Just Because We Can, Doesn’t Mean We Should: Algorithm Aversion as a Principled Resistance

Abstract

This paper problematizes the literature on the non-use of algorithmic decision-making systems (ADMS), commonly examined as algorithm aversion. Whilst prior literature attributes algorithm aversion primarily to human bias and irrationality, assuming utility-based evaluations, we argue that it may also stem from values-based evaluations of technology, which are overlooked. Through an integrated ethical analysis, drawing upon the “big three” ethical theories of consequentialism, deontology, and virtue ethics, we examine implicit normative judgments within the algorithm aversion literature. Consequently, we positively reframe algorithm aversion as a potentially principled resistance to ADMS, expanding prior views of the phenomenon. We argue that such resistance may be constructive and lead to a better alignment of ADMS with societal needs and values. Thus, we call on IS scholars to explore this phenomenon as an ethical and sociotechnical issue, rather than as a costly problem to be mitigated, as prior literature might suggest.