Show simple item record

dc.contributor.authorWickstrøm, Kristoffer Knutsen
dc.contributor.authorHöhne, Marina Marie-Claire
dc.date.accessioned2024-02-16T13:42:15Z
dc.date.available2024-02-16T13:42:15Z
dc.date.issued2023
dc.description.abstractExplainable AI (XAI) is a rapidly evolving field that aims to improve transparency and trustworthiness of AI systems to humans. One of the unsolved challenges in XAI is estimating the performance of these explanation methods for neural networks, which has resulted in numerous competing metrics with little to no indication of which one is to be preferred. In this paper, to identify the most reliable evaluation method in a given explainability context, we propose MetaQuantus—a simple yet powerful framework that meta-evaluates two complementary performance characteristics of an evaluation method: its resilience to noise and reactivity to randomness. We demonstrate the effectiveness of our framework through a series of experiments, targeting various open questions in XAI, such as the selection of explanation methods and optimisation of hyperparameters of a given metric. We release our work under an open-source license1 to serve as a development tool for XAI researchers and Machine Learning (ML) practitioners to verify and benchmark newly constructed metrics (i.e., “estimators” of explanation quality). With this work, we provide clear and theoretically-grounded guidance for building reliable evaluation methods, thus facilitating standardisation and reproducibility in the field of XAI.en_US
dc.descriptionManuscript submitted to <a href=https://jmlr.org/tmlr/index.html>https://jmlr.org/tmlr/index.html</a>.en_US
dc.identifier.citationWickstrøm KK, Höhne MM. The Meta-Evaluation Problem in Explainable AI: Identifying Reliable Estimators with MetaQuantus. Transactions on Machine Learning Research (TMLR). 2023en_US
dc.identifier.cristinIDFRIDAID 2184130
dc.identifier.issn2835-8856
dc.identifier.urihttps://hdl.handle.net/10037/32956
dc.language.isoengen_US
dc.relation.journalTransactions on Machine Learning Research (TMLR)
dc.relation.projectIDNorges forskningsråd: 303514en_US
dc.relation.projectIDNorges forskningsråd: 315029en_US
dc.relation.projectIDNorges forskningsråd: 309439en_US
dc.rights.accessRightsopenAccessen_US
dc.rights.holderCopyright 2023 The Author(s)en_US
dc.titleThe Meta-Evaluation Problem in Explainable AI: Identifying Reliable Estimators with MetaQuantusen_US
dc.type.versionsubmittedVersionen_US
dc.typeJournal articleen_US
dc.typeTidsskriftartikkelen_US


File(s) in this item

Thumbnail

This item appears in the following collection(s)

Show simple item record