MCAS AI Grading Glitch Mis-scores 1,400 Student Essays Across Massachusetts — A Cautionary Case in Automated Assessment

During the 2025 Massachusetts Comprehensive Assessment System (MCAS) administration, an automated AI scoring tool — used to grade open-response essay questions, with roughly 10% of AI-scored essays normally re-checked by human readers — produced a temporary scoring glitch. A teacher at the Reilly Elementary School in Lowell noticed a discrepancy between a student's essay and its assigned score and alerted school and district leaders in mid-July 2025.

The Massachusetts Department of Elementary and Secondary Education (DESE) subsequently found that roughly 1,400 of the state's 750,000 MCAS essays had been mis-scored, with some student responses wrongly marked '0' when they should have scored as high as 6 on a 1-to-7 scale. Districts across the state were affected, including Attleboro, Fall River, New Bedford and Taunton in Bristol County; DESE reported that close to 200 schools in roughly 145 districts were touched by the error, though the average impact was only one or two mis-scored students per affected district. All flagged essays were rescored and every affected district was notified, with corrected results issued by August 2025.

This is a documented failure, not a success story: an AI scoring tool operating at full state scale mis-scored student work, and the error was caught by a classroom teacher's vigilance, not by the AI system or its own quality-assurance process — with only about 10% of essays otherwise receiving routine human regrading. The case is included here as a cautionary example: it shows both the risk of automated assessment at state scale and the value of human oversight, and it illustrates a comparatively transparent post-incident response by the state education agency, which does not offset the fact that the underlying error and the thin human-review safety net were the root cause.