Ethical Implications of Implicit Bias in AI: Impact for Academic Libraries

Abstract

Academic libraries are exploring artificial intelligence (AI) applications that have the potential to create new or improved user experiences, streamline ways of working, and deliver new insights to their activities. Nevertheless, it is now clear that AI applications are not neutral technological solutions. They can embed and magnify prejudices and stereotypes, and they can perpetuate errors and limitations in training and accumulated datasets. At the same time, academic libraries abide by ethical considerations of social responsibility. If datasets and algorithmic black boxes in AI systems replicate or aggravate inappropriate discrimination in their use of information, or if they simply lack or ignore data, they can produce distorted outcomes. The ethical implications for academic libraries and end-users can be profound. This chapter examines these issues, illustrates problematic outcomes, and identifies both the need for caution and some paths to the ethical use of AI applications in academic libraries. After a general exploration of the essence of machine learning (ML), this chapter explains what implicit bias is, how it enters ML applications, and why the problem is insidious and challenging. The authors present an illustrative review of the ethical foundations of the work of academic libraries and draw analogies to other professional interfaces with AI and implicit bias. Possible scenarios of ethically problematic outcomes in academic libraries are explored.