Limits of XAI application-grounded evaluation: an e-sport prediction example - Laboratoire des Sciences et Techniques de l'Information, de la Communication et de la Connaissance, site ENIB Brest Accéder directement au contenu
Communication Dans Un Congrès Année : 2022

Limits of XAI application-grounded evaluation: an e-sport prediction example

Résumé

EXplainable AI (XAI) was created to address the issue of Machine Learning's lack of transparency. Its methods are expanding, as are the ways of evaluating them, including human performance-based evaluations of explanations. These evaluations allow us to quantify the contribution of XAI algorithms to human decision-making. This work performs accuracy and response time measurements to evaluate SHAP explanations on an e-sports prediction task. The results of this pilot experiment contradict our intuitions about the beneficial potential of these explanations and allow us to discuss the difficulties of this evaluation methodology.

Mots clés

XAI
Fichier principal
Vignette du fichier
Limits_of_XAI_task_performance_evaluation_ECML_PKDD (3).pdf (516.05 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-03847499 , version 1 (10-11-2022)
hal-03847499 , version 2 (18-09-2023)

Identifiants

  • HAL Id : hal-03847499 , version 1

Citer

Corentin Boidot, Olivier Augereau, Pierre de Loor, Riwal Lefort. Limits of XAI application-grounded evaluation: an e-sport prediction example. XKDD 2022: 4th International Workshop on eXplainable Knowledge Discovery in Data Mining, Sep 2022, Grenoble, France. ⟨hal-03847499v1⟩
50 Consultations
119 Téléchargements

Partager

Gmail Facebook X LinkedIn More