LLM Personality Expression Affects User Perceptions in Goal-Oriented Tasks

A recent study published on arXiv (2509.09870) investigates the impact of personality expression in LLM-based conversational agents on user perceptions during goal-oriented activities. In this between-subjects experiment, 150 participants engaged in travel planning, where agents displayed varying levels of the Big Five traits—low, medium, or high—using the Trait Modulation Keys framework. Findings revealed an inverted-U relationship, with medium expression leading to the best ratings in areas such as Intelligence, Enjoyment, Anthropomorphism, Intention to Adopt, Trust, and Likeability. Additionally, a better personality match between user and agent enhanced results, particularly with Extraversion and Emotional Stability being the most significant factors. Cluster analysis revealed three unique user groups.