The Algo Taming Effect: An Experimental Study of Human vs Algorithmic Dismissals
Brice Corgnet (EM Lyon)
Abstract: Using two laboratory studies that employ a combination of performance metrics and questionnaires, we compare workers’ reactions to human and algorithmic dismissals that automatically demote the least productive workers. We find that algorithmic dismissals can mitigate workers’ negative reactions to dismissals, a phenomenon we refer to as the algo taming effect. We demonstrate that this effect results from algorithmic decisions moderating perceptions of distributive justice among dismissed workers and reducing the negative behavioral reaction of spiteful workers to dismissals. We also find that human managers who are given the option to delegate dismissal decisions to algorithms only do so in one-third of the cases, thus exacerbating workers’ negative reactions. Our findings indicate that using algorithmic dismissals can help reduce workers’ negative reactions. Yet, implementing algorithmic dismissals may face resistance, not necessarily from workers, but from human managers.