Automated processes take over more and more tasks from humans. This is also the case in performance-related pay systems. The objective of this paper is to examine whether human performance is influenced by the usage of an automated performance appraisal system (automated PAS), and whether it matters if the company or the direct line manager decides to use an automated PAS. We present a modified Principal-Agent Game in a real job market using Amazon Mechanical Turk (Amazon MTurk). Depending on the treatment, either the principal or the system choose if an automated PAS or a manual PAS is used. We find that performance is significantly lower under an automated PAS system than under a manual PAS. However, performance does not differ significantly whether the principal decides compared to when the system decides to use an automated PAS or a manual PAS.
Presented at: 11th IMPRS Uncertainty Thesis Workshop