The medical algorithmic audit

Xiaoxuan Liu, Ben Glocker, Melissa M McCradden, Marzyeh Ghassemi, Alastair K Denniston, Lauren Oakden-Rayner

Research output: Contribution to journalReview articlepeer-review

86 Downloads (Pure)

Abstract

Artificial intelligence systems for health care, like any other medical device, have the potential to fail. However, specific qualities of artificial intelligence systems, such as the tendency to learn spurious correlates in training data, poor generalisability to new deployment settings, and a paucity of reliable explainability mechanisms, mean they can yield unpredictable errors that might be entirely missed without proactive investigation. We propose a medical algorithmic audit framework that guides the auditor through a process of considering potential algorithmic errors in the context of a clinical task, mapping the components that might contribute to the occurrence of errors, and anticipating their potential consequences. We suggest several approaches for testing algorithmic errors, including exploratory error analysis, subgroup testing, and adversarial testing, and provide examples from our own work and previous studies. The medical algorithmic audit is a tool that can be used to better understand the weaknesses of an artificial intelligence system and put in place mechanisms to mitigate their impact. We propose that safety monitoring and medical algorithmic auditing should be a joint responsibility between users and developers, and encourage the use of feedback mechanisms between these groups to promote learning and maintain safe deployment of artificial intelligence systems.

Original languageEnglish
Pages (from-to)e384-e397
JournalThe Lancet Digital Health
Volume4
Issue number5
Early online date5 Apr 2022
DOIs
Publication statusPublished - 1 May 2022

Bibliographical note

Funding Information:
We acknowledge Inioluwa Deborah Raji and her co-authors for their seminal work on internal algorithmic auditing, 19 upon which we based our medical algorithmic audit. We are grateful for her comments and feedback on our adaptation of this framework. XL and AKD receive a proportion of their funding from the Wellcome Trust, through a Health Improvement Challenge grant (200141/Z/15/Z). BG receives funding from the European Research Council under the EU Horizon 2020 research and innovation programme (grant agreement number 757173, project MIRA, ERC-2017-STG).

Publisher Copyright:
© 2022 The Author(s). Published by Elsevier Ltd. This is an Open Access article under the CC BY 4.0 license

Fingerprint

Dive into the research topics of 'The medical algorithmic audit'. Together they form a unique fingerprint.

Cite this