neur.ro

Circle of competence

mental model · origin: practice · evidence: contested

In short

Your circle of competence is the area where you really know what you’re doing. The idea comes from investing. In his 1996 letter to shareholders, Warren Buffett wrote that an investor doesn’t need to be an expert on every company. They only need to be able to evaluate the companies within their circle of competence: “The size of that circle is not very important; knowing its boundaries, however, is vital.” The second part is the hard one. Psychology shows that people judge their standing relative to others poorly, and part of the research suggests that the weakest performers in a domain are also the least able to see their limits.

What it says

The core study. Kruger and Dunning (1999) gave Cornell students tests of humor, logical reasoning and grammar. They then asked them to estimate their percentile relative to their peers. Participants in the bottom quarter greatly overestimated their results: although their scores put them at the 12th percentile on average, they placed themselves at the 62nd. Those in the top quarter slightly underestimated themselves.

The authors’ explanation: the skills you need to do something well are often the same skills you need to recognize that you did it well. Someone who doesn’t know the rules of grammar can’t spot their own grammar mistakes either. Two results support the explanation:

The authors themselves discussed a statistical explanation, regression to the mean. Someone with a very low score has almost no room to underestimate themselves. They argued that regression doesn’t explain the whole effect, because the logic lesson changed the estimates.

The logic test in Kruger and Dunning (1999). The bottom quarter was at the 12th percentile on average but estimated the 62nd. The top quarter was at the 86th percentile but estimated the 68th. Where they were and where they thought percentile on the logic test (Study 2) Bottom quarter actual 12 estimated 62 Top quarter actual 86 estimated 68 average = 50
Data from Kruger and Dunning (1999), Study 2: 45 students, split into quarters by test score.

The critiques. The finding became famous as the “Dunning–Kruger effect”, but its interpretation has been contested for more than 20 years:

The replies. Ehrlinger, Dunning, Kruger and colleagues (2008) found the same lack of self-insight among weak performers in real-world settings and when participants had an incentive to be accurate. Jansen, Rafferty and Griffiths (2021) built a mathematical model of self-assessment and repeated the original study at scale, with about 4,000 participants in each of two studies. Their results support the idea that, in grammar and logic, weak performers are less able to tell whether an answer is correct.

In short: everyone judges their standing relative to others poorly. That weak performers overestimate themselves the most shows up consistently in the data, but part of the pattern is a statistical effect. How much remains once you subtract it is still debated.

Example

An example built for this text: a programmer who is good at web applications decides to run their own email server. They don’t know what they don’t know: DNS records, sender reputation, anti-spam rules. Everything seems to work, until their messages start landing in spam. The problem isn’t that they’re bad at email servers. The problem is that they didn’t notice they had left their circle.

How to apply it

The steps below are a practical approach we propose, not a tested method.

  1. Draw the circle explicitly. Write down the areas where you have verifiable results, not just experience or interest. Buffett says the size of the circle matters little. The boundary is what you need to know.
  2. Check the boundary with outside feedback, not with a feeling. All the studies above show that self-assessment is noisy. Write down in advance what you expect to happen, then compare it with the actual result.
  3. Be more careful on hard or new tasks. In Burson et al. (2006), on difficult tasks even the best performers misjudged themselves. A feeling of confidence is a weak signal exactly where it matters most.
  4. Learn a little before you judge. In Kruger and Dunning (1999), a short lesson made estimates more realistic. Sometimes the fastest way to see your limits is to learn the basics of the domain.
  5. At the edge, borrow other people’s competence. When a decision falls outside your circle, ask someone inside theirs, or lower the stakes.

Limits and nuances

Sources

See also: Planning fallacy, Regression to the mean, Via negativa