Failure of AI to predict Baccalaureate grades

What Happened

In August 2020, following the cancellation of A-Level and GCSE exams due to the COVID-19 pandemic, the UK's Office of Qualifications and Examinations Regulation (Ofqual) used an algorithmic model to predict and assign grades to hundreds of thousands of students. The algorithm combined teacher assessments with historical school performance. The system resulted in mass grading failures, systematically downgrading 40% of students' teacher-assigned grades, disproportionately penalizing students from public schools and disadvantaged backgrounds while favoring private schools. Following massive student protests, the government withdrew the algorithm and restored the teacher-assigned grades.

How MEG 1 would have acted

The grading algorithm failed both the Dynamic Accuracy Index (DAI) and the Index of Safety and Responsibility (ISR) audits due to its systemic inaccuracy, bias, and violation of procedural fairness and educational access.

MEG 2 Analysis - Legal Framework

Agent level: N2 (management) - automated decision-making system used by a public authority to allocate critical life-altering scores (grades).
Cause (6.1):
(a) system defect - design flaws and bias built into the predictive grading model (design defect), attributable to the developer.
(b) autonomous decision error - the system's decision to mass-downgrade student grades based on historical school performance proxies.
Attachment of liability: Ofqual as the operator is fully liable for the systemic harm caused by deploying an unverified, biased system; the developer is liable for the model's design defects (6.1a).

Procedural Mechanism

6.2 (architectural human confirmation): Decisions involving critical educational grading or certifications must require active, architectural human confirmation. The AI must act as a suggestion tool; a human educational board must verify and sign off before final grades are locked.
7.1 (independent forensics): The input parameters, the model's prediction weights, and the system's decision path must be securely logged in an independent ledger to allow for educational and public audits.
9.4 (least privilege): The grading system was granted direct authority over student certifications without independent validation layers, violating safe operational parameters.