An Agent by Any Other Name: A Exploration of the Influence of Mental Models, Priming, and Perception on Users’ Views of Chatbot Character and Evaluation of Agent Effectiveness

Abstract

This thesis endeavors to delineate the evolving interface between human cognition and artificial intelligence (AI), aiming to contribute to the corpus of knowledge within this rapidly advancing domain. It investigates the ability of psychological priming and the placebo effect to alter one’s mental models to enhance user interactions with AI technologies. Through a randomized control experiment, surveys initially assessed users’ attitudes towards AI and subsequently introduced supraliminal priming via descriptions of a purportedly unique AI mental wellness chatbot. The findings establish that priming users measurably impacts their experience with AI. A micro-diverse set of priming objectives (Avery vs. Charlie vs. Jamie vs. Bot characters) yield observable differences in response. Character priming was found to improve positive participant experiences relative to the neutral Wellness-Bot priming, to a similar degree as initial attitudes towards AI, which also enhanced the overall experience. This research underlines the importance of personalized character creation in the design and deployment of digital wellness interventions, as well as the essential role of fostering positive AI attitudes for successful integration of chatbots in user-centric applications. By investigating the overlap of cognitive science and AI research, this work provides foundational insights for fostering more intuitive and rewarding human-AI relationships, especially in enhancing mental wellness through digital interventions.Acknowledgments