Our team lead, Dr. Jessica Szczuka, recently contributed a psychological perspective to an investigation by the Süddeutsche Zeitung. Journalists Natalie Sablowski and Berit Kruse developed their own empirical approach: 25 psychologically grounded fictional personas interacted with ChatGPT over extended conversations, generating a dataset for analysis. It was a genuinely interdisciplinary effort, with our team member Anna Neumann involved directly and Sven Mayer contributing background expertise.
The timing was fitting: we had just submitted a paper investigating how AI systems themselves foster emotional bonds through their interactional behavior, including the asymmetry in how much systems disclose compared to their human partners. The SZ data showed a strikingly similar pattern: chatbot responses were, on average, three times longer than user messages. Combined with constant affirmation and proactive questions, this creates a dynamic Dr. Szczuka described in the interview as “love bombing.”
While not a replication of our own study, seeing similar patterns emerge from an independent approach is a compelling reminder of how actively AI systems shape human-AI interaction and a great example of science communication at the intersection of journalism and research.
Read more at: https://www.sueddeutsche.de/projekte/artikel/politik/chat-gpt-ki-kuenstliche-intelligenz-suizid-experiment-psychotherapie-e688126/?reduced=true
Leave a comment