GENERATIVE AI, TRUST AND TRANSPARENCY IN AN AGE OF ALGORITHMIC KNOWLEDGE

Abstract

Generative artificial intelligence (AI) is increasingly used in education, research, health, professional decision-making, public administration and communication, and it poses new challenges of trust, reliability, transparency and accountability. This article explores the interplay between trust and transparency in generative AI, focusing on the conditions that allow human users to rely on AI-generated outputs appropriately. The analysis distinguishes trust from trustworthiness and from reliance, and it treats generative AI as a sociotechnical and epistemic system rather than a purely technical one. It examines several dimensions of transparency – system-level, output-level, data, epistemic and interactional transparency – and highlights the shortcomings of disclosure, interpretability and explainability as stand-alone tools for building trust. The discussion suggests that when users cannot evaluate the sources, evidential support, limitations or origins of generated information, the ease of conversational interaction and the apparent authority of the output may lead to overconfidence and epistemic overreliance. The study concludes that trustworthy generative AI is not about maximising trust or disclosure but about enabling calibrated reliance. Technical reliability needs to be accompanied by epistemic and functional transparency, so that generative AI can be used in knowledge practices without assuming the status of an epistemic authority.