Let me decide: Increasing user autonomy increases recommendation acceptance

The use of algorithm assistance for decision making is vastly increasing in recent years, largely in online settings wherein users receive advice from recommendation systems. Despite the prevalence of such augmented decision making, users sometimes demonstrate aversion toward algorithms and even reject the recommendations of highly accurate systems. Research in behavioral decision making on the one hand, and in information systems on the other, has yet to fully comprehend these phenomena. We build on insight from these two research streams, as well as from self-determination theory, in identifying the psychological need for autonomy as a driver of decreasing algorithm aversion and increasing recommendation acceptance. We identify two concrete autonomy mechanisms—choice autonomy and control autonomy—which may increase the willingness of users to accept algorithmic recommendations. In three online experiments, simulating the process of selecting a vacation package, we consistently find that showing users multiple recommendations instead of a single recommendation, thereby increasing their choice autonomy, significantly increases recommendation acceptance, both when the system is perfectly accurate (Study 1) and when it is not (Studies 2 and 3). Furthermore, we find in Study 3 that allowing users control over whether they receive single or multiple recommendations, thereby increasing their control autonomy, has a positive moderating effect on the relationship between choice autonomy and recommendation acceptance, suggesting a complementarity between the two autonomy mechanisms. In so doing, we provide theoretical grounding and empirical evidence for the positive consequences of increasing user autonomy on the effectiveness of recommendation systems.