Bridging the causation gap: examining the utility of the Res Ipsa Loquitur doctrine for AI-related harms under US tort law

Abstract

The increasing reliance on artificial intelligence (AI) systems has introduced complex legal challenges, particularly in determining liability for AI-caused harms. The opacity, autonomy, and unpredictability of AI systems complicate the application of traditional tort law principles, particularly the doctrines of causation and foreseeability. The inability to trace an AI system’s decision-making process exacerbates information asymmetry between AI developers and injured parties, making it difficult to establish fault under a negligence framework. This study examines whether the doctrine of Res Ipsa Loquitur can be applied to AI harms within the context of US tort law. It explores how AI systems operate and how their characteristics disrupt traditional liability doctrines of causation, both in fact and proximate cause. The study employs a qualitative research method and relies on doctrinal analysis as well as a deductive reasoning approach to reach its conclusions. The study proposes Res Ipsa Loquitur as a viable solution, arguing that in cases where plaintiffs cannot access critical evidence due to information asymmetry between themselves and defendants, then the burden should shift to the opposite side. The study finds that the doctrine is appropriate in situations involving unintentional AI harms that result from either goal misspecification, lack of robustness in the model or unrepresentative training data.