Circularity, and how to spot it

Examples of circularity

The best way to understand circularity is through some examples:

– On the one hand, high IQ is associated with higher education and higher-paying jobs. On the other hand, school exams and job interviews contain IQ tests – not actual IQ tests, but questions that are better measures of the candidate's IQ than their job-related abilities. Hence, there is a circularity: part of the reason high-IQ people have higher-paying jobs is that we designed exams and job interviews to be IQ tests.

– In finance, when a stock rises in price, it often triggers mechanisms that cause more traders to want to buy it, further increasing its price. The fact that the price is both outcome and cause is a circularity.

– Have you noticed, when studying history, how exceptional it is that the good guys always win in the end? Victors write the history books, and historians belong to the winners. This creates circularity: those being measured (the winners) are the same doing the measurement.

– On the 8th of January 2025, the New York Times titled "Meta [Facebook's holding company] says fact-checkers were the problem. But fact-checkers rule that false." Fact-checkers ruling on themselves creates an obvious circularity.

These examples show that modernity is full of circularities: contexts in which those doing the measuring and those being measured overlap significantly, or when the act of measurement itself influences the outcome.

Is circularity bad?

Circularity is not always bad, though it often is, especially when we are unaware of it. For example, in science and truth-seeking, circularity makes correlations appear as causations. In finance, it might mislead traders into believing an asset is better or more stable than it actually is, causing them to enter trades whose risk they do not understand. Hence, the importance of detecting circularity: to prevent costly mistakes.

Partial circularity

Note that circularity is not a binary property, either fully present or fully absent; often, circularity is only partial. For example, finance is full of circularities, because current prices influence future ones, as highlighted by the fact that some traders buy a stock for the sole reason that it has risen. However, circularity is seldom total, for past prices are only one of the factors considered in evaluating an asset, and estimations of intrinsic value matter, too.

How to detect circularity?

Here are two tests useful to detect circularity:

1. Given two metrics A and B, does the act of measuring A influence B? For example, in the case of IQ and life outcomes, the question is not only whether A (one’s IQ) influences B (their life outcomes), but also whether the act of measuring A (administering tests that, directly or indirectly, measure IQ) influences B (life outcomes). In this case, it does, for school tests and job interview tests have an obvious effect on one’s access to good schools and good jobs.

2. Given a metric, who defines how it should be measured, and were they selected based on that same metric? For example, to some extent, it can be said that academicians are both those who decide who is smart and are selected for being smart, leading to at least some partial circularity in the definition of smartness.

A positive answer to either of the points above indicates circularity.

Note: A reader submitted the topic of this article. If you also have questions you would like me to answer, let me know!