Jelena Zdravkovic “From Powerful to Responsible: Rethinking Digital Twin Architectures”

From 28 June to 1 July 2026, the University of Tartu Institute of Computer Science organised the 17th International Baltic Conference on Digital Business and Intelligent Systems (DB&IS), a biennial international event dedicated to advancing research, innovation and practice across the broad spectrum of digital business, intelligent systems and data-driven technologies. During the event, four keynote talks were given, and recordings are available online. ------ Abstract Digital Twins have been promoted as a wonder of digital innovation, promising unique capabilities to observe, predict, and optimise complex real-world systems. By digitally replicating physical entities, processes, and environments, Digital Twins integrate heterogeneous data streams, track system dynamics in real time, explore alternative futures, and support decision-making across various domains such as healthcare, industry, and critical infrastructures. These capabilities depend on intensive data processing and, today, rely on artificial intelligence and machine learning models to generate predictions and recommendations. As Digital Twins evolve from passive representations to closed-loop systems, they also become actors of the systems they mirror. Through feedback loops, Digital Twins influence decisions and interventions that reshape the physical world, often with direct consequences for human actors. Therefore, optimisation alone is no longer a sufficient design objective. When Digital Twins learn, act, and adapt, questions of responsibility cannot be treated as a post-hoc concern. This keynote argues that Digital Twin architectures need to be by responsible by design. Many current solutions produce excellent outcomes, but fail to make explicit how those outcomes are derived, who is accountable for them, and how harmful or unintended effects can be anticipated and corrected. A responsible Digital Twin must therefore go beyond accuracy and performance. It must be traceable, explainable, corrective, and preventive—not only accountable for its actions, but designed with foresight, ownership, and ethical self-direction. Grounded on information systems engineering and responsible AI, the keynote proposes architectural principles for Digital Twins that can be trusted not only because they work, but because they can act rightly, be understood, traced, and when necessary - governed.