“Wow, we are playing poorly right now. What’s the deal?” – Coach
“They look tense, I think they are too wound up.” – Assistant
“I think you’re right. Time-out!” – Coach
This is an example of a retroductive (sometimes called abductive) inference, or an inference from a surprising fact to another possible fact that explains the surprise. In the case above, poor play is the surprise, and retroduction supplies the cause: anxiety.
Logician Charles Peirce argues that retroduction is the first step of any scientific inquiry. Moreover, he argues we are wired to make this kind of inference. Think about how long it takes you to guess at an explanation for some surprise event. Not long. Think about how many guesses you have to go through in order to find the truth. Not many. In fact, Peirce suggests that just as the gift of the spider is to weave webs, the gift of human beings is to guess well, that is, to find causes quickly by retroductive inference. In some cases he even uses the language of religion. He writes in a letter: “Retroduction gives hints that come straight from our dear and adorable Creator…It is the side of human intellect that is exposed to influence from on High.”
In a newly published book, Thinking, Fast and Slow, psychologist Daniel Kahneman shows (via experimental research) that the region of our mind he calls System 1 (roughly the unconscious) is, in fact, very adept at making causal inferences (Peirce’s retroductions.) They appear, just as Peirce notes, to our conscious minds (Kahneman’s System 2) as suggestions, as intuitions. Kahneman, however, also shows that System 1 generates other intuitions and suggestions to problems that are not logically of the form of a retroduction. Consider the following exchange:
“What do you think, is she good enough for our team? Should we recruit her?” – Coach
“I think so, she looks pretty good. She certainly looks the part of an outside of hitter.” – Assistant
The assistant in this case has an intuitive suggestion supplied by System 1. But the suggestion is not of the nature of a retroductive conclusion. The coach is not asking the assistant for a specific cause that might explain something. Rather, the coach is asking for an evaluation. Making evaluations is essentially a kind of inductive inference that typically involves statistical reasoning. A good outside hitter, by one measure, is one that kills the ball 40% of the time and makes less than 10% errors.
It turns out, Kahneman claims, that System 1 is terrible at induction and statistical reasoning in general. Indeed, System 1 will not even attempt to make an induction, but will perform some other type of inference (usually something like an inference to similarity – which it is good at – as in, “she looks like other good outside hitters.) Sometimes this will work, but most of the time it won’t. Moneyball is an extended analysis of what happens when we substitute conscious evaluations (System 2-guided statistical studies) for intuitive suggestions in order to do our player evaluations: essentially, we win more.
What does all this mean? When we have to make evaluations of players for recruiting, playing time, etc., don’t trust your intuitions. Instead, spend time doing stats and trust those. But, when the situation calls for an explanation, as in, what is going on here?, what is leading to this on the court?, trust your gut. You certainly could be wrong, but after just a few educated guesses, you’ll likely hit on the cause.