Emotion, Reason, Models

Jonah Lehrer has a piece in Wired recently that says the following:

While there is an extensive literature on the potential wisdom of human emotion – David Hume was a prescient guy – it’s only in the last few years that researchers have demonstrated that the emotional system (aka Type 1 thinking) might excel at complex decisions, or those involving lots of variables. If true, this would suggest that the unconscious is better suited for difficult cognitive tasks than the conscious brain, that the very thought process we’ve long disregarded as irrational and impulsive might actually be more intelligent, at least in some conditions.

The latest demonstration of this effect comes from the lab of Michael Pham at Columbia Business School. The study involved asking undergraduates to make predictions about eight different outcomes, from the Democratic presidential primary of 2008 to the finalists of American Idol. They forecast the Dow Jones and picked the winner of the BCS championship game. They even made predictions about the weather.

Here’s the strange part: although these predictions concerned a vast range of events, the results were consistent across every trial: people who were more likely to trust their feelings were also more likely to accurately predict the outcome. Pham’s catchy name for this phenomenon is the emotional oracle effect.

No doubt this is true in many situations, and Lehrer pointed to many of them in his book How We Decide. But it’s important to distinguish between individual reason and basic formal modeling.

I’m in the middle of a lecture on modeling right now and via that course, here’s a graph from Tetlock:

Hedgehogs, in Tetlock speak, are people who reason based off of a single mental model – they know one big thing and it informs their reasoning. Foxes, who do much better at prediction, know many things, and reason using many mental models.

But note up top that formal models beat both of those. Of course, neither of these is speaking exactly to intuitive thinking exactly, but I’d wager the same effect would be seen in the cases Lehrer cites.

Kahneman addressed this in Thinking, Fast and Slow, (I wrote about that part of it here):

…it is possible to develop useful algorithms without any prior statistical research. Simple equally weighted formulas based on existing statistics or on common sense are often very good predictors of significant outcomes. In a memorable example, Daws showed that marital stability is well predicted by a formula:

frequency of lovemaking minus frequency of quarrels

You don’t want your result to be a negative number.

That’s why, for instance, I basically outsource my estimations of the upcoming presidential campaign to Nate Silver, who essentially has built a fairly simple model based on a few variables known to be important. His model is unlikely to be perfect, but highly likely to be better than both my rational and my intuitive guess.