The Missing Teacher In AI: Involving Teachers In Metadesign of AI to Ensure Fairness

The Project is focusing on potential offered, and challenges posed, by adaptive AI-based digital teaching materials. The aim is to involve teachers in developing design, metadesign and use of the systems. This is intended to ensure that the systems will be both fairer and better adapted to the learning environment.

Adaptive systems can be seen as a digital learning environment that automatically adapts teaching and teaching resources to the abilities and needs of the individual student. Intelligent adaptivity is when artificial intelligence (AI) is used for this adaptation. AI in teaching, for instance, can be used to monitor students’ progress, ascertain their current strengths and difficulties, and give rapid feedback in the form of explanations and appropriate tasks.

So far these systems, usually abbreviated AIEd – AI in Education – have mainly been designed by computer scientists, designers and systems developers with little, or no, involvement of teaching professionals.

The researchers see a pressing need for greater awareness of the possibilities and risks entailed in the use of AIEd, and the extent to which it can discriminate. 

The project team therefore wants to involve teachers who are specialized in students’ educational needs. In this way they plan to ensure the systems do not become entirely controlled by algorithms, which they consider can result in discrimination, injustices, and other adverse effects. In their view, the knowledge, skill and experience of teachers are essential to the design and acceptance of fair AIEd.

They also point out that the systems have been developed without knowledge or consideration of how teachers will use them in their teaching. They think there is therefore a need for further research into ethical and pedagogical approaches in the use of AIEd.

One issue they raise is that if an AIEd system is designed to predict and alert the teacher to students who run the risk of falling behind, how should that category of students be defined? And how should the result of any such screening process be presented to the teacher?

Their earlier research in the field has shown that current AIEd applications on the Swedish market are “blackboxed”, rendering it difficult or impossible to interpret and explain the recommendations given to teachers and students.

The researchers have found that when teachers are unable to understand, use and influence an AI system, they lose confidence in the systems. They therefore consider it essential to involve teachers, since teachers ought to be able to understand and challenge algorithmic decisions and predictions, but also so they can derive practical benefit from the support offered by the systems.

Affiliated with WASP-HS

This project is affiliated with WASP-HS and generously funded by the Marianne and Marcus Wallenberg Foundation.

Principal Investigator(s)

Johan Lundin
Professor, University of Gothenburg

Project Member(s)

Marisa Ponti
Senior Lecturer, University of Gothenburg

Martin Tallvid
Researcher, University of Gothenburg

Marie Utterberg Modén
PostDoc, University of Gothenburg

1 July 2022 until 1 July 2025