Beyond Efficiency: Trust, AI, and Surprise in Knowledge Work Environments.

 One of the major challenges in contemporary work and job design lies in identifying effective strategies for leveraging the affordances of artificial intelligence (AI) and machine learning technologies to jointly improve outcomes for workers and organizations. The demands of knowledge work and other non-routine task environments make productivity difficult to measure and assess. Moreover, the technological interventions deemed effective in more predictable work contexts typically undermine the key levers for supporting and motivating knowledge workers. Nonetheless, algorithmically-managed contexts are increasingly leveraging these new technologies to direct behavior and evaluate workers. In this study we conduct an online experiment examining the mechanism by which automated feedback might improve individual perceptions about the trustworthiness of algorithmic evaluations of their work. Our results show that automated feedback leads to an increase in the perceived trustworthiness of an algorithmically-determined performance score through improved knowledge of results that reduce evaluation-based surprise under conditions of higher task uncertainty. This work contributes to the literature on algorithmic management and the calibration of trust in AI