational case, in which people do discount, predicts a drive toward behavioural perfection but not a better self-image on average, because a person who knows they are signalling learns less from the signal.

In effect

The experimental anchor is a 1984 paper by George Quattrone and Amos Tversky. Students were told that the way exercise changed their tolerance of cold water signalled either a long, healthy life or a shorter one. On the second trial their tolerance moved toward the good-news pattern. Prelec and Bodner read this as diagnostic utility at work: the students were not trying to change their health, only the news about them.

The chapter lists everyday cases. Giving to charity, drinking before noon, jogging in a heat wave, buying an expensive item on credit and voting all have effects that are small or nil, and all can be read as messages. In self-control, they argue, the anticipated news raises the temptation needed to make someone indulge, so that diagnostic motives support restraint. They also cite Bodner's earlier work on contingent pledges: people who saw themselves as insufficiently altruistic pledged more when the chance of having to pay was low.

A later line of work on costly good deeds fits the same shape. Ayelet Gneezy and four colleagues found in a laboratory experiment and a large field experiment that costly prosocial acts signalled a prosocial identity and were followed by consistent behaviour, whereas costless acts signalled less. Cost, on this reading, is part of what makes the news believable to the person receiving it.

For consumer behaviour, the implication is a question and not a conclusion: how much of a purchase is bought for the item and how much for what buying it says about the buyer. The experiments above do not measure that split for ordinary shopping.

What it does not say

It does not say that the model has been tested in the shop. The Prelec and Bodner chapter is theoretical and reports no new experiments. It leans on the cold-water study and on work on self-reports, problem solving, charitable pledges, self-handicapping and motivated recollection.

It does not say the assumptions are comfortable. The authors acknowledge that the central assumption, which requires people to discount their own motives when reading their actions, may not be psychologically accurate. They also say the model assumes people do not learn across repeated choices, which they call implausible, and that it explains neither willpower nor effort. They describe their most extreme case, paranoid self-scrutiny, as dubious as a model of normal behaviour.

It does not answer the objection from The Elephant in the Brain, whose authors argue that incentives which begin and end inside one's head lead nowhere (Simler and Hanson, 2018, p. 295). That rule is asserted in the book and not tested, but the experiments behind diagnostic utility do not separate private signals from observed ones either, so the question stays open.

It does not make diagnostic utility a synonym for self-perception. Bem's theory describes inference from behaviour that has already happened. Diagnostic utility describes a reward that can steer the behaviour before it happens.

---