Work

What I Cannot Predict, I Do Not Understand: A Human-Centered Evaluation Framework for Explainability Methods

Year: 2021

Type: article

Abstract: = 1,150) to evaluate the usefulness of representative attribution methods in three real-world scenarios. Our results demonstrate that the degree to which individual attribution methods help human participants better understand an AI system varies widely across these scenarios. This suggests the need to move beyond quantitative improvements of current attribution methods, towards the development of complementary approaches that provide qualitatively different sources of information to human end-users.

Source: PubMed

DOI: 10.48550/arxiv.2112.04417

Authors: Colin, Julien, Fel, Thomas, Cadene, Remi, Serre, Thomas

Institutions: Brown University, ELLIS Alicante, Université de Toulouse, Société Nationale des Chemins de Fer Français (France), Centre National de la Recherche Scientifique, Sorbonne Université

Language: English

Topic: Explainable Artificial Intelligence (XAI)

Subfield: Artificial Intelligence

Field: Computer Science

Domain: Physical Sciences

Open Access status: green

Funders: National Science Foundation, Brown University, Agence Nationale de la Recherche, National Institutes of Health, Conselleria de Innovación, Universidades, Ciencia y Sociedad Digital, Generalitat Valenciana, Office of Naval Research, NIH Office of the Director

Awards: G2076488875, G3206771630, Brain Science Compute Cluster, CRCNS US-France Research Proposal: Oscillatory processes for visual reasoning in deep neural networks, Collaborative Research: Origins of Southeast Asian Rainforests from Paleobotany and Machine Learning, G8280078382, G8876996369

OpenAlex ID: W4200630041

FWCI: 2.972

Cites: 26

Cited by: 56

Related to: 10