New Publication in the Journal of Behavioral Decision Making

Professur Lars Hornuf published a new article titled “Similarity and Consistency in Algorithm-Guided Exploration” together with Ludwig Danwitz, Sebastian Fehrler, Hsuan-Yu Lin, Yongping Bao, Fabian Dvorak, and Bettina von Helversen in the Journal of Behavioral Decision Making. The study investigates when people are willing to follow algorithmic advice in complex tasks that require balancing exploration and exploitation.In an online experiment, participants received advice from reinforcement learning algorithms designed to prioritize either exploration or exploitation. Surprisingly, participants did not prefer advice that matched their own exploration tendencies. Even though the advice from exploratory algorithms led to better outcomes for the participants, they were more likely to follow consistent, exploitative algorithms, seemingly interpreting consistency as a sign of competence.

The findings highlight a key challenge for human-algorithm collaboration: effective but exploratory algorithms may fail to promote behavioral diversification if users are reluctant to trust and follow their advice.

The full article is available here.