AI Controlling AI? A Potentially Dystopian View of Automatic Systems

In this short paper, after reviewing the IS literature concerning AI, we call for a careful reconsideration of how these automatic systems are theorized, as this holds profound implications for research and practice. As an example, we comment a recent paper on AI published in MIS Quarterly (Kane et al. 2021). The authors outline the possibility that the problematic societal implications associated with ML systems can be addressed by ad-hoc designed ML. While we applaud Kane et al.' effort to address these issues (our community lacks this focus), we outline several concerns and conclude that Kane et al.'s way to address a dystopian future dominated by ML using an adversary system is a dystopian idea by itself. We make our case by questioning a design science approach to address ML issues. We advocate for a practice-based approach to make sense of the societal ramification of ML use in everyday practices.