How the Introduction of Machine Traders and Disclosure of Their Presence Affect Prediction Accuracy: An Online Controlled Experiment in Prediction Market

Abstract

ABSTRACT Incorporating prediction models developed based on machine learning algorithms into the traditional prediction market creates hybrid intelligence. We design and conduct an online controlled experiment to investigate the impacts of two dimensions of human–machine interaction, whether to introduce machines as traders and whether to disclose their presence, on the prediction performance. The results of the experiment reveal that the introduction of machines creates two competing effects on prediction accuracy. The positive influence comes from the intensified competition brought by machines, which fosters a strong desire to win among human participants and motivates them to engage in more deliberate decision-making efforts. Conversely, in the context of intensive competition, humans are inclined to trade at a large magnitude, consequently leading to a decrease in prediction performance. Furthermore, the results indicate that simply disclosing the presence of machines can have a detrimental impact on prediction performance, as it may lead to a reduction in human deliberation efforts. Furthermore, this article delves into the potential mechanisms involved. This study contributes to the understanding of human behaviors in hybrid prediction markets and highlights the need for careful human–machine interaction design to optimize prediction market performance