Assessment of damages ex machina? A lawyer's perspective on ML-supported assessment of damages in tort law

This paper examines, from a lawyer’s perspective, under what conditions machine learning (ML) can usefully and responsibly support damages assessment in tort law. It argues that damages assessment is not merely a prediction problem: the role that ML can play depends on the type of loss involved and on the way that loss is assessed in legal practice. The paper discusses two application areas: the standardised assessment of routine pecuniary losses, with vehicle damage as the main example, and the equitable assessment of non-pecuniary losses, such as bereavement damages and reputational harm. In the first area, ML can improve efficiency; in the second, it can help reveal patterns in past decisions and support more consistent awards. The paper then identifies the main conditions for responsible use, ranging from data quality and explainability to compatibility with the tort law principle of full reparation and regulatory constraints such as the EU AI Act. It concludes that ML can make damages assessment faster, more consistent and better informed, but only if it remains a tool to support human legal decision-making rather than replace it.