On the Interaction of Belief Bias and Explanations

Publikation: Bidrag til bog/antologi/rapportKonferencebidrag i proceedingsForskningfagfællebedømt

A myriad of explainability methods have been proposed in recent years, but there is little consensus on how to evaluate them. While automatic metrics allow for quick enchmarking,
it isn’t clear how such metrics reflect human interaction with explanations. Human evaluation is of paramount importance, but previous protocols fail to account for belief biases affecting human performance, which may lead to misleading conclusions. We provide an overview of belief bias, its role in human evaluation, and ideas for NLP practitioners on how to account for it. For t o experimental paradigms, we present a case study of gradientbased explainability ntroducing simple ways to account for humans’ prior beliefs: models of varying quality and adversarial examples. We show that conclusions about the highest performing methods change when introducing such controls, pointing to the importance of accounting for belief bias in evaluation.
1 Int
OriginalsprogEngelsk
TitelFindings of the Association for Computational Linguistics: ACL-IJCNLP 2021
Antal sider13
UdgivelsesstedOnline
ForlagAssociation for Computational Linguistics
Publikationsdato1 aug. 2021
Sider2930-2942
DOI
StatusUdgivet - 1 aug. 2021
BegivenhedJoint Conference of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing, ACL-IJCNLP 2021 - Virtual, Online
Varighed: 1 aug. 20216 aug. 2021

Konference

KonferenceJoint Conference of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing, ACL-IJCNLP 2021
ByVirtual, Online
Periode01/08/202106/08/2021
SponsorAmazon Science, Apple, Bloomberg Engineering, et al., Facebook AI, Google Research

ID: 285387796