Breaking the Hyperreal Loop : AI-generated content, compulsive engagement, and design interventions for digital wellbeing in short-form video
Year: 2026
Type: dissertation
Abstract: Short-form video feeds are now a primary media channel for young adults, and AI-generated content increasingly circulates within them, so what a user encounters may be machine-selected and machine-made at once. The concept of hyperreality suggests this should erode the distinction between the recorded and the generated, but whether users experience it that way has not been examined directly. This thesis asks in what ways AI-generated content figures in compulsive engagement - here meaning scrolling that continues without a decision to continue, and that ends only when something outside the app intervenes - and how design might make the workings of a feed legible from inside it. Following a Research through Design methodology, I conducted masked observation and in-depth interviews with eight high-frequency users of TikTok and Instagram Reels, aged 18 to 25. The observation was masked in the sense that participants were told the study concerned short-form video use and interface design, while the focus on AI-generated content and habitual scrolling was withheld until the debriefing. I analysed the material using reflexive thematic analysis and built a low-fidelity prototype of four interface components, layered onto an unmodified feed, from the findings. Participants described scrolling as automatic to begin and effortful to end and reported that the feed's own accuracy often produces the boredom that ends a session. They had not lost the ability to identify generated content, and where their cues failed, they had improvised a collective verification practice in the comment section. What had changed was the default: they approached the feed already assuming production. Hyperreality here operates not as a failure of perception but as the retirement of a question: participants could still tell recorded from generated content, but nothing in the feed gave them any occasion to ask, so they had stopped asking. That reframes hyperreality as an orientation that design might reach rather than a fixed condition of the medium. No participant raised data use, network load or environmental cost at any point. I argue that this is not a matter of inattention on their part: the interface holds no representation of the infrastructure that delivers the feed, so its demands cannot enter a user's decisions even in principle.
Author: Radyvonenko, Anastasiia
Language: English
Subfield: Human-Computer Interaction
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7223162712
