Responsible Artificial Intelligence in Knowledge Work: User Experience Design Problems and Implications.

With ongoing technological advances, artificial intelligence (AI) tools are becoming increasingly capable of handling not only repetitive tasks, but also so called "white collar" knowledge work involving abstract cognition and communication to accomplish complex tasks. However, AI's limitations regarding adapting and processing exception events, finding creative solutions, and making ethical and moral evaluations render the involvement of human users as AI overseers or ultimate decision-makers necessary. The effectiveness of such a human-in-the-loop approach will depend on a multitude of factors, including those related to user interactions afforded by the design of the AI. This paper reports the current state of a work-in-progress research project aiming to identify key design considerations and their implications to keep humans in the loop when AI handles a complex, knowledge-intensive task. Out of an initial pool of 3,698 peer-reviewed articles, 720 were screened-in for relevance. From a subset of 18 articles used in this paper, preliminary findings, implications for theory and practice, and recommendations for future research are discussed.