Artificial Intelligence and the Transformation of Legal Decision-Making

Abstract

Artificial intelligence is increasingly used to support legal and public decision-making, including legal research, document review, risk assessment, administrative processing, and other forms of decision support. Its use can improve speed, consistency, and institutional capacity, but it can also create risks concerning opacity, discriminatory effects, automation bias, procedural fairness, privacy, and the allocation of responsibility. This article examines those risks through qualitative doctrinal analysis and focused comparative policy analysis. It considers the Wisconsin Supreme Court’s decision in State v. Loomis, the European Union Artificial Intelligence Act, Canada’s Directive on Automated Decision-Making, Council of Europe guidance on AI and judicial systems, and selected interdisciplinary scholarship. The article argues that the central legal problem is not simply whether an AI system is accurate, but whether its use can be reconciled with requirements of lawful authority, reasoned decision-making, equality, contestability, institutional accountability, and effective remedy. The article further distinguishes decision-support systems from systems that materially automate or determine legal outcomes and argues that the required safeguards should be proportionate to the legal stakes and degree of automation. The analysis also considers Pakistan, where constitutional fair-trial, equality, dignity, and due-process principles provide an existing legal vocabulary for evaluating public-sector AI, while the National Artificial Intelligence Policy 2025 provides a policy framework rather than a substitute for legislation. The article concludes that responsible public-sector AI requires layered governance: lawful authority, impact assessment, procurement controls, validation and monitoring, meaningful human review, reasons and notice, auditability, access to remedy, and institutional technical capacity.