Exploring How Personalization and Source Expertise of Information from Healthcare Chatbot Affect Users’ Health Beliefs and Usage Intention

With the advance of the artificial intelligence (AI) and growing demands for virtual health assistance, healthcare chatbot services are emerging in the global market. In China, chatbots equipped with intelligent self-diagnosis can provide a solution to the underuse of the primary health care system. Several AI diagnosis apps have currently offered chatbot services (e.g., Left Hand Doctor, Medical Deer), which greatly contribute to remote triage, improve access to medical knowledge, and reduce healthcare costs, especially in the COVID-19 crisis. Although healthcare chatbots provide a reliable approach to search health information online (Codete, 2020), there is little research on exploring how Chinese users deal with health advice from chatbots and how users’ dual model processing of information works on their perceptions and attitudes. Applying the heuristic-systematic model (HSM) and the health belief model (HBM), this study is aimed to analyze the influence of users’ information processing mechanism on their health beliefs and continuous usage intentions in the context of healthcare chatbots in China.

HSM explains two modes in which persuasive messages can induce attitude change (Chaiken, 1980). This study assumes that personalization of health information is the systematic processing while heuristic processing is used for emotional appeals. Meanwhile, three variables were selected in the HBM (Rosenstock, 1974), namely perceived benefits, perceived barriers (privacy concerns), and self-efficacy, which are suitable for healthcare chatbot scenarios to measure subsequent perception changes. Health consciousness and perceived trust will also be included as they have been proven to play a moderating role (Guo et al., 2020) and a mediating role (Lee, 2017) in mobile health settings, respectively.

This study raises three questions regarding healthcare chatbots in China:

RQ1: How do Chinese users’ dual-mode processing affect their health beliefs?

RQ2: To what extent do users’ dual-mode processing and health consciousness affect their health
beliefs interactively?

RQ3: Do users’ health beliefs influence their intentions to use via perceived trust?

A between-subject experiment of 2 (personalization vs. non-personalization) × 2 (emotional appeals vs. non-emotional appeals) × 2 (health consciousness: high level vs. low level) will be conducted. Firstly, participants complete questions about health consciousness and demographic variables. Secondly, they are randomly assigned to one of four conditions. Real scenarios will be then developed to make participants directly interact with a chatbot via Sanuker (a chatbot provider). Thirdly, they are asked to look for answers about hemorrhoids in a standardized way. Finally, they will complete a questionnaire that includes the measurements of health beliefs, perceived trust, and usage intention. For data analysis, a series of MANOVAs will be conducted to explore the primary influence and interaction effects. The bootstrapping PROCESS macro will be used to examine the mediating role of perceived trust in relationships between health beliefs and usage intentions.

This study is expected to contribute to the extension of dual process theory through interaction with healthcare chatbots. Furthermore, this study may provide evidence for practitioners to highlight the significance of personalized recommendations and emotional appeals for changing uses’ attitudes.