Exploring companionship and social connectedness in interactions of groups vulnerable to social isolation with conversational virtual humans (CVHs)

The significance of social connectedness for emotional resilience and mental health is crucial in today's modern world. However, many individuals and communities are facing challenges in establishing and maintaining meaningful relationships, leading to a widespread issue of loneliness that negatively impacts mental well-being and societal cohesion. Certain groups, such as those with physical, cognitive, or social limitations, are particularly vulnerable to difficulties related to socialization and social connectedness, which can result in social exclusion and isolation, further hindering the individuals' ability to engage in social interactions.

The exacerbation of social isolation among vulnerable groups is a complex issue that requires a comprehensive approach to address. Specifically, groups like autistic adults, mourners, and those living with dementia encounter unique barriers to forming and maintaining meaningful social connections. These vulnerabilities emphasise the urgent need for inclusive and empathetic strategies to promote social connectedness among these vulnerable groups to foster meaningful social connections and combat the loneliness epidemic in our society.

This PhD thesis addresses this complex problem by investigating the role of conversational virtual humans (CVHs) in enhancing social connectedness among those three vulnerable groups: autistic adults, mourners, and individuals living with early-stage dementia. By analysing interactions between these groups and CVHs, this research aims to uncover how such technologies can support emotional well-being, mitigate feelings of isolation, and potentially enhance companionship and social connectedness. A multi-case approach was used to facilitate a comparative analysis that enriches our understanding of the phenomena across different contexts, based on qualitative and quantitative data analyses from interviews and questionnaires. The choice of research methodology was driven by the specific needs of the specific groups being studied, the nature of the research question, and the resources available for the

study. Hence, an experimental approach with an off-the-shelf conversational agent was used for autistic adults, an exploratory approach with existing chatbots and griefbots was used for mourners, and an experimental approach with a custom-built conversational app (through a co-design process) was used for people with early-stage dementia.

Autistic adults found CVHs to be a non-judgmental platform for social interaction, appreciating the controlled and predictable nature of these interactions. CVHs helped users feel more understood and less isolated, suggesting potential for improving social skills and confidence. Mourners utilised CVHs for emotional support, finding comfort in the ability to express grief without burdening others, and fostering a sense of presence for the deceased, aiding the grieving process. The CVH we designed and developed for people with early-stage dementia offered cognitive and social benefits to individuals with dementia, including memory stimulation and daily routine support. These interactions contributed to a sense of companionship highlighting the potential of CVHs in dementia care.

This thesis provides a comprehensive analysis across the studies, identifying common themes such as the importance of personalisation, the need for advanced emotional intelligence in CVHs, and the potential of these technologies to serve as complementary tools in care and support systems. The analysis highlights the transformative potential of CVHs in addressing the unique needs of vulnerable populations, while also pointing out the technological and ethical challenges that need to be addressed.

The findings from these studies have also several implications for the future development of CVHs and related research. Firstly, there is a clear need for advancements in emotional intelligence in CVHs. Secondly, the development of CVHs should be guided by a deep understanding of the target user group, focusing on user-centric design. Thirdly, ethical considerations, especially in sensitive contexts, are paramount in ensuring the responsible use of CVHs.

Fourthly, longitudinal studies are essential to assess the long-term impact of CVH interactions on users' social well-being and mental health. Finally, addressing the current technological limitations and exploring new opportunities in AI will be crucial for the evolution of CVHs. Research in this thesis has implications for CVH researchers, user groups and practitioners urging them to consider the deployment of the CVHs as well as the ethical design towards a user-centric and responsible approach.