Our latest research publication, “Trustworthiness Optimisation Process: A Methodology for Assessing and Enhancing Trust in AI Systems,” by  Mattheos Fikardos, Katerina Lepenioti, Dimitris Apostolou and Gregoris Mentzas was published in Electronics 2025.

In today’s AI-driven world, ensuring that our systems are not only powerful but also trustworthy is more critical than ever.

Our paper introduces TOP, a comprehensive methodology that operationalises trustworthiness across the entire AI lifecycle.

🔍 Four Stages of TOP:
• Identify: Collect & document socio-technical information and requirements.
• Assess: Quantitative and Risk assessment of the AI System.
• Explore: Investigate a range of mitigation techniques tailored to address specific challenges.
• Enhance: Implement and continuously monitor improvements.

✏️ TOP’s properties:
• AI system lifecycle compatibility
• Extensibility
• Conflict consideration
• Human-in-the-centre
• Multidisciplinary engagement

📋 Key enablers:
Documentation Cards: We leverage standardized cards (use case, data, model, and method cards) to capture crucial system details, ensuring transparency and accountability at every stage.
Risk Management Integration: Coupling the process with risk management practices enables targeted, effective interventions that align with ethical and regulatory standards.

Our goal with TOP is to bridge the gap between high-level ethical principles and practical, actionable strategies—empowering AI stakeholders to build systems that are trustworthy.

🔗 Read the full paper here.

Research within the THEMIS 5.0 project.