idw – Informationsdienst Wissenschaft

Nachrichten, Termine, Experten

Grafik: idw-Logo
Grafik: idw-Logo

idw - Informationsdienst
Wissenschaft

idw-Abo

idw-News App:

AppStore

Google Play Store



Instanz:
Teilen: 
11.05.2026 11:50

Bayreuth Study Reveals Memory Gaps Regarding AI-Generated Content

Theresa Hübner Pressestelle
Universität Bayreuth

    From August 2026, an EU-wide AI regulation will come into force requiring the labelling of AI-generated content. However, a research team from the University of Bayreuth and Aalto University in Helsinki, Finland, has found that users of AI systems can no longer reliably recall after just one week whether content was generated by AI or not. The researchers presented their findings at the CHI conference, the most important and largest international conference in the field of Human–Computer Interaction.

    ---
    Why it matters

    Interactive and generative AI systems are increasingly used in creative work, such as brainstorming ideas or drafting texts. If AI use is to be disclosed at a later stage—as required by the EU regulation from August 2026 for content created for commercial purposes—an accurate memory of how the content was produced is essential. The findings of the Bayreuth study have implications for how individuals use AI and for the design of AI systems and requirements for AI labelling.
    ---

    Whether summaries, social media captions, infographics, image editing, slogans, or prompts for reflection—the digital world is now filled with AI. To prevent deception and to ensure transparency about the origin of content, a labelling obligation for content intended for the public will apply from this summer.

    “However, after just one week, significant false memories already occur regarding which ideas and texts were generated by AI and which were created by oneself,” says Tim Zindulka, doctoral researcher at the University of Bayreuth’s Mobile Intelligent User Interfaces research group and leading author of the study.

    A total of 184 study participants developed texts either with or without the assistance of AI. After one week, they were asked about how the ideas and the final texts had been produced.
    “Workflows in which human and artificial intelligence were combined were particularly prone to errors—that is, when the idea came from AI but the formulation was done by the human, and vice versa,” explains Professor Dr. Daniel Buschek, Chair of Mobile Intelligent User Interfaces at the University of Bayreuth. After seven days, the probability of correctly recalling the attribute “AI-generated” falls to 37.7% when an AI-generated idea is formulated by a human. Conversely, there is a 64% chance of correctly identifying one’s own idea after a week if the final wording was produced by AI.

    “Our study shows that retrospective questioning about how content was created does not lead to realistic results. As individuals using AI, people may come to adopt AI-generated ideas as their own, or conversely underestimate their own capabilities because they mistakenly recall a self-generated idea as AI-generated,” says Zindulka. He adds: “These findings should also be considered in the design of AI systems and in guidelines for AI labelling. For example, in a university context, it is unrealistic to expect students to retrospectively disclose details of AI use in coursework based on memory alone. Instead, it is more effective to document the process of content creation from the outset.”

    The study was conducted in collaboration with Aalto University in Helsinki, Finland, and funded by the German Research Foundation (DFG; project number 525037874).


    Wissenschaftliche Ansprechpartner:

    Prof. Dr. Daniel Buschek
    Mobile Intelligent User Interfaces
    University of Bayreuth
    Phone: +49 (0)921 / 55-7775
    E-mail: daniel.buschek@uni-bayreuth.de


    Originalpublikation:

    The AI Memory Gap: Users Misremember What They Created With AI or Without. Tim Zindulka et al. CHI ’26 Proceedings of the 2026 CHI Conference on Human Factors in Computing Systems (2026)
    DOI: https://doi.org/10.1145/3772318.3791494


    Bilder

    Merkmale dieser Pressemitteilung:
    Journalisten, Studierende, Wissenschaftler, jedermann
    Informationstechnik
    überregional
    Forschungsprojekte, Wissenschaftliche Publikationen
    Englisch


     

    Hilfe

    Die Suche / Erweiterte Suche im idw-Archiv
    Verknüpfungen

    Sie können Suchbegriffe mit und, oder und / oder nicht verknüpfen, z. B. Philo nicht logie.

    Klammern

    Verknüpfungen können Sie mit Klammern voneinander trennen, z. B. (Philo nicht logie) oder (Psycho und logie).

    Wortgruppen

    Zusammenhängende Worte werden als Wortgruppe gesucht, wenn Sie sie in Anführungsstriche setzen, z. B. „Bundesrepublik Deutschland“.

    Auswahlkriterien

    Die Erweiterte Suche können Sie auch nutzen, ohne Suchbegriffe einzugeben. Sie orientiert sich dann an den Kriterien, die Sie ausgewählt haben (z. B. nach dem Land oder dem Sachgebiet).

    Haben Sie in einer Kategorie kein Kriterium ausgewählt, wird die gesamte Kategorie durchsucht (z.B. alle Sachgebiete oder alle Länder).