Who needs rationality?
It seems that irrationality is everywhere. Economics long assumed that people are perfectly rational, but now behavioral economics is the hottest sub-field and Daniel Kahneman got a Nobel prize for showing how people make biased decisions. Even so, there is no shortage of criticism of the remaining reliance on rationality in economic models. Books like Nudge: Improving Decisions About Health, Wealth, and Happiness and Predictably Irrational: The Hidden Forces That Shape Our Decisions become bestsellers. A smart, earnest community of people who want to become more rational has grown up around websites like Less Wrong and Overcoming Bias.
In the face of all this, Fast and Frugal: The Tools of Bounded Rationality (a part of this book) argues that people are actually excellent decision-makers. The author concedes that people rarely act as perfect Bayesians who do constrained optimization of their utility function. However, he claims that the heuristics that ordinary people use actually perform very well in most environments. These heuristics work more quickly and with less information than the corresponding “rational” method. The paper includes criticisms of many of the studies showing people are biased, as well as other studies showing examples of heuristics with good predictive power.
Some questions stand out to me:
1. To what extent are markets an environment in which these heuristics perform well? As we find out what heuristics people actually use, economists can build models based on them; to some extent this is being done already with the general idea of “bounded rationality”. If some markets are not a friendly environment for these heuristics, how can public regulation or entrepreneurial information-providers improve them?
2. What are the arbitrage opportunities? The investment funds of behavioral economists have not been especially successful as far as I know, but perhaps the fund that really figures out how people apply a heuristic will be, or perhaps there is another market where this would work. Of course, casinos and credit card companies seem to have done this already.
3. In some situations (in markets or otherwise) there are returns to cognitive diversity. People with less common heuristics or who are proper rationalists may succeed because they are doing something different rather than because their strategy is strictly dominating per se. Of course, the reverse may also be true in other situations and heuristics could pull everyone to a bad equilibrium.
4. Heuristics may work well for people in most environments, but our environments are becoming increasing different from those we evolved to succeed in. More and more decision-makers are not human, and the question of how to program them is a very different one from the question of how humans should think.
HT: This Less Wrong post, it is good that they consider arguments which seem to undermine much of their project. Of course, for all the people working on AI the computer exception is a major one.
I realize my writing style is trying to imitate Tyler Cowen‘s. This attempt is mostly unconscious and is doomed to failure as I can never be so succinct.