Year: 2025
Type: dissertation
Abstract: The increasing integration of AI-generated microcontent summaries in everyday use lacks clear evidence of impact on user decision quality compared to simpler methods or original content, raising "AI-washing" concerns. This study empirically investigated the effects of AI-generated summaries of microcontent on user decision quality, considering content collection length, to assess if and when AI integration is advantageous. A controlled experiment (n=22) measured objective and subjective outcomes, constructed to operationalize decision quality as a measure. Results showed users subjectively preferred original microcontent, reporting higher subjective quality (satisfaction and confidence). Objectively, neither AI nor programmatic summaries sig-nificantly improved or harmed accuracy or comprehension compared to original content. AI summaries offered no significant advantage over simpler programmatic techniques. Content length did not alter these comparative effects, though this result should be observed in an ex-plorative manner. Findings suggest AI summarization may offer no tangible benefit for micro-content within this scope, supporting AI-washing critiques. A cautious, user-centric approach, potentially offering original content options, is recommended.
Authors: Melander, Erik, Bjelvér, Jakob
Institution: Lund University
Language: English
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7007884582