UC Davis

Chatbot use is becoming increasingly prominent in mental health management, but research focusing on the mechanisms through which chatbots can successfully provide informal social support to users is still in its infancy. This dissertation addresses this gap in the literature by focusing on the supportive capabilities of a chatbot as an explanation for the influence of the chatbot-human interaction on mental well-being for college students by highlighting the importance of two key communication processes: perceptions of the interaction partner and self-disclosure. Study one examined if and how the perceived identity (human versus chatbot) and perceived expertise (expert versus non-expert) of a chatbot support provider interact to influence four support outcomes: support quality, perceived coping ability, state wellbeing, and support seeking intentions. Additionally, study one tested the mediating role of perceptions of the support provider in terms of empathy, similarity to the support seeker, and trustworthiness. College students participated in an online survey experiment where they had a conversation with a chatbot that was framed as an expert human, non-expert human, expert chatbot, or non-expert chatbot. After completing the interaction, participants answered questions regarding their perceptions of their support provider and the four support outcomes. Results showed that perceive human versus chatbot identity drove perceptions of similarity and trust, with those assigned to the chatbot conditions reporting lower levels of state wellbeing compared to the perceived human/non-expert condition can be attributed to the negative influence of the chatbot conditions on perceived similarity. Additionally, lower levels of perceived support quality and more negative support seeking intentions in the chatbot/non-expert condition compared to the perceived human/non-expert condition can be attributed to the negative influence of the chatbot condition on perceptions of trustworthiness. Study two involved manipulating the type of chatbot’s self-disclosure—describing having helped others with the same stressor or having dealt with the same stressor themselves—in addition to perceived support provider identity (chatbot versus human). The study examined if the interaction of perceived provider identity and self-disclosure type influenced participant self-disclosure during the interaction and reactions to the received support, however, results showed there was no significant associations. While participant self-disclosure was not found to drive support outcomes in study two, similarity and trust were identified as key mediators in study one, suggesting that certain provider characteristics are more salient and influential in human-chatbot support.