EPoS Winter Summit 2025

Ernst Schraube is Professor of Social Psychology of Technology in the Department of People and Technology at Roskilde University in Denmark. His research centers on the social and political implications of modern technologies in everyday life, with a current focus on digitalization, learning, and educational practice. His books include the co-edited volumes Psychology and the Conduct of Everyday Life (Routledge), Subjectivity and Knowledge: Generalization in the Psychological Study of Everyday Life (Springer), and Psychological Studies of Science and Technology (Palgrave Macmillan), and his most recent book is Digitalization and Learning as a Worlding Practice: Why Dialogue Matters (Routledge). He is former President of the International Society for Theoretical Psychology.

The Wordlessness of ChatGPT: Can Dialogue with AI Chatbots Facilitate Learning as a Worlding Practice?

ABSTRACT

Learning and knowing unfold through dialogue with oneself, with others, and with the learning matter. Generative AI systems, such as ChatGPT, constitute a new form of dialogue. Today, many learners, including university students, use generative AI in their learning activities. They use it not only for practical, operative aspects, such as searching for and accessing learning materials, but also for creative, content-related, and world-engaging learning, such as explaining concepts, interpreting text passages, or providing insights into particular aspects of the world. Drawing on critical and action-based psychology as well as science and technology studies—fields that offer situated, world-disclosing conceptions of learning and knowing—this talk explores the significance of dialogue with AI chatbots in human learning. It describes the politics of generative AI and how the generated responses constitute a new form of synthetic text that is disconnected from the world. Furthermore, it examines the extent to which wordless text can facilitate learning as a worlding practice.

Dr. Johanna L. Degen is a social psychology scholar affiliated with Europa-Universität Flensburg. Her research focuses on parasociality in various forms of intimacy, sexuality, and social relationships, ranging from online dating and social media use to subscription platforms and AI–human interaction. Drawing on her work in clinical and therapeutic work contexts, she combines conceptual analysis with applied practice in ongoing knowledge-transfer processes. Her work bridges theoretical inquiry and reflection (productive critique), empirical research, and practical validation in “real-life settings of everyday people” from a humanist stance. Recent books: The Shaping of the Parasocial Self (Palgrave) and Swipe, Like, Love: Intimität und Beziehung im digitalen Zeitalter (Psychosozial).

“I dont understand what it says, but I think it is correct”: Parasocial Mechanisms and Epistemic Orders in Human-AI Interaction

ABSTRACT

Recent developments in psychological and media studies suggest that being online is not primarily driven by fear of missing out, pragmatic organization of everyday life, such as information seeking, or habitualization, but increasingly by social relatedness. It is shown how parasocial involvement and the relationships it engenders become central to the social self, with significance for social positioning, orientation, decision-making, and even bodily regulation. Such parasocial mechanisms operate beyond cognitive reflection and contribute to the structuring of subjectivity and sociality. When applied to AI environments—such as chat bots (ChatGPT), companion apps (Repiika), and therapy bots (AI Esther)—parasocial dynamics take on distinct qualities, with studies indicating accelerating effects and significance, particularly in terms of ease of access, predictability, endless validation, centering of the user, and epistemic authority. This talk explores the psychological mechanisms of parasociality in Human-AI interactions and introduces evaluated interventions and their implications for media literacy, policies, social work and therapeutic settings.

From Cyber Risk to AI Risk

ABSTRACT

Conventional cyber risk is notoriously difficult to quantify. Can we hope for any improvement in the age of generative AI? In this talk, I will revisit how cyber risk is modelled and measured conventionally, present selected risk breakdown structures for the risk associated with generative AI, and draw some preliminary conclusions on how AI risk could be managed. The lack of data and technological regime shifts make predictions more challenging, but no less relevant.