#231 Productive failure with Manu Kapur

Craig Barton
Apple PodcastsSpotifyGoogle PodcastsAnchorPodbeanStitcherPocketCasts

Episode sponsor

If you’re a Head of Maths or an aspiring Head of Maths, please join our HOD Space community, where you can find CPD and resources as well as a whole community of people doing the exact same job as you. Start your free, no-obligation, no-card trial at hodspace.com.

Episode Overview

Craig speaks to Manu Kapur, Professor of Learning Sciences and Higher Education at ETH Zurich and the researcher behind productive failure. They cover what the framework is and is not, the mechanisms behind it, how long students should struggle, what AI tutors mean for learning before instruction, and the strongest challenges to his work.

Talking points

  • What productive failure actually is: a precisely specified design framework for initial learning, where failure means not reaching the solution to a deliberately designed problem. Manu distinguishes it from broader umbrella terms such as productive struggle and problem-based learning.
  • Why it works: Manu’s four As of activation, awareness, affect and assembly, and his argument that these mechanisms are present in all instruction, including direct instruction, but that productive failure optimises them for deep learning and transfer.
  • How long students should fail: Craig contrasts prolonged inquiry with Dan Meyer’s headache–aspirin approach, and Manu describes two protocols: quick-fire intuition building over five to ten minutes, and a longer invention activity of thirty to forty minutes in which generating three to five ideas is the sweet spot.
  • Subject, age and prior knowledge: why maths gives a clear failure signal, why the 2021 meta-analysis did not find the same benefits for younger children, and the distinction between diagnostic priors and prospective priors.
  • Students who give up at the first sign of failure: how intuition-building items allow students to experience some success before they have been taught, so that success can drive motivation.
  • The most convincing argument against productive failure: people’s aversion to failure, and Manu’s three kinds of failure, from failure through lack of effort, to high-stakes failure, to the safe, designed failure that aims to reduce the high-stakes kind.
  • AI tutors and productive failure: the distinction between learning and performance, why a tutor that can be pestered into giving the answer is a problem when learning something new but less so for experts looking to be more productive, and his lab’s work fine-tuning models on productive failure principles.
  • Individual AI work versus whole-class teaching, why students and adults need support to have AI conversations that are good for learning, and Manu’s preference for training bespoke models over relying on prompt wrappers, set against Eedi’s human-on-the-loop approach with Google DeepMind.
  • The school of the future and the role of the teacher: why Manu expects schools to look largely the same, his concern about shallow knowledge, and his view that domain knowledge will matter more, not less, for teachers.
  • Craig’s closing challenge from the Direct Instruction tradition of Siegfried Engelmann and Kris Boulton’s logically faultless communication, including the role of non-examples and whether productive failure loses the argument on efficiency.

Links from the show

Video

My usual plugs

Leave a Reply