When the COVID-19 pandemic cancelled exams in 2020, England’s qualifications regulator Ofqual needed a way to award A-level and GCSE grades without students sitting papers. As described in Ofqual’s official interim report, published on GOV.UK on August 13, 2020, grades were instead based on “evidence of their likely performance in the exams had they gone ahead.” Teachers submitted estimated grades and rank orders, and a statistical standardisation model adjusted those using each school’s historical grade distribution and the prior attainment of its students.
When A-level results were released on August 13, the standardisation produced a public crisis. Around 36 percent of grades came out one grade lower than teachers had predicted, and the model’s reliance on a school’s past performance meant high-achieving students at historically lower-performing schools were disproportionately downgraded, while small class sizes at some private schools were treated more leniently. Students protested with the slogan “the algorithm stole my future.”
Within days the government reversed course. Ofqual withdrew the calculated grades and awarded students the higher of the calculated grade or the teacher-assessed grade, one of the largest U-turns in UK education policy. The chief regulator resigned. Prime Minister Boris Johnson later blamed the episode on a “mutant algorithm,” a framing critics rejected as deflecting responsibility from the policy choices behind the model.
The episode is a textbook case of why a statistically defensible model can still be socially and politically unacceptable: aggregate accuracy meant little to an individual student whose university place vanished because of where they went to school. When automated decisions carry life consequences, fairness to individuals and the ability to appeal matter as much as overall calibration.