Wednesday, June 11, 2008

Experts Don't Know Shit



Tetlock shows conclusively two key points: First, the best experts in making political estimates and forecasts are no more accurate than fairly simple mathematical models of their estimative processes. This is yet another confirmation of what Robyn Dawes termed "the robust beauty of simple linear models." The inability of human experts to out-perform models based on their expertise has been demonstrated in over one hundred fields of expertise over fifty years of research; one of the most robust findings in social science. Political experts are no exception.
Secondly, Tetlock demonstrates that experts who know something about a number of related topics (foxes) predict better than experts who know a great deal about one thing (hedgehogs). Generalist knowledge adds to accuracy.
Tetlock's survey of this research is clear, crisp, and compelling. His work has direct application to world affairs. For example he is presenting his findings to a conference of Intelligence Community leaders next week (Jan 2007) at the invitation of the Director of National Intelligence.
"Expert Political Judgment" is recommended to anyone who depends on political experts, which is pretty much all of us. Tetlock helps the non-experts to know more about what the experts know, how they know it, and how much good it does them in making predictions.

The Myth of the Rational Voter: Why Democracies Choose Bad Policies

Mistakes Were Made (But Not by Me): Why We Justify Foolish Beliefs, Bad Decisions, and Hurtful Acts