Sunday, November 18, 2012

Kahneman's Muddles: Consistency and Coherence

I'm reading Daniel Kahneman's Thinking, Fast and Slow. It is a very interesting book. When Kahneman is on his own ground, psychology, he seems to be a brilliant thinker. (Not being a psychologist, I feel the need to hedge this with a "seems to be": I'm really not qualified to judge!) I think he probably deserved the Nobel Prize he received.

But as soon as Kahneman starts talking philosophy, he begins to make terrible errors: we might say he is suffering from the "illusion of understanding." Let us examine a few of these, starting with consistency and coherence.

Kahneman litters his text with statement such as "we are prone to exaggerate the consistency and coherence of what we see," or we "produce a representation of reality that makes too much sense." This is because "we are pattern seekers, believers in a coherent world."

At first glance, it might seem that this is just a psychological truth, something Kahneman has discovered in the course of his research. But that is absolutely wrong, and it is important to see why in order to understand Kahneman's mis-step.

What are the actual research results that "back" Kahneman's claims? They are the results of many experiments which consistently demonstrate -- and I fully believe they do demonstrate this! -- that people are very prone to jumping to a conclusion, to believing they have an explanation for something or see a pattern in events without actually having an explanation or actually having detected a pattern. All well and good. So what can we conclude from that?

Well, imagine we study gamblers at the roulette wheel at casinos. We find them often saying things such as, "Fourteen is going to hit this time: I know it will!" We can say that people have unjustified confidence that a particular spin of the wheel will result in some particular number as a result.

What would be completely unjustified, though, is concluding that "people are prone to exaggerate the likelihood that spins of the roulette wheel achieve a definite result," because, of course, every spin of the wheel achieves a definite result. The gamblers problem was not their belief in definite results, but their too hasty conclusion about which particular definite result would be achieved.

Here is the (faulty) syllogism being used by our gambling investigator who does reach the conclusion stated in the previous paragraph:

p1) Some events of category Y have a property from property set X with some frequency between 0 (no events in Y have a property from X) and 1 (all events in Y have a property from X).
p2) People can be shown to pick out a property x from set X and assign that property to a member of Y too readily.
c) Therefore the frequency of events from Y having a property from X is lower than people believe it to be.

That syllogism is obvious nonsense, and Kahneman's statements about us exaggerating the consistency and coherence of the world have exactly its form. The fact that people are hasty in deciding upon explanations for events says absolutely nothing about how explainable events really are. In fact, reality could be far more consistent and coherent than we realize, and only the fact we have missed that leads us to reach such bad explanations!

But the situation is really even worse than that for Kahneman: his entire book rests upon an assumption that the world is consistent and coherent, because the whole book rests on experimental findings, and in an inconsistent, incoherent world, the result of experiments is utterly uninformative: whatever we "find" now may well be directly contradicted by the same experiment run one minute from now. If reality is really inconsistent, why not simultaneously hold that our experiments show many cognitive biases, but people do not suffer from any cognitive biases? (If the world were inconsistent, what would be the problem of having inconsistent beliefs?) So Kahneman's philosophical conclusions are directly contradicted by the excellence of his own work in psychology.

No comments:

Post a Comment

Current review queue

Pearce: British Journal for the History of Philosophy Deneen: The American Conservative Chao-Reiss: Computing Reviews