The starting point of the research is that AI is turning from a tool into an actor. When agents write code, make decisions and form organisations among themselves, the established assumptions of software engineering about division of labour, responsibility and quality must be re-examined. The aim is to understand this shift empirically and to build the methods needed to govern it.
AI-native software engineering. How should software development be organised when much of requirements engineering, design, implementation and testing is carried out by autonomous agents? The research develops and evaluates multi-agent systems for different phases of software development and identifies where human judgement cannot be replaced.
Trust and governed autonomy. How much decision-making power can agents be given, and how should their actions be overseen? The goal is practical methods for deploying trustworthy AI in organisations where the cost of error is high.
Zero-human organisations. The research examines companies and services run almost entirely by agents, and what this means for work, management and business.
AI as a scientist. When AI can produce research independently, scientific quality assurance must change. The goal is protocol-driven methods that make the work of autonomous research agents transparent and reproducible.
AI and societal resilience. The research examines the effects of AI on democracy, media and security: information influence, fraud, misuse targeting children and young people, and new forms of public decision-making. The method is working simulations and experiments in which future systems can be studied before they become everyday reality.
All of these themes share one question: what is the place of humans when systems begin to think, write and decide on our behalf?
