{"id":69251,"date":"2025-07-23T14:18:48","date_gmt":"2025-07-23T12:18:48","guid":{"rendered":"https:\/\/www.audiotranskription.de\/resources\/serendipity-prompting\/"},"modified":"2026-08-12T11:19:25","modified_gmt":"2026-08-12T09:19:25","slug":"serendipity-prompting","status":"publish","type":"page","link":"https:\/\/www.audiotranskription.de\/en\/resources\/serendipity-prompting\/","title":{"rendered":"Serendipity Prompting"},"content":{"rendered":"

[vc_section][vc_row row_padding=””][vc_column]

\"audiotranskription<\/div>

AI in interpretative qualitative research<\/p>

Serendipity Prompting<\/span><\/h1><\/div><\/div><\/div>[\/vc_column][\/vc_row][\/vc_section][vc_section row_padding=”vpt-7 vpb-7″][vc_row row_padding=”custom” row_padding_bottom=”vpb-5″][vc_column width=”3\/4″ col_sticky=”” offset=”vc_col-lg-offset-2 vc_col-lg-8 vc_col-sm-offset-1″]

Large Language Models as Sources of Maieutic Inspiration in Qualitative Data Analysis
<\/div><\/h2><\/div><\/div>[vc_column_text]<\/p>\n

The debate over generative AI in qualitative research is polarizing. Some call for abstaining from using it, while others adopt LLM outputs largely without scrutiny. Our concept of \u201cSerendipity Prompting\u201d<\/strong> (Uwe Kr\u00e4hnke, Thorsten Dresing, Thorsten Pehl) represents a third approach and, to that end, reverses the conventional use of AI. Language models typically function as response engines. The LLM generates results that humans use, and with that convenience, agency and interpretive authority shift to the machine. Studies on cognitive offloading, cognitive surrender, and deskilling show what is at stake here. With Serendipity Prompting, the LLM provides maieutic prompts in the form of questions<\/strong> rather than ready-made answers.<\/p>\n

The prerequisite is the participant\u2019s own initial interpretation<\/strong>: The participant selects a short passage from the material, notes down their interpretation, and submits both along with a brief profile and the prompt template in a single message. This issue consists of three staggered sections: The \u201cValidation Impulse\u201d<\/strong> examines the extent to which one\u2019s own interpretation of the material can be developed and substantiated. The falsification principle<\/strong> interprets the same finding as another plausible possibility. The synthesis prompt<\/strong> asks open-ended “W” questions that delve deeper into the same topic. Every interpretation remains in the subjunctive; quotations are reproduced verbatim; praise and definitive conclusions are ruled out\u2014the model is framed as a co-reader, not as an expert system. The researcher retains full interpretive authority throughout the process.<\/p>\n

In theory, the procedure is anchored in two ways. In Socratic maieutics\u2014 <\/strong>the art of using questions <\/strong>to help people discover their own insights\u2014and in the principle of serendipity<\/strong>\u2014the productive use of the unexpected. The goal is methodical curiosity\u2014to engage in thinking and deliberation. The interpretive space is opened up rather than prematurely narrowed; abductive inferences and the discovery of sensitizing concepts are set in motion, while every impulse simultaneously drives the process toward validation through the material. The interpreters are productively challenged, encouraged to \u201cview things differently than usual,\u201d to \u201cthink outside the box,\u201d to be confronted with their \u201cown blind spots,\u201d and to solve a problem by \u201cdetaching themselves from the problem\u201d (see Kr\u00e4hnke, Uwe 2017: \u201cThe Social Scientist\u2019s Diagnosis of the Times as a Point of Reference.\u201d In: Matthias Junge [ed.]: Metaphors of Sociological Diagnoses of the Times. Wiesbaden: Springer. pp. 7\u201319). In this way, the approach counteracts the \u201cproving-the-obvious\u201d problem of generative AI and strengthens researchers\u2019 analytical agency rather than replacing it (see Dr\u00f6ge 2025, https:\/\/blogs.surrey.ac.uk\/caqdas\/2025\/05\/27\/why-ai-has-a-proving-the-obvious-problem-and-what-we-can-do-about-it\/<\/a>). To explore a text passage analyzed in this way in greater depth, the hybrid interpretation approach using dialogically integrated LLMs<\/a> (Kr\u00e4hnke, Pehl & Dresing 2025) provides a direct follow-up.<\/p>\n

Scientific publication<\/b><\/p>\n