Actionable Approaches to Promote Ethical AI in Libraries

TL;DR

This paper reviews approaches to promote ethical AI, emphasizing their adaptation to library information systems, addressing the lack of practical guidance for ethical AI deployment in libraries.

Contribution

It introduces adaptable approaches for ethical AI in libraries, filling a gap in practical guidance for planning, evaluating, and auditing AI ethics in library services.

Findings

Several promising approaches for ethical AI promotion.

Adaptation of methods from other contexts to library AI.

Addresses stages of AI lifecycle in libraries.

Abstract

The widespread use of artificial intelligence (AI) in many domains has revealed numerous ethical issues from data and design to deployment. In response, countless broad principles and guidelines for ethical AI have been published, and following those, specific approaches have been proposed for how to encourage ethical outcomes of AI. Meanwhile, library and information services too are seeing an increase in the use of AI-powered and machine learning-powered information systems, but no practical guidance currently exists for libraries to plan for, evaluate, or audit the ethics of intended or deployed AI. We therefore report on several promising approaches for promoting ethical AI that can be adapted from other contexts to AI-powered information services and in different stages of the software lifecycle.