Academic Interests
Preferences and collective decision-making. Human preferences are at the heart of many AI systems, yet they are notoriously difficult to model and reason about. My research develops methods for learning, eliciting, and aggregating preferences, with applications ranging from recommender systems to collective decision-making and the alignment of language models with diverse human preferences.
Fairness in AI. I study how fairness should be understood in AI systems that make decisions based on human preferences. This includes analysing different notions of fairness and their implications, as well as designing methods that balance fairness with other objectives in collective decision-making and AI-assisted decision support. I am also interested in the relationship between fairness and algorithmic recourse, particularly in settings with limited resources where recommendations made to one individual can affect opportunities available to others.
Methods. My work is primarily theoretical, focusing on mathematical models and algorithms for AI and decision-making. Depending on the problem, I draw on techniques from areas including combinatorial optimisation, graph theory, logic, learning theory, and online learning, such as dueling bandits, while maintaining a strong emphasis on principled modelling.
Previous work. Previously, I worked on computational social choice, with a particular focus on approval voting, liquid democracy, stable matchings and methods for preference modelling and inference.