By definition, algorithmic decision-making (ADM) transfers human agency (i.e., decisional control) to the machine in exchange for optimal outcomes. Whereas algorithmic aversion literature suggests a default hesitance to use machine decisions, these conclusions have recently been disputed. This study examines the conditions under which individuals choose to delegate decisions to algorithms versus human agents, focusing on how trade-offs between agency and ADM benefits influence comparative evaluations. In three experiments (Ntotal = 841) in financial and romantic contexts, participants made a joint evaluation of algorithmic and human agents manipulated to offer varying levels of decisional agency and benefits. Contrary to prior research, we find limited evidence for algorithmic aversion as a default phenomenon. Instead, 1) both agency and benefits play critical roles in shaping ADM acceptance, and 2) individuals weigh these factors differently depending on the context. The findings highlight the complex context-dependent trade-offs between control and benefits shaping delegating preferences, and call for a more nuanced understanding beyond generalized algorithmic aversion.
Schaap et al. (2025) studied this question.