Thorsten Dresing, Uwe Krähnke, Jan Ulrich Hense, Thorsten Pehl
Pseudonymisation and Anonymisation are not Enough!Page 394–407
Methodological, Ethical and Data Protection Challenges in the AI-assisted Processing of Personal Qualitative Data
Shortlink
: https://www.waxmann.com/artikelART106756
.doi: https://doi.org/10.31244/zfe.2026.02.10
Abstract
The use of generative artificial intelligence (AI) for qualitative data analysis is widely debated in research and evaluation, with particular attention to stakeholders’ ethical and data protection concerns. This paper shows that conventional techniques for masking qualitative data, namely pseudonymization and anonymization, are reaching their limits under today’s technological and legal conditions. Even after obvious personal identifiers have been removed from data sources, research findings, evaluation reports, or other published material, AI systems can still identify individuals from the field under study. We argue that common masking practices for personal information raise serious ethical and data protection concerns once data is shared with external AI providers, and we propose practicable options for handling such data appropriately.
Keywords
data protection, anonymization, pseudonymization, research ethics, artificial intelligence
APA citation
Dresing, T., Krähnke, U., Hense, J. & Pehl T. (2026). Pseudo- und Anonymisierung reichen nicht!: Methodische, ethische und datenschutzrechtliche Herausforderungen bei der KI-gestützten Verarbeitung personenbezogener qualitativer Daten. Zeitschrift für Evaluation, 25(2), 394-407. https://doi.org/10.31244/zfe.2026.02.10
