Does Loneliness Affect Mental Health Chatbot Experiences? The function of social motivation on anthropomorphism of and information sharing with virtual agents.

Abstract

Mental health chatbots are virtual agents often used by patients of mental disorders. Since loneliness often accompanies mental disorder, we aim to explore the effects of loneliness on two attributes that shape virtual agent functionality. Mental health chatbots are designed to replicate human interactions; therefore, anthropomorphism, the attribution of human-like traits to non-humans, is essential to enabling such functionality. Further, information sharing is a social behavior that allows chatbots to gather information from users. The role of both anthropomorphism and information sharing brings us to explore the potential effects of different social motivations on the efficacy of mental health chatbots. We hypothesize that lonely people are more likely to anthropomorphize and exhibit information sharing behaviors towards virtual agents than socially satisfied people and those that interact with a human.

To test this, participants were induced to either lonely or socially satisfied motivation states prior to engaging in a conversational task with what they perceived to be either a “virtual agent” or a “human.” Ultimately, we found three items of importance: (1) Anthropomorphism and information sharing hold no relationship in the context of our virtual agent. (2) Loneliness does not affect anthropomorphism of virtual agents, and (3) does not affect information sharing with virtual agents. These findings indicate that such attributes, which are important to the functionality of mental health chatbots, are not impacted by social motivation. These virtual agents can function with comparable success across users of differing social motivation states based on anthropomorphism and information sharing.