Autism does not limit strategic thinking about another person’s thoughts in an economic game

Autism is often associated with an impaired ability to reason about another person’s thoughts and actions. But new research suggests that individuals with autism spectrum disorder can predict the actions of others just as well as those without ASD in some cases.

In the study, 30 high-functioning adults with ASD and 51 adults without ASD participated in a simple game known as the Beauty Contest. In order to win the game, players must think of the strategies the other players will use based on their beliefs about how other people will behave.

The study found no discernable difference in the manner in which adults with and without ASD responded in the Beauty Contest.

The findings were recently published in the peer-reviewed scientific journal Cognition. PsyPost interviewed the study’s corresponding author, Peter C. Pantelis of Indiana University-Bloomington. Read his responses below:

PsyPost: Why were you interested in this topic?

Pantelis: I study what psychologists call ‘theory of mind’—our ability to reason about other people’s mental states (goals, beliefs, intentions, etc.). When I was in graduate school, a paper by Camerer & Fehr (2006) got me interested in an economic game called the Beauty Contest, because I thought it really tapped into theory of mind in a very elegant way.

The game goes like this: Everyone picks a (whole) number between 0 and 100. The winner is the person whose selection comes closest to the 2/3 of the mean of all numbers chosen.

Consider the number you would choose. Although according to game theory, “rational” players should all say “0,” people typically give a wide variety of answers. Some will answer randomly (on average, 50). Some will assume the other players will answer randomly, and so give an answer that’s 2/3 of what random players would say, on average (~33). But then some will take a more sophisticated strategy, accounting for *the beliefs players may have about other players’ beliefs.* For example, they may *believe that others believe other players will answer randomly*, and therefore select 2/3 of 33 (something like 22). Or they may dive even deeper into this reasoning process, selecting 2/3 of 2/3 of 33…and so on and so forth. Analyzing the responses a group of individual players make in this game gives you some very interesting insight into how they might be reasoning about the strategies of the other players—a potentially complex reasoning process about other people’s minds.

At the same time, I was also becoming interested in autism spectrum disorder (ASD), because it has been a longstanding hypothesis that people with this condition have particular trouble with theory of mind—Especially when it’s the kind of situation where you have to reason about the mind of another person who, *in turn,* is representing the mental states of others. Not that there aren’t many perspectives and hypotheses about how to fundamentally characterize ASD, but this has been a pretty popular take for a long time (Simon Baron-Cohen made it popular in the 1980s).

So when I got the opportunity to work with adults with ASD as a postdoc, I finally got a chance to see how these individuals would approach the Beauty Contest.

What should the average person take away from your study?

First of all, I’m trying to build an important bridge between two areas of scientific study: theory of mind (which has been a special focus for developmental psychologists for decades), and game theory (especially as adapted relatively recently by behavioral economists, in a way that often captures actual human behavior quite well). In both cases, we’re concerned with how people reason about how people are likely to act, based on what they know, and what they are trying to do. But there hasn’t been much of a connection made between the two fields until fairly recently (through the work of researchers like Rosemarie Nagel, Colin Camerer, Wako Yoshida, and a few others).

I think studying theory of mind using the tools and modeling approaches of game theory (especially the ‘behavioral game theory’ version) could yield some great insights, especially in the study of ASD. I was surprised to learn how few experiments have actually been done to see how people with ASD approach economic games—I could find less than five studies of this kind in the entire literature.

And that literature is a bit of a mixed bag. I found that high-functioning adults with ASD (on average with the same IQ as the control group in the study) gave a virtually identical distribution of responses when they played the Beauty Contest game. That was surprising to me, given that this game to me seems very likely to tap into some pretty complex reasoning about other people’s minds—and people with ASD are supposedly not very good at that. On the other hand, it’s an explicit, abstract task for which participants are extrinsically motivated (they get money if they win the game). Increasingly, I get the sense that the most persistent deficits in ASD—the ones most likely to persist into adulthood—might be more related to how attentional and perceptual processes spontaneously interact with “theory of mind” in dynamic, real-world scenarios.

Are there any major caveats? What questions still need to be addressed?

Some people didn’t buy that the Beauty Contest is a “real” test of theory of mind. I would agree that when people answer randomly, then of course they aren’t really tapping into their theory of mind. But when they have to explicitly reason about the strategies other people may bring to the game, and when that often involves some pretty complex reasoning about the beliefs other players may have about what *you* might do—well, if that’s not theory of mind, I don’t know what is.

I like this task because, although the reasoning process you can bring to it can be quite complex, the game itself is very simple. There are no complex social cues you have to perceptually process. The math involved is trivial (and we allow subjects to use a calculator or pen and paper if they want).

Some other, more “canonical” tests of theory of mind actually mix quite a few things together, which makes interpreting performance on them difficult. For example, consider the “Reading the Mind in the Eyes” task (developed by Baron-Cohen and colleagues in 1997), which has been used in hundreds of studies and cited thousands of times. You see a cropped photo of the region around a person’s eyes, then must choose (out of four provided options, in the revised version of the task) what emotion you think the person is experiencing.

One critical part of that task is the social ability to read subtle cues to emotion from the eye region of another person. But then there is the vocabulary of emotion words you bring to the study, and your ability to map the emotion you infer from the eyes to the “correct” label. And some of the word choices might be a priori more salient than others to begin with—another variable thrown into the mix.

This is often framed as a bona fide “mentalizing” or “theory of mind” task, but even though it draws on quite a few perceptual and cognitive processes (like those I just listed), one ability that it doesn’t actually require is explicitly representing the mental state of the person. You can “read” the social cue; you don’t need a richer representation of the inner workings of the person’s mind to arrive at a response. There isn’t even any context that might make it a more complex inference.

Is there anything else you would like to add?

I am encouraged that there is an interdisciplinary literature emerging that is finally building bridges between developmental psychology, economics, and cognitive science/computer science. More and more you see computational modeling or game theoretic approaches being brought to bear on the study of theory of mind. Our study extends this collaboration into clinical psychology, an approach which I think should yield some truly novel insights into disorders like ASD in the near future. Be on the lookout for more and more “computational psychiatry” papers, in which rigorous modeling approaches are applied to clinical populations (there’s been some especially interesting schizophrenia research being done with this approach) in ways that have only recently been made possible through intense multidisciplinary collaborations.

The study, “Autism does not limit strategic thinking in the ‘beauty contest’ game“, was also co-authored by Daniel P. Kennedy.