Let the user have a say - voice in automated decision-making

Abstract

Automated decision-making is on the rise and systems now make decisions traditionally made by humans. As initial evidence suggests that automation of decisions can be associated with low fairness perceptions, features for the design of fair systems are needed. One such feature could be voice. Voice means having a say in decision processes and is associated with fair processes. In the present experiment, we tested whether perceptions of the opportunity to voice one's views to a system resulted in higher perceived overall fairness of the system. Drawing on fairness heuristic theory, we furthermore investigated, whether perceived overall fairness of the system related to trust in the system, job satisfaction, and task performance. In the experiment, participants (N = 138) performed a monitoring task on temperature scales. The system decided how many scales participants had to monitor in a final work round. We varied voice (participants could or could not voice their opinion on how many scales they wanted to monitor). The results indicate that voice leads to fairness perceptions. Overall fairness mediates the relationship between voice and job satisfaction and voice and trust. Therefore, voice is an important design feature for fair automated decision-making.