Reconstructing Legal Responsibility and Causation in the Age of Algorithms: From Linear Liability to Distributed and Probabilistic Models

This paper develops a foundational rethinking of legal responsibility and causation in the context of algorithmic systems and artificial intelligence. It argues that traditional legal doctrines—built upon linear causation and individual fault—are structurally inadequate to address the complexity of algorithmic environments characterized by opacity, autonomy, and distributed agency. The study critically examines the classical two-tier model of causation (factual and proximate causation), demonstrating its limitations when applied to AI-mediated decision-making, where outcomes often result from multi-layered interactions between human actors, data infrastructures, and algorithmic processes. Adopting a doctrinal and interdisciplinary approach, the research proposes a reconstructed model of legal responsibility based on distributed accountability and probabilistic causation. It integrates insights from legal theory, philosophy of causation, and computational systems to redefine the relationship between action, outcome, and liability. The paper further explores the implications of this reconstruction for liability regimes, judicial reasoning, and regulatory frameworks, emphasizing the need for adaptive legal concepts capable of capturing the complexity of algorithmic causation. The study contributes to contemporary legal scholarship by offering a novel theoretical framework that bridges the gap between traditional legal reasoning and the realities of algorithmic decision-making, positioning causation as a dynamic and context-dependent construct rather than a fixed linear relation.