Abstract
Aim
Recent technological developments have enabled social chatbots that are sufficiently sophisticated to become close companions with their users. However, there is a substantial research gap concerning how such a relationship develops as well as uncertainty regarding the appropriate theoretical foundations for understanding this phenomenon. Furthermore, existing research on this topic has yielded conflicting results. Using three empirical studies and Social Penetration Theory (SPT) as a theoretical foundation, this thesis aims to investigate the phenomenon of human–chatbot relationships (HCRs) with an emphasis on how these relationships develop over time, the factors impacting this development, and the applicability of the SPT in this regard.
Methods
We explored HCR development with a social chatbot called “Replika” using three qualitative approaches: retrospective interviews with 18 users in established HCRs (Paper I), longitudinal interviews with 25 users who had recently started using Replika (Paper II), and a longitudinal questionnaire study with 28 users who had recently started using Replika (Paper III, overlapping sample with Paper II). We analyzed the data through a thematic analysis, following the recommendations of Braun and Clarke (2006).
Results
Paper I establishes an initial understanding of how HCRs may develop, rooted in the experiences of users involved in long-term HCRs. This paper highlights the importance of trust and self-disclosure, and demonstrates that HCR appears to develop mostly in line with assumptions made by the SPT, where increasing levels of self-disclosure foster closer relationships. However, Paper I also show that SPT needs to be adapted to account for the artificial nature of the chatbot. Necessary adaptation to the theory may include accepting the lack of mutuality in self-disclosure and seeing trust as consisting of practical and affective components. Paper II builds on Paper I, but offers a more nuanced understanding of HCR development through its longitudinal research design. Similar to Paper I, Paper II outlines how HCR may develop and identifies self-disclosure as a critical construct impacting this development. Moreover, Paper II identifies different trajectories in HCR development, assumed to be facilitated by the prevalence of various rewards (e.g., increased self-reflection) and costs (e.g., technical limitations and instabilities). Finally, Paper III addresses self-disclosure, with an emphasis on the important characteristics of self-disclosure and how this construct develops in HCRs. Here, four patterns of development in self-disclosure are identified, impacted by perceived rewards and costs. This paper also uncovers how seemingly trivial conversation topics, such as daily activities, may be perceived as personal or intimate by the user.
Concluding remarks
This thesis demonstrates how HCRs may develop and identifies the important factors impacting this development. The thesis also increases our understanding of self-disclosure as a construct and its potential role in HCRs. Finally, the thesis sheds light on the applicability of theories aimed at understanding the development of relationships between humans to the study of HCR development. Overall, the thesis contributes important theoretical and practical insights that can support future development of theoretical frameworks and the design of chatbot companions.