Home / Research
Research

Co-PI Dr. Feng Liu of the Facial Expression Recognition and Affective Computing Research Group Publishes an Interdisciplinary Perspective Article on AI and Law in The Innovation

2026-11-18

Recently, the team of Dr. Feng Liu, Co-PI of the Facial Expression Recognition and Affective Computing Research Group at the School of Psychology, Shanghai Jiao Tong University, including Hailong Sun, Junhan Gu, published an article entitled “LLMs are reshaping common law reasoning: the challenge of technical reason to artificial reason” in the comprehensive international journal The Innovation. From the interdisciplinary perspective of artificial intelligence and law, the article examines how large language models (LLMs) deeply intervene in common law reasoning and the new challenges this intervention poses to the traditional boundary between “technical reason” and “artificial reason.”

Figure 1. The left panel represents the “artificial reason” of the common law; the middle panel represents the “technical reason” introduced by LLMs; and the right panel represents normative reflection and constrained human–AI collaboration. Figure 1 also illustrates the three impact pathways: adjudication, legal knowledge production, and professional autonomy.

The common law is often misunderstood as “judges deciding cases based on experience.” In fact, as early as 1608, in Prohibitions del Roy, Coke emphasized that the common law embodies an “artificial reason.” It is neither “instrumental rationality” nor “natural reason” shared by all human beings, but a “professional rationality” gradually accumulated by the legal professional community through long-term study, apprenticeship, and judicial practice. A.W.B. Simpson further pointed out that in the common law, it is difficult to sharply separate what is “legally correct” from what is “rational, just, and fitting.” This means that common law reasoning relies on analogy, historical continuity, and case-by-case weighing, and also on the collective memory of judges and lawyers regarding precedents (Figure 1).

LLMs are not merely faster search engines or rule-based expert systems; they are probabilistic models trained on massive text corpora. They can generate legal summaries, draft clauses, predict outcomes, and perform formal analogies, while also producing hallucinations, exhibiting prompt sensitivity, and generating answers in a “quasi-argumentative” manner. This article refers to this AI rationality, centered on efficiency, calculability, and formal consistency, as “technical reason,” to distinguish it from traditional instrumental rationality. It challenges the common law through three pathways: adjudicative decision support, legal knowledge production, and institutional governance. In fact, technical reason is not entirely new; similar logic has long existed in codification and bureaucratic management, but its power becomes unprecedented when embedded in algorithms.

Its impact can be summarized as a “threefold reshaping.” First, the objective of adjudication may drift from individualized substantive justice toward standardized data-driven output. Judges’ intuition and legal sensibility come from years of practice and are part of artificial reason. When courts increasingly rely on risk scores or LLM-generated outcome predictions, this tacit experience may be marginalized. Second, the grasp of precedent and historical memory may shift from deep study to “just-in-time retrieval.” AI-assisted retrieval can significantly improve efficiency, but it may also weaken lawyers’ internal motivation to trace the lineage of legal doctrine, making common law reasoning increasingly “ahistorical.” Third, the professional autonomy of the legal community may be eroded. When algorithmic systems become indispensable tools for research, document drafting, and even judicial deliberation, professional authority may shift partly from legally trained persons to data scientists and system architects. The “black-box” nature of algorithmic systems also makes it difficult for judges to understand, verify, and challenge their outputs in specific contexts, thereby creating a new form of “technological dependence.”

However, the article does not therefore deny the value of LLMs. When used appropriately, they can facilitate access to legal information, reduce routine burdens, flag contradictory precedents, and provide legal information assistance to litigants without legal representation; for judges, they can also serve as a “second pair of eyes.” What truly warrants vigilance is not technological progress itself, but the uncritical subordination of legal judgment to technical logic.

In the future, the relationship between technical reason and artificial reason should not be viewed as a zero-sum game. The key is that technological application must preserve those values that algorithms cannot optimize and should not replace, including the irreducible particularity of individual cases, the requirement that judicial reasoning be subject to public accountability, the narrative coherence of legal doctrine, and human moral agency. It can be said that the common law has survived for centuries by adapting rather than rejecting change; while respecting the core of artificial reason, establishing constrained collaboration with LLMs will be a path worth exploring.

Dr. Feng Liu is the corresponding author of the article. The first author, Hailong Sun, is a doctoral candidate in law at the KoGuan School of Law, Shanghai Jiao Tong University. The second author, Junhan Gu, is an undergraduate student in psychology at the School of Psychology and Cognitive Science, East China Normal University, and is also a student recommended for postgraduate study by corresponding author Dr. Feng Liu. This work was supported by the Startup Fund for Young Faculty at SJTU (Grant No. 25X010506040) and the 2026 Research Topics on the Work of the Communist Youth League in Shanghai Schools (Grant No. 2026LXP366).

Authors: Hailong SunΨ, Junhan Gu, Feng Liu*

Affiliations: KoGuan School of Law, Shanghai Jiao Tong University; School of Psychology and Cognitive Science, East China Normal University; School of Psychology, Shanghai Jiao Tong University

Article information: Hailong Sun, Junhan Gu, Feng Liu. LLMs are reshaping common law reasoning: the challenge of technical reason to artificial reason. The Innovation, 2026, 101636. DOI: 10.1016/j.xinn.2026.101636.

Online publication URL: https://www.cell.com/the-innovation/fulltext/S2666-6758(26)00383-8

Published online: November 05, 2026.