Goodhart’s Law and the Measurement of English Proficiency

Goodhart’s Law was first proposed by the British economist Charles Goodhart. In essence it states that, “When a measure becomes the goal, it ceases to be a good measure.” Measurements are often used as a proxy for performance. For example, it’s sometimes reported that in Soviet Russia, when the success of nail production was measured by quantity of nails, many tiny nails were produced. When it was measured using weight of nails, smaller numbers of large nails were produced. The measure became the target, and gaming the system created the illusion of success.

In U.S. university admissions, the TOEFL is the most common measure of the English proficiency of international applicants. It’s easy to understand why the complexity of language proficiency needs to be reduced to a small set of numbers when large quantities of applications have to be evaluated. Unfortunately, TOEFL preparation is very often a great example of Goodhart’s Law in action: many students focus on attaining the necessary score rather than comprehensively working on the cognitive-academic language skills and cultural skills they need to succeed in the U.S. university, and this can result in serious challenges for those students. Once matriculated, as those students seek to earn good grades – a proxy measure for learning – they may wind up trying to game the system by plagiarizing, using online essay services, cramming at the last minute, or begging the instructor for a better grade.

Although it would present practical difficulties, it would serve everyone better – schools and students – if the schools used a broader set of mechanisms to determine English proficiency. These might include evidence of English (not just test prep) study, Skype, phone, and in-person interviews, recorded presentations by applicants, synchronous online discussion groups, and reports from instructors in intensive English programs who have first-hand – not proxy – knowledge of the students’ English.