The level of strength of an explanation: A quantitative evaluation technique for post-hoc XAI methods

Marilyn Bello, Rosalís Amador, María-Matilde García, Javier Del Ser, Pablo Mesejo, Oscar Cordón

Abstract: Highlights•The evaluation of a post-hoc XAI approach becomes a multi-criteria decision problem.•This paper proposes a new quality estimator: the LSE measure.•Our approach evaluates another category of evaluation, its Strength.•The LSE quantifies the extent to which the explanation supports neural inference.