Mr Barton Maths Podcast #215 AI in Education with Daisy Christodoulou
Mar 11, 2026
Daisy Christodoulou, Director of Education at No More Marking and author on assessment, discusses AI in assessment and the history of automated marking. She covers comparative judgment versus absolute marking, why human oversight remains crucial, AI transcription pitfalls, effects on teacher development, and visions for blended high-tech infrastructure alongside traditional classrooms.
AI Snips
Chapters
Transcript
Episode notes
AI Caught A Teacher's Handwriting Bias
- In a trial Daisy showed a teacher a pair the AI and teacher disagreed on; the teacher admitted her handwriting bias caused the mistake.
- Large disagreements so far have often revealed human error, not AI error.
Transcription Errors Can Mask True Performance
- Handwriting transcription remains a weakness: LLMs sometimes 'improve' messy student writing, producing a better-seeming transcript.
- That transcription error, not the judgment step, causes many apparent AI-marking mistakes.
Embed Lightweight Human Oversight
- Keep a human in the loop but design it efficiently so it doesn't double workload; e.g., 10% human judgments can mean every script is seen twice by humans.
- Humans deter gaming and preserve teacher development by keeping engagement with student work.