Skip to main content

TECHNOLOGY AND CODE article

Front. Comput. Sci.
Sec. Human-Media Interaction
Volume 6 - 2024 | doi: 10.3389/fcomp.2024.1412341
This article is part of the Research Topic Human-Centered Artificial Intelligence in Interaction Processes View all 5 articles

How informative is your XAI? Assessing the Quality of Explanations through Information Power

Provisionally accepted
Marco Matarese Marco Matarese 1,2*Francesco Rea Francesco Rea 1Katharina J. Rohlfing Katharina J. Rohlfing 3Alessandra Sciutti Alessandra Sciutti 1
  • 1 Italian Institute of Technology (IIT), Genova, Italy
  • 2 University of Genoa, Genoa, Liguria, Italy
  • 3 University of Paderborn, Paderborn, North Rhine-Westphalia, Germany

The final, formatted version of the article will be published soon.

    A growing consensus emphasizes the efficacy of user-centered and personalized approaches within the field of explainable artificial intelligence (XAI). The proliferation of diverse explanation strategies in recent years promises to improve the interaction between humans and explainable agents. This poses the challenge of assessing the goodness and efficacy of the proposed explanation, which so far has primarily relied on indirect measures, such as the user's task performance. We introduce an assessment task designed to objectively and quantitatively measure the goodness of XAI systems, specifically in terms of their "information power". This metric aims to evaluate the amount of information the system provides to non-expert users during the interaction. This work has a three-fold objective: to propose the Information Power assessment task, provide a comparison between our proposal and other XAI goodness measures with respect to eight characteristics, and provide detailed instructions to implement it based on researchers' needs.

    Keywords: XAI, goodness, assessment, Human-centered, Objective, quantitative

    Received: 30 Oct 2024; Accepted: 18 Dec 2024.

    Copyright: © 2024 Matarese, Rea, Rohlfing and Sciutti. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

    * Correspondence: Marco Matarese, Italian Institute of Technology (IIT), Genova, Italy

    Disclaimer: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.