The Evolving Role of AI in Legal Judgment – Law, Innovation and Technology

‘In the 2010’s, scholars had reached a consensus that the role of artificial intelligence in legal judgment was constrained by limitations inherent in machine learning and other forms of AI. AI could not reason by analogy, draw on a wider context, be sensitive to quickly shifting social norms, or effectively address algorithmic bias and opacity. But the AI critiqued here mainly involved tools for predicting scores or probabilities of an outcome. This paper argues that AI involving language models fundamentally challenges these earlier assumptions, calling for a reassessment of the prior consensus. Unlike earlier forms of AI, language models are increasingly effective at engaging in analogical reasoning, being context aware, and mitigating bias and opacity in judgement. The paper demonstrates this by canvassing an experiment using briefs and factums from apex court cases to render decisions with these models that closely match the actual outcomes in those cases. The findings point to legal judgment becoming increasingly collaborative between humans and machines, and distinctions between automated processes and human intuition becoming increasingly blurred.’

Link: https://dx.doi.org/10.2139/ssrn.5289203