When is overconfidence useful (if ever)?

(This post is part of a series of Open Questions, in which I describe important questions about which thoughtful people disagree)

This might be the most persistent divide I see between the tech/entrepreneurship communities on the one hand (who tend to be very pro-overconfidence), and the finance/rationalist/academic communities on the other hand (who tend to think overconfidence is harmful).

Key arguments against overconfidence:

  1. You end up choosing bad projects if you’re overconfident — you could instead have been doing something else with a higher expected payoff
  2. You lose the ability to fix flaws (in your own skills, or your strategy) if you’re convinced they don’t exist

Key arguments for overconfidence:

  1. Without overconfidence you’d never have the motivation to achieve anything ambitious
  2. Overconfidence is self-fulfilling; believing in yourself makes you more effective
  3. Overconfidence convinces others (investors, employees) to trust and follow you

Looking at these arguments suggests a few candidates for “cruxes” on which this disagreement might rest. For example, if you believe overconfidence is net good, that might be because you believe that a person’s confidence and motivation are more pivotal in determining his success than his objective skills or strategy. (Whereas someone who thinks overconfidence is net bad might reverse the relative weights on those factors.)

Another possible crux involves the type of overconfidence. Perhaps overconfidence about one’s abilities is net good, because the social benefits of that type of overconfidence are especially high, but overconfidence about one’s knowledge is net bad.

One resolution that’s been suggested to the tradeoff between overconfidence’s costs and benefits is to separate your emotional/motivational attitude from your epistemic beliefs. That is, you want to be able to adopt a cheerfully confident attitude while saying “Yup, this project has a low chance of success, but it’s totally worth trying anyway.” I know several people who do this successfully, and their attitude seems to inspire confidence in others, as well (taking care of the third concern in the “pro-overconfidence” list).

The crux here seems to be whether that ability is easily replicable or not. If it’s too difficult for most people to be (emotionally) confident while being (epistemically) well-calibrated, then overconfidence could be the best patch for humans in some cases.

Also, it’s important to note that all of the above arguments are about why overconfidence is bad or good for the individual. But there’s another set of arguments at the group level: even if overconfidence is bad for you, some people argue, it’s good for society. Overconfidence means that 10,000 individual entrepreneurs end up worse off because they all wrongly think they’re going to create the next Google — but as a result, society gets one Google.

Related open questions: Rationality vs. Postrationality (coming soon)

Relevant reading: