Harvesting Trust Indicators: A Method for Evaluating Trustworthiness of RAG-Driven Advice
Abstract
AI-generated advice is increasingly used in hybrid human-AI settings for both personal and professional decision-making. In high-stakes domains such as agriculture, incorrect guidance can cause significant economic and environmental harm. With the growing adoption of Large Language Models (LLMs) for advisory tasks, it remains unclear whether typical quantitative evaluations accurately reflect trustworthiness. In this work, we propose a method to identify and analyze indicators of trustworthy AI advice, by investigating the relationship between system- and user-based metrics with human-perceived trustworthiness. We apply this method to an agricultural decision-support setting, in which advice is generated through knowledge-driven Retrieval-Augmented Generation (RAG). To support this analysis, we introduce AgriAdviceQA, a benchmark consisting of 100 question-answer-context entries of information and advice questions, derived from 3,000+ agricultural best-practice documents and their metadata. We compare retrieval and generation performance of three different AI system designs (LLM-only, RAG, and metadata-augmented RAG) against AgriAdviceQA’s ground truth. A user study is performed with domain experts, who rate Perceived Fluency, Relevance, Simplicity, and Trustworthiness. Results indicate significant correlations between system and user metrics, supported by qualitative insights. These findings demonstrate that our method can meaningfully identify trust indicators, which…
Read the paper
Related studies
- Knowledge Management Perspective of Generative Artificial Intelligence
- Navigating Generative Artificial Intelligence Promises and Perils for Knowledge and Creative Work
- Characterizing the evolving research on enterprise content management
- Large language models and multimodal foundation models for precision oncology
- From Generative AI to Extended Reality: Multidisciplinary Perspectives on the Challenges, Opportunities, and Future of Educational Computing
- Generative Artificial Intelligence: A New Engine for Advancing Environmental Science and Engineering
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.