The research field Explainable Artificial Intelligence (XAI) already provides approaches to generate automated explanations for Artificial Intelligence (AI) decisions. In order to be able to design and use explanations effectively, it is essential to find out how explanations affect the interaction of users with AI. In the German-Israeli project "Guess The City / explAIn - lehAsbIr", we are investigating what role explanations of modern XAI methods can play in the context of AI-supported learning - "Lehasbir" means "explain" in Hebrew.

The question of how XAI can support users in learning has hardly been considered scientifically. In practice, too, XAI is not yet frequently implemented in learning contexts, despite its great potential. In this project, we therefore develop a learning app that generates explanations for its users based on the current state of XAI research and investigate its effectiveness. Users are presented with a learning game in which they match Google Streetview images from four cities in Germany and Israel - Berlin, Hamburg, Tel Aviv, and Jerusalem - to the respective city. The AI attempts the same task. Users are shown how the AI arrived at its decision by highlighting the image components that are essential for the assignment ("feature relevance"). The app is implemented with a user-friendly interface and can be accessed low-threshold by users either via smartphone or via a web browser on a computer. At the end, users will receive a score for their answers by which they can compare their performance with that of the AI.

Cooperation partner: Prof. Dr. Lior Fink from the Behavioral Information Technology (BIT) Lab at Ben Gurion University in Beer Sheva, Israel.

Project period: September 2021 - September 2022