They Behave For Me Can AI give teachers quality feedback? With Raj Chande
13 snips
Mar 6, 2026 Raj Chande, a classroom teacher and senior research associate focused on teacher expertise. He debates whether AI can help teachers improve. They discuss recording lessons, linking observations to pupil progress, training models on mentor judgements, and what transcripts can and cannot capture.
AI Snips
Chapters
Transcript
Episode notes
Researcher Turned Classroom Teacher
- Raj moved from behavioural economics research into full-time classroom teaching to close the gap between theory and practice.
- He trained via Researchers in Schools, found classrooms humbling, and joined Mossbourne where he now teaches maths and leads economics.
Data Can Reveal Teacher Value Add
- Teacher quality can be estimated more reliably than current vibes-based judgements using rich, linked pupil data.
- Rob Coe's approach projects pupil outcomes from prior year performance and controls for factors like prior attainment, pupil premium and cohort effects to surface residual teacher effects.
Within Cohort Comparisons Reduce Bias
- Comparing teachers within cohorts and controlling for many contextual variables reduces biases from set allocation or pupil selection.
- The model uses end-of-year projections and peers' performance across subjects to avoid crediting teachers for prior gains they didn't produce.