Designing Trustworthy Genai: Citation Mechanisms For Calibrating Employee Trust In Organizational Contexts
Eylem Taş, Eva Bittner · Journal of the Association for Information Systems · 2026
AI-generated evidence extraction, verified across multiple analytical personas. Not a substitute for the peer-reviewed original.
Methodology & findings
Study design
Design Science Research (DSR) approach integrating a structured literature review with 21 semi-structured interviews at a tier-one German bank piloting an enterprise GenAI assistant, followed by expert evaluation of design principles..
Primary method
Design Science Research (DSR)
Main result
The study identifies nine recurring issues and twelve user requirements, which are "synthesized into sixteen prescriptive design principles clustered across four thematic areas: (1) content accuracy and verifiability, (2) source credibility and institutional endorsement, (3) transparency and explainability, and (4) personalization and cognitive alignment." Expert evaluation confirms their sociotechnical utility while highlighting context-dependent applicability.
Research paradigm
Design Science Research (DSR) with qualitative exploratory approach
Author conclusions
The authors conclude that "This work contributes theoretically by extending IS trust research to the sociotechnical design of GenAI citation," demonstrating that systematically derived design principles can address the challenge of how "their tendency to generate inaccurate or misattribute sources undermines employee trust and limits responsible adoption" of large language models in organizational contexts.
Risk of bias
Selection bias: participants from single organization (tier-one German bank); Potential confirmation bias in semi-structured interview design; Limited geographic and sectoral diversity; Expert evaluation panel composition not specified in abstract; Selection bias: Single organization (tier-one German bank) limits generalizability to other organizational contexts and sectors; Sample representativeness: 21 semi-structured interviews may not represent diverse employee perspectives across different roles and departments; Potential response bias: Employees volunteering for interviews about trust in GenAI may hold different views than non-participants; Limited expert evaluation details: No information provided about expert selection criteria, number of experts, or evaluation methodology rigor; Single organizational context (tier-one German bank); Limited sample size (21 interviews); Potential selection bias in interview participants; Context-specific findings may not generalize to other organizational settings
Open questions raised
- The paper identifies the gap that "Large language models hold transformative potential for knowledge-intensive work, yet their tendency to generate inaccurate or misattribute sources undermines employee trust and limits responsible adoption," and addresses this through development of design principles for citation mechanisms in GenAI systems.
Explore related topics
Related papers
- Academic Integrity considerations of AI Large Language Models in the post-pandemic era: ChatGPT and beyondMike Perkins · 2023 · 668 citations
- Generative AI tools and assessment: Guidelines of the world's top-ranking universitiesBenjamin Luke Moorhouse · 2023 · 343 citations
- AI-assisted peer reviewAlessandro Checco · 2021 · 261 citations
- Fighting reviewer fatigue or amplifying bias? Considerations and recommendations for use of ChatGPT and other large language models in scholarly peer reviewMohammad Hosseini · 2023 · 209 citations
- Using ChatGPT in academic writing is (not) a form of plagiarism: What does the literature say?Adeeb M. Jarrah · 2023 · 207 citations
- Artificial intelligence to support publishing and peer review: A summary and reviewKayvan Kousha · 2023 · 137 citations