Start Date: May 22, 2025
Michael Papademas
The research is driven by a simple but tough question: when is an AI system actually worthy of trust? Trustworthiness is treated not as a single technical score, but as something that emerges from the entire AI ecosystem, namely models and data, but also people, roles, procedures, and governance across the system’s lifecycle. Common assessment approaches (including self-assessment frameworks like ALTAI) are reviewed, with attention to where they fall short, especially when evaluations become subjective or miss the bigger picture. Building on this analysis, the work develops complementary ways to assess trustworthiness that combine algorithmic tools with non-algorithmic approaches. Methodologically, it draws on socio-technical systems theory, link analysis, graph theory, and mathematical programming, grounded in applied ethics and the philosophy of technology.