Beyond efficiency: Trust, AI, and surprise in knowledge work environments

Abstract

Contemporary managemenet practices are often designed with the needs of knowledge-based workers in mind, but an increasingly pressing challenge today is how to manage and effectively handle non-routine work. This paper revisits the job characteristics model through the lens of self-determination theory, specifically in the context of algorithmic performance management. Non-routine work is inherently unpredictable, and individuals often struggle with prolonged uncertainty. However, automated interventions that help individuals make sense of their work in uncertain conditions may help overcome the challenges of non-routine work and increase worker performance. In a randomized, controlled experiment delivered in a novel online task environment, we find that automated, real-time feedback increases the perceived trustworthiness of an algorithmic performance rating under conditions of high task uncertainty. Our research demonstrates the potential of artificial intelligence to automate certain tasks in non-routine work environments that positively augment human work performance while simultaneously enhancing trust in these automated work systems.