70

Recent work on the acceptability of Artificial Intelligence (AI)-based professional tools has focused on the consideration of human factors such as user confidence as well as technical factors like model explainability as determinants of user adoption and usage. These links are mainly established in simulated contexts or environments with low deployment constraints. However, the acceptability of AI tools in professional settings often seems to be restricted to the acceptance of performance and short-term return on investment (ROI), without any questions being raised about the technical suitability of AI capabilities to user’s needs (and fears) or the tasks at hand. Three deployment sites of AI-assisted decision-making tools have been observed where AI acceptability was limited to the artifact’s measurable time-based efficiency. In this article, we examine the other User experience (UX) design considerations in these machine learning (ML)-enhanced dashboards with rushed production cycles that produced AI gadgets rather than user-friendly tools. These AI gadgets ultimately resulted in negative user experiences, inconclusive ROI, and/or sidelining of projects. We then suggest ways of limiting risks of end-user rejection through a more holistic approach to UX design considerations. These include ascertaining the needs of the stakeholder network, avoiding technological solutionism, optimizing user interfaces to reduce unnecessary complexity, and scaling technical solutions appropriately to fit the task in hand.