User Perception of Text-Based Chatbot Personality.

This work explores the effect of chatbot personality on user experience and investigates how users perceive agent personality when conveyed through text. Building on previous work in the field of human-computer interaction on designing chatbot personality, we investigate whether users in a low-stakes conversation have a preference for a specific personality type when the agent does not use voice, is not visually represented, and does not provide identity cues such as gender. We developed two chatbots that interact with users in a multi-turn conversation and designed them to have distinct personalities along two axes of the Five Factor Model (extraversion and agreeableness). We conducted a user study to evaluate user engagement, user perception of the agents, and the effect of user personality on user experience.