An AI grading tool used on Massachusetts's statewide MCAS tests mis-scored roughly 1,400 student essays across 192 school districts -- including one Lowell student whose essay should have scored six out of seven but got a zero.

Massachusetts uses an AI tool to help score parts of the MCAS state test, working from human-scored essay exemplars as a reference point at each score level. This year, the tool produced significantly low scores on about 1,400 essays across 192 districts, including examples reported from Lowell, Malden, and Medford. One documented case: a Lowell elementary student's essay should have scored six out of seven points, and instead scored zero.

The state's testing contractor rescored the affected essays before results were released publicly. Every updated score came out higher than the original AI score, and no student ended up with a lower score than they should have received. The errors never reached families as final results.

Massachusetts Education Commissioner Pedro Martinez called it a 'learning experience,' saying, 'I don't want to discourage the use of technology because it allows us to get more thoughtful on the assessments.' The state's education department (DESE) confirmed that humans already review 10% of AI-scored essays specifically to catch discrepancies like this one.

An AI tool made a real, measurable error grading real students' work -- at meaningful scale, across nearly 200 districts. It was caught before it affected any family. That's the system working as a safety net should. But it's also a clear example of why human review of AI-scored assessments matters, not a hypothetical concern.

▶ Watch the 30-second video version: https://youtube.com/shorts/8SVIB9jNoLM

Source: https://www.nbcboston.com/news/local/mcas-ai-glitch-school-districts/3832222/

Want practical AI guidance for parents and educators every week?
Subscribe: https://www.aibyage.com/?modal=signup&utm_source=beehiiv&utm_medium=newsletter