From Prompt to Practice: Evaluating Unedited AI-Generated Psychometric Scales
Survey Practice · 16.03.2026 · 20 (1) · 1-13
DOI: 10.29115/sp-2026-0002 ↗Kurz gefasst
Fachartikel in der Zeitschrift Survey Practice, online erschienen am 16.03.2026. Verfasst von Peter J. Martini und 2 weiteren Autor:innen.
Abstract (Originalsprache)
The current study assesses whether a non-expert researcher can take an unrefined, AI-generated scale for survey dissemination and obtain results comparable to those derived from conventional, psychometrically developed scales. A sample of 140 respondents completed both established measures (e.g., the Rosenberg Self-Esteem Scale and a 10-item variant of the Right-Wing Authoritarianism Scale) and unaltered AI-derived counterparts. Analyses included factor structure comparisons, reliability assessments, paired t-tests of standardized scores, and Bland-Altman plots of agreement. Results showed that AI-generated items often reproduced similar factor solutions and reliability levels as traditional scales. Agreement was strongest at the whole-scale level, though subscales revealed nuanced differences. These findings suggest that even without expert refinement, AI-generated measures can approximate established scales in many respects. For practitioners, this raises important considerations about efficiency and rigor in survey design: although AI can provide usable starting points, careful evaluation remains essential before deployment in applied research.
Zitieren
APA
Martini, P. J., Scagnelli, P., & Medina, M. E. T. (2026). From Prompt to Practice: Evaluating Unedited AI-Generated Psychometric Scales. Survey Practice, 20(1), 1-13. https://doi.org/10.29115/sp-2026-0002
Vancouver
Martini PJ, Scagnelli P, Medina MET. From Prompt to Practice: Evaluating Unedited AI-Generated Psychometric Scales. Survey Practice. 2026;20(1):1-13. doi:10.29115/sp-2026-0002
BibTeX
@article{Martini2026,
author = {Martini, Peter J. and Scagnelli, Patrick and Medina, Marisol E. Trujillo},
title = {From Prompt to Practice: Evaluating Unedited AI-Generated Psychometric Scales},
journal = {Survey Practice},
year = {2026},
volume = {20},
number = {1},
pages = {1--13},
doi = {10.29115/sp-2026-0002}
}
EUVIMED