EdTechnical Assessment in Education: To AI or Not to AI?
Aug 14, 2025
Dylan Wiliam, Emeritus Professor at UCL Institute of Education, dives into the transformative intersection of formative assessment and AI in education. He reveals the reasons why formative assessment is underutilized despite its effectiveness. Wiliam discusses AI's potential to automate summative assessments while warning against its limitations in providing meaningful feedback. He advocates for live oral exams and portfolio assessments to enhance learning authenticity and reduce cheating, offering a fresh perspective on the evolving teacher-student dynamic.
AI Snips
Chapters
Books
Transcript
Episode notes
Assessment As The Teaching–Learning Bridge
- Effective teaching starts from what learners already know and assesses to draw conclusions about their thinking.
- Formative assessment is the bridge between teaching and learning and should inform what to do next.
Add One Planned Checkpoint Per Lesson
- Plan at least one checkpoint per lesson to check understanding using simple, quick techniques like whiteboards or finger voting.
- Use the checkpoint to make smarter decisions about what to teach next rather than relying on a single confident student's answer.
LLMs Good At Grading, Weak At Learning Models
- Large language models are strong at scoring student work against rubrics and can reduce teacher marking workload.
- However, LLMs lack a model of student cognition and so struggle to deliver feedback that reliably moves learners forward.
