Abstract
This study aims to improve trust to performative algorithms (PA) by providing social incentives (SI) to users of PA. Currently, algorithms outperform humans in many areas. However, even though algorithms are proven to be superior, they have been shown to prefer human assistance over algorithmic assistance. Algorithms include PA, which can take independent action without human intervention by gathering information, making and executing decisions on their own, and advisory algorithms (AA), which provide support for human decisions. With PA, the user is forced to make a decision between themself and the algorithms. With AA, on the other hand, the final decision is left to the human. Compared to PA, AA has the disadvantage of being time-consuming, although users are less averse to algorithms. Therefore, we conducted a survey experiment focusing on changing trust in PAs using SI. This study reveals that SI is not an effective means of improving trust in PA, and provides implications for the use of SI in settings where algorithms aversion is expected to occur.
