Abstract:

With the rapid development of artificial intelligence (AI) technology, AI chatbots have shown great potential in simulating human guidance to improve the engagement and efficacy of Internet-based self-help interventions (ISIs). Given that studies on the mechanisms of chatbots are still in the early stages, there is a need to conduct research that can help deepen our rational understanding of designing more targeted chatbots and further promoting the effectiveness of ISIs. Recently, researchers have focused on human-chatbot relationships (HCRs) and have attempted to explain the mechanisms of chatbot effectiveness from this perspective. As HCRs share some similarities with human-human relationships (HHRs), some HHR theories, such as social penetration theory, may be used to explain HCRs in ISIs. However, relying solely on HHR theories to explain HCRs in ISIs has some limitations, such as overlooking the early cognitive processing of human-computer interactions (HCIs) and ignoring the core goals of psychotherapy in ISIs. In response to these limitations, we propose a theoretical model that is particularly suitable for the ISI context from the perspective of HCI. Our model suggests that chatbots can gradually develop relationships with users through four stages: anthropomorphic attribution, utilitarian value judgment, attachment relationship development, and the establishment of the digital therapeutic alliance (DTA). These can ultimately promote user engagement through HCRs. First, when users interact with chatbots, they unconsciously treat them as if they were actual persons due to their human-like qualities or conversational ability, resulting in anthropomorphic attribution. As a result, users tend to interact with chatbots through interpersonal communication strategies. The effect of anthropomorphism will be amplified as the frequency of interaction increases, thereby promoting the development of HCRs. Consequently, we propose that anthropomorphic attribution be the primary starting point for developing HCRs. Second, users tend to judge the utilitarian value of chatbots in the early stages of HCR development to determine whether their actual needs can be met. Therefore, whether or not chatbots can demonstrate their true effectiveness based on user expectations in terms of usability, ease of use, and expectation confirmation is likely to influence user acceptance and engagement with them. With the growth in users’ recognition, familiarity, and confidence in chatbots, their notion of anthropomorphism improves, indirectly enhancing user attitudes and further developing HCRs. Third, user participation in ISIs is influenced by anthropomorphic attribution and utilitarian value judgment in the short term. Given that emotional factors are becoming increasingly important in sustaining user engagement over time, users are likely to further anthropomorphize chatbots based on social motivation and establish an attachment relationship with them. The positive emotions experienced when interacting with chatbots deepen the HCRs, shifting relationships from “tools” to “partners.” When users’ attachment to chatbots is transferred to their attachment to ISI tasks, they are more likely to actively engage in the treatment, allowing for the effective use of chatbots in psychotherapy. Finally, to achieve better therapeutic goals of ISIs, HCRs should be further developed into DTA—a deliberate and purposeful HCR model. Additionally, the stage-wise development of HCRs has laid a good foundation for establishing DTA. As the development of DTA can enhance and stabilize user engagement, future research can make valuable contributions by evaluating key HCR theories and constructing chatbots based on these theories. Apart from the topics stated above, there is a need to examine additional variables that affect HCRs, standardize operational definitions of engagement, and develop appropriate methods for measuring engagement. Ultimately, by developing our theoretical model, we contribute to the improvement of chatbot effectiveness in the field of psychotherapy through the promotion of a deeper rational understanding of HCRs.

Key words: chatbot, engagement, human-chatbot relationships