Developing engaging digital humans for psychotherapeutic applications

Abstract

Digital humans offer a promising virtual platform for the remote delivery of psychological interventions. Digital humans are scalable, yet capable of forming relationships with patients to promote engagement. However, little research is available to indicate what relationship building techniques are effective for digital humans. In addition, digital humans are a new technology and research has yet to evaluate their use in delivering psychological interventions. This thesis aimed to investigate: (1) techniques for building relationships with digital humans in psychotherapeutic applications; and (2) the feasibility and acceptability of digital humans at delivering psychological interventions. To identify relationship building techniques for digital humans, an opinion piece, a systematic review, and a large experimental study (presented in three manuscripts) were conducted. The opinion piece proposed that an evolutionary neuropsychiatry theory, the Mammalian Behavioural Triad, could inform the behavioural design of artificial agents (e.g., robots, computer agents) in psychotherapeutic applications. The systematic review synthesized the existing literature on design features that have been shown to improve relationships with embodied conversational agents, and proposed a new theoretical framework to inform experimental research. The experimental study found that emotional expressiveness in a female digital human was an effective closeness building technique, however only in female users. A digital human with emotional expressiveness was associated with improved psychological and physiological outcomes in females after a self-disclosure conversation. Conversely, males experienced improved closeness, psychological, and physiological outcomes with a neutral digital human. A qualitative analysis found that a range of design features contributed to participantsð€‚¶ perceptions of closeness and willingness to seek emotional support with a digital human. Findings suggest that behavioural, language, emotive, and appearance-based techniques can be used by digital humans to build relationships in psychotherapeutic contexts. To evaluate the feasibility and acceptability of digital humans at delivering psychological interventions, two randomised pilot trials were conducted. The digital humans delivered relationship building techniques alongside evidence-based psychological interventions. The first pilot trial found that digital humans were a feasible and acceptable way to deliver a remote loneliness and stress intervention to older adults living independently and adults with a chronic health condition during the COVID-19 pandemic. In the second pilot trial, a digital human was found to be as feasible and acceptable as teletherapy with a human therapist and a self-guided manual at delivering one session of Cognitive Behavioural Stress Management (CBSM) therapy to healthy adult women. Preliminary evidence was found in support of the effectiveness of digital human delivery, with significant improvements in psychological and physiological stress with large effect sizes. Overall, this thesis demonstrated that digital humans can build quality relationships with patients by using behavioural, language, emotive, and appearance-based techniques. However, it is important to tailor relationship building techniques to user characteristics and use context. Digital humans were also found to be a feasible and acceptable technology for delivering psychological interventions to a range of populations. Future research should further examine relationship building techniques in experimental studies, and evaluate the effectiveness of digital human psychotherapy in randomised controlled trials.