Saturday, February 07, 2015

Monkeys, Hot Hands, and Human Irrationality


Tom Stafford is a psychologist who blogs at Mind Hacks. A recent blog post discussed two common types of human irrationality: the gambler's fallacy and the hot hands fallacy. The gambler's fallacy is believing that a future random result has a greater-than-average chance of being different from a previous random result. An example is a gambler's belief that a streak of "red" results on a roulette wheel makes a "black" result more likely on the next spin. The hot hands fallacy is believing that a future random result has a greater-than-average chance of being the same as a previous random result. An example is a spectator's belief that a basketball player is on a "streak" and so should continue to be given possession of the ball as much as possible by his teammates.

Stafford claims that recent research on monkeys shows that humans are not being irrational when they commit the hot hands fallacy. His justification for this is that monkeys in the experiment were observed to make the same mistake as humans.
The reason the result is so interesting is that monkeys aren’t taught probability theory as school. They never learn theories of randomness, or pick up complex ideas about chance events. The monkey’s choices must be based on some more primitive instincts about how the world works – they can’t be displaying irrational beliefs about probability, because they cannot have false beliefs, in the way humans can, about how luck works. Yet they show the same bias.
Stafford's argument is flawed. Monkeys are not taught probability theory in school, they don't learn theories of randomness, nor do they pick up complex ideas about chance events. But this seems irrelevant to whether their behavior is rational or not. Being taught probability theory in school can hardly be a necessary condition for displaying irrational beliefs about probability. What makes beliefs irrational is that they are systematically biased in a way that does not reflect the underlying reality, This could be the case regardless of whether a person (or a monkey!) has taken a course on probability theory. If Stafford's argument were sound, then unreflective humans or humans who had no training in probability theory would not be capable of irrationality, and reflective humans and those with special training would be the only ones capable of irrationality. This seems obviously false.

Stafford next argues that there must be a good reason why the monkeys are behaving this way:
What’s going on, the researchers argue, is that it’s usually beneficial to behave in this manner. In most of life, chains of success or failure are linked for good reason – some days you really do have your eye on your tennis serve, or everything goes wrong with your car on the same day because the mechanics of the parts are connected. In these cases, the events reflect an underlying reality, and one you can take advantage of to predict what happens next. An example that works well for the monkeys is food. Finding high-value morsels like ripe food is a chance event, but also one where each instance isn’t independent. If you find one fruit on a tree the chances are that you’ll find more.
I don't doubt that this hypothesis (or something like it) explains why both monkeys and humans behave the way that they do. The problem is that this argument does not prove that monkeys and humans are always rational. Whether behavior is rational is context-dependent. The fact that the hot hands style of reasoning is rational in some contexts does not imply that it is rational in all contexts.

Finally, Stafford seems to conflate two senses of the word 'rational' here. The first sense is the usual one discussed in the biases and heuristics literature--something like "free from systematic bias". The second sense of 'rational' that Stafford introduces is something like "increases biological fitness". The latter comes into play in the previously quoted paragraph, and is made clear when he discusses the possible role played by evolution in shaping human and monkey behavior:
The wider lesson for students of human nature is that we shouldn’t be quick to call behaviours irrational. Sure, belief in the hot hand might make you bet wrong on a series of coin flips, or worse, lose a pot of money. But it may be that across the timespan in evolution, thinking that luck comes in clumps turned out to be useful more often than it was harmful.
The problem is that these two senses of 'rational' are distinct. Even if the hot hands bias is rational in the sense of "adaptive" or "increases biological fitness", this does not prove that it is rational in the sense of "free from systematic bias".

Stafford is correct to focus on the rhesus monkey study, which is genuinely interesting, and his blog is truly excellent. But I think Stafford's blog post shows the benefits of having some training in philosophy when trying to discuss the implications of scientific research. Of course, rare is the bird who has mastered both the science and the philosophy, which explains the difficulty here.
Post a Comment