Conference Agenda

As universities experience an uncertainty caused by Generative Artificial Intelligence (GenAI), learning is moving from physical spaces (classrooms, libraries, home) to digital dyads between students and Large Language Models (LLMs). This shift poses the question on how educational research shall investigate spaces of learning that are largely invisible and inherently private. Existing research methods including elicited interactions in experiments might fail to understand the actual interaction patterns between students and GenAI, making these processes invisible to educational researchers.

This research proposes to treat conversations between user and GenAI as ethnographic field sites rather than static artifacts. It draws on Constructivist Grounded Theory (Charmaz, 2006) rooted in symbolic interactionism to gain deeper understanding of students’ learning and GenAI’s role in it. By tapping into this “black box” of interactions between students and GenAI, this study aims to locate the university’s place in knowledge construction and learning. The method of constant comparison that this framework requires makes it possible to construct a comprehensive picture of student, GenAI, and university relations.

The study utilizes data from an ongoing research project comparing university students’ interactions with GenAI in Austria and Kazakhstan. The methodology consists of the triangulation of three distinct data layers: 1) The raw logs of student-AI interactions, treated here as "authentic recorded conversations" rather than text outputs; 2) Intensive interviews where students reconstruct their broader academic life and learning habits with GenAI; 3) Students’ comments and reflections on their GenAI logs with the researcher. These three data sources provide insight into students’ learning with GenAI from their point of view, and can help decode the "taken-for-granted assumptions" embedded in these interactions.

By treating the interaction logs as a field site, this methodology reveals the usually invisible shared understandings of learning with GenAI: feelings of guilt and shame, navigation of internal integrity rules, and epistemic negotiations that students develop in the absence or vagueness of institutional guidance. This approach responds to the network's call for "methodological pluralism" bringing in ethnographic approaches towards human-machine interactions. The study argues that gaining access to these "silent classrooms" requires a shift from observation to collaborative reconstruction, where participants act as guides through their own digital learning histories.

In the post-pandemic era, Chinese adolescents are situated at the precarious intersection of rapid technological acceleration and intense educational competition, locally termed neijuan (involution). This specific socio-cultural context is characterised by a hyper-meritocratic evaluation system where the margin for error in real-world social interactions is perceived as shrinking. Simultaneously, Generative Artificial Intelligence (GenAI) has permeated the daily lives of these digital natives. While prevailing educational discourses often pathologise students’ use of GenAI as academic dishonesty or passive dependency, this research identifies an under-theorised phenomenon termed 'algorithmic rehearsal'. This concept refers to the agentic practice where adolescents utilise GenAI not merely as information retrieval tools, but as a 'digital sandbox' to simulate future actions, pre-enact social conflicts, and experiment with potential identities before stepping into the high-risk reality.

The primary objective of this study is to challenge the technological determinist narrative that views youth as passive consumers of algorithms. Instead, it aims to uncover how Chinese adolescents actively appropriate GenAI to construct a safe house for identity exploration. Specifically, the study seeks to: map the diverse scenarios where 'algorithmic rehearsal' occurs; analyse the psychological and sociological mechanisms behind the preference for GenAI over human interlocutors; critically evaluate the implications of this practice for the development of adolescent subjectivity.

Therefore, three research questions are raised:

1. In what specific scenarios (e.g., emotional regulation, career planning, interpersonal conflict) do Chinese adolescents employ GenAI for rehearsal?

2. How does the 'non-judgmental' affordance of AI compensate for the lack of psychological safety in a high-stakes, evaluation-oriented society?

3. Does this reliance on algorithmic simulation constitute a form of empowerment that enhances real-world agency, or does it lead to a dependency that detaches youth from the complexities of authentic human interaction?

This research is grounded in a synthesis of Erving Goffman’s dramaturgical theory and Gert Biesta’s philosophy of education. Firstly, this study appropriates Goffman’s (1959) distinction between front stage and back stage for the digital age. This study conceptualises GenAI as a unique 'digital back stage'. In the Chinese context, the school and family environments often function as a relentless front stage subject to constant surveillance and evaluation. GenAI, therefore, provides a private backstage where the performance of the self can be scripted, rehearsed, and refined without the immediate risk of social embarrassment or losing face. Secondly, this study engages with Biesta’s (2022) notion of subjectification—the process of becoming a subject who can act responsibly and independently. Consequently, this study hypothesises that 'algorithmic rehearsal' represents a struggle for subjectivity. In a system that often reduces students to data points, the active manipulation of GenAI to explore who I might become reflects a desire to reclaim the interpretative power of one’s own life narrative against a deterministic future.

Generative artificial intelligence is fundamentally changing how knowledge is generated, used, and subjectively acquired—especially outside of formal educational settings. This study addresses these changes by examining the use of generative AI by young people as part of open media learning cultures.

Advancing digitalization is increasingly shaping the lives of young people and is accompanied by profound changes in their media use (Suter et al., 2024; Vodafone Foundation Germany, 2025; Feierabend et al., 2025). Smartphones and tablets have become constant companions, enabling young people to be online anytime and (almost) anywhere. In addition to social networks, search engines, and messenger services, generative AI applications in particular have been gaining in importance since ChatGPT became publicly available in November 2022. Using text-, image-, and audio-based AI tools is now part of everyday life for many young people (Feierabend et al., 2025, SINUS Institute, 2025, Yu et al., 2025).

It is well known from the history of technology and media that new technologies are quickly appropriated for sexual purposes once they become widely available (Döring et al., 2021). Initial research findings indicate that sexuality-related activities—such as searching for information, advice, or engaging with erotic content—are increasingly shifting to the realm of generative AI (Döring et al., 2024, Moothedan et al., 2025, Marcantonio et al., 2023).

While current research focuses primarily on sexual interactions between adults and chatbots (Huang et al., 2025; Pentina et al., 2023) in the form of romantic or erotic relationships with AI systems like Replika, the use by adolescents remains largely unexplored.

In particular, there is a lack of empirical studies dealing with the initial, experimental use of generative AI by adolescents. This differs fundamentally from lasting, binding AI relationships, as described in isolated cases among adults. For adolescents, the focus is often on experimentation, exploration, and testing, for example with regard to sexual questions, relationship models, fantasies, or identities. Reasons for using generative AI in this context may include anonymity, freedom from judgment, permanent availability, and low-threshold access to information and spaces for interaction that appear free of embarrassment, stigmatization, or social sanctions.

The availability of digital sexual education resources on platforms such as TikTok, Instagram, or in the form of specialized apps indicates that adolescents are actively using social media to seek information on topics such as contraception, body knowledge, or relationships (Döring, 2022; Ritter & Stephan, 2024). Against this backdrop, it stands to reason that generative AI will increasingly be used as a medium for information, advice, and relationships. However, empirical research on this topic has been very limited to date. One of the few studies involving young people aged 11 to 17 examines predominantly hypothetical usage scenarios and was largely conducted before generative AI became widely available (Woodley & Williams, 2023).

This field of research is particularly relevant with regard to queer adolescents, who often have difficulty accessing sex education due to social stigmatization, queerphobia, and digital violence. Studies on digital hate and anti-feminist online structures show that queer and multiply marginalized youth in particular are exposed to increased risks in the digital space (Ballaschk et al., 2022). At the same time, unlike adult users, they often do not have established protection and coping strategies.

This study addresses this gap by examining how young people experiment with generative AI in contexts of sexual education as well as romantic, sexual, and platonic relationships, with particular attention to queer perspectives. It investigates the extent to which (queer) young people use generative AI tools in relation to questions of sexual knowledge, intimacy, and relationships, and explores the experiences, practices, and meanings they associate with these uses.