Evaluation Criteria for Synthetic Respondents in Qualitative Research: Limits of Applicability
Inter · 03.07.2026 · 18 (2) · 147-160
DOI: 10.19181/inter.2026.18.2.5 ↗Kurz gefasst
Qualitative Studie in der Zeitschrift Inter, erschienen am 03.07.2026. Verfasst von Kirill Zhulikov. Die Arbeit nennt 28 Literaturangaben.
Abstract (Originalsprache)
Large language models (LLMs) are increasingly used to generate synthetic respondents — textual simulations of informants’ responses based on a specified socio-demographic profile. For quantitative synthetic data, evaluation approaches have already been developed that rely on comparing statistical distributions with the results of large-scale surveys. For qualitative data, however, established criteria are lacking. In the literature, this absence is described both as a temporary gap that will be filled as methodological tools develop and as a limitation stemming from a deeper mismatch between the foundations of qualitative methodology and the nature of synthetic data. This article provides an analytical review of approaches to evaluating synthetic respondents across three research traditions: computational social science, natural language processing, and qualitative methodology. A review of publications from 2022 to 2026 shows that the most formalized approaches — statistical comparisons of distributions and automated text metrics — measure properties that are secondary to qualitative research. Approaches developed within the qualitative tradition address substantively important properties, such as contextual specificity, internal contradictions, and narrative resistance, but remain at the level of observations and have not been operationalized. Comparing the properties of qualitative data with existing evaluation criteria reveals areas that remain insufficiently covered, including internal contradictions within narratives, speech features, and the capacity of data to generate non-trivial interpretations. On this basis, the article outlines areas in which the use of synthetic respondents in qualitative research appears justified, as well as areas in which it remains problematic. Minimal transparency standards for cases of their use are also discussed.
Zitieren
APA
Zhulikov, K. (2026). Evaluation Criteria for Synthetic Respondents in Qualitative Research: Limits of Applicability. Inter, 18(2), 147-160. https://doi.org/10.19181/inter.2026.18.2.5
Vancouver
Zhulikov K. Evaluation Criteria for Synthetic Respondents in Qualitative Research: Limits of Applicability. Inter. 2026;18(2):147-160. doi:10.19181/inter.2026.18.2.5
BibTeX
@article{Zhulikov2026,
author = {Zhulikov, Kirill},
title = {Evaluation Criteria for Synthetic Respondents in Qualitative Research: Limits of Applicability},
journal = {Inter},
year = {2026},
volume = {18},
number = {2},
pages = {147--160},
doi = {10.19181/inter.2026.18.2.5}
}
EUVIMED