Al ethics auditingdrivers, strategies, and challenges

Resumen

The goal of this study is to assess practices for evaluating the ethical development of artificial intelligence (AI) systems according to the perceptions of AI ethics 'auditors.’ In assessing the adoption of AI ethics auditing strategies, the study begins by determining the range of strategies by auditors used across organizations. Subsequently, the rationales driving the adoption of a given AI ethics auditing strategy amongst a range of choices by an external AI ethics auditor and their client organizations is explored. The study then establishes what factors drive the effectiveness of AI ethics audits, and explores the perceived benefits and challenges associated with each strategy. The study concludes with a discussion of the possible sources of variation between the auditing strategies available from external AI ethics auditors, and what this implies for the state of AI ethics auditing considering current ethical, organizational, and regulatory challenges. The study is being developed as a joint effort between four researchers from Georgia Tech (USA), Queens’ university (Canada), University of Southampton (UK) and Universidad Pontificia Comillas (Spain). The academic benefits of this study are understanding the different AI ethics auditing strategies available to organizations; the factors that drive the adoption of a given AI ethics auditing strategy by an auditor and by an organization; the perceived benefits and challenges associated with each AI ethics auditing strategy; and the factors that might drive the effectiveness of AI ethics audits. The findings from the study could help practitioners using AI to select appropriate AI ethics auditing strategies for more effective prevention of AI ethics issues. It could also inform AI ethics auditors of alternative auditing strategies, and the accompanying benefits, challenges, and factors that lead to more responsible adoption and implementation of AI systems. The study involves semi-structured interview with 40-60 participants who are employees or associates of AI ethics audit firms and/or audited firms and participate/have participated directly in the AI ethics audit team. Participants are employed in companies operating out of Canada, the United States, Singapore, the United Kingdom, and Spain.