Conference Agenda

Given that the adoption and acceptance of emerging technologies can be shaped by perceptions on social media, this study examines discussions about communicative artificial intelligence (ComAI) on Facebook (n = 2,551) and Instagram (n = 3,348) between November 2022 and May 2024 using LLM-based coding and topic modeling. The analysis focuses on the prevalence of risk and benefit frames, the expression of functions and reasons for the use of ComAI, and the inductive identification of topics. Our preliminary findings indicate a multifaceted discussion about ComAI on social media, ranging from topics such as “Technological Developments & Functions” and “Mis- & Disinformation” to topics focusing on the testing of and experiences with ComAI tools in everyday contexts (e.g., “Daily Use & Recipes”, “Events & Workshops”). While broader societal implications are discussed, users also share experiences with concrete applications, suggesting that they make sense of these technologies through shared experiences.

This study examines the cognitive strategies users employ when interacting with a conversational AI system across three experimentally assigned tasks: creative writing, concept explanation, and text proofreading. Using an LLM-based classification pipeline, 524 US participants were coded along eight psychological dimensions grounded in motivation, self-regulation, and epistemic cognition theory. Latent class analysis identified six distinct cognitive strategy profiles that predict engagement depth more strongly than any individual dimension. A two-stage hurdle model reveals that engagement operates through separate mechanisms: self-regulation strategy determines whether users engage beyond a single turn, while need for cognition drives depth among those who do. Task difficulty did not differ across conditions and thus cannot account for the observed strategy differences. The findings offer a psychologically grounded typology of user engagement with conversational AI and demonstrate the feasibility of LLM-based conversation classification at scale.

Humor is widely considered a mechanism for fostering rapport and relational closeness in human–machine communication (HMC), yet which chatbot humor styles work—and why—remains poorly understood. Drawing on humor style theory and co-constructive perspectives on HMC, this study examines how affiliative, aggressive, self-defeating, and self-enhancing humor styles shape user evaluations. In a between-subjects experiment (N = 231), participants interacted with a chatbot employing one of four humor styles, rated on funniness, sociality, pleasantness, and popularity. Results show that humor effects depend more on interpretation than on design: affiliative readings enhanced evaluations across all styles, while misalignments proved consequential. Most strikingly, self-defeating humor received the highest evaluations, while self-enhancing humor underperformed most severely. This reflects an ontological asymmetry constitutive of HMC: whereas human–human communication rewards affiliation, HMC appears to reward hierarchy—the machine that jokes at its own expense affirms the proper status order; the machine that signals confidence transgresses it.

Generative AI is increasingly used to produce media content, yet it often reproduces or amplifies existing social biases. To understand how users perceive such biases, evaluate them morally, and see their own and others' role in addressing them, we conducted 36 semi‑structured interviews with individuals differing in AI experience. Participants expressed ambivalent views: Many recognized that biased training data, probabilistic generation, and AI’s limited reflective capacities can lead to biased outputs, while AI was also perceived as neutral and bias attributed primarily to human creators and users. Moral evaluations centered on perceived agency and responsibility, with humans judged as more accountable than AI. At the same time, participants worried that AI could amplify stereotypes due to its capability to produce large amounts of media content. Responsibility for countering AI bias was attributed across micro‑ (users, developers), meso‑ (companies, platforms), and macrolevels (policy), with effective mitigation seen as requiring coordinated action.