Roxana Rădulescu, DECIDE co-applicant and WP5 co-leader at Utrecht University, gave a keynote talk in May 2026. The talk, “From Friction to Synergy: Building Human-Aligned Agents with Multi-Objective Reinforcement Learning,” was presented at the Citizen-Centric Multiagent Systems (C-MAS) workshop. The workshop took place as part of the Autonomous Agents and Multiagent Systems (AAMAS 2026) conference.
Why it matters for transparent AI
Many real-world problems involve multiple stakeholders with conflicting goals. Traditional reinforcement learning often reduces these trade-offs to a single reward signal. Rădulescu’s talk showed how Multi-Objective Reinforcement Learning (MORL) keeps these trade-offs visible instead. This approach supports explainability, transparency, and trust — directly relevant to DECIDE’s work on human-aligned decision-making.
Readers interested in the technical foundations can find an accessible introduction in Rădulescu’s co-authored guide to MORL.