Interdisciplinary | New Developments in Legal Doctrine Empowered by Generative AI

Interdisciplinary | New Developments in Legal Doctrine Empowered by Generative AI

by
Generative Artificial Intelligence (i.e., generative AI) has sparked numerous discussions in recent years. Many countries and regions have a strong interest in integrating AI into law, but until now, all discussions have been from a judicial rather than a legal doctrinal perspective. From a constitutional perspective, introducing AI into judicial power, which involves binding decisions, faces legitimacy challenges, but legal doctrine that does not involve binding decisions does not. For civil law countries that value legal doctrine, if the state plans to gradually integrate generative AI into practical doctrinal studies, it may lead to a certain kind of disruptive innovation—reversing the doctrinal structure of “practical expansion, theoretical inadequacy” for leading countries; for latecomer countries, it could effectively build a doctrinal structure that balances practicality and theory, quickly surpassing the transitional phase.

The Dilemma of Legal Doctrine in Civil Law Systems

The study of legal doctrine in civil law systems differs from that in common law systems: first, the structure of legal sources is different, i.e., whether cases have independent binding force apart from statutes; second, the research methods differ, with the two major legal systems forming two different research paradigms of “pre-emptive/systematic” and “post-hoc/case”; third, the subjects of action differ, with the understanding of law in common law systems centered around judicial decisions, while civil law systems form a co-constructive relationship between academic theory and court practice.
The “pre-emptive/systematic” research in civil law legal doctrine equally emphasizes the interpretation of statutes and the verification of judicial cases, and involves a significantly higher human investment from individuals with complete legal education participating from various legal professions compared to common law systems. The latter, by rejecting systematic research between statutes, retains a larger space for the formation of individual norms beyond precedent binding, which reduces barriers to the introduction of social science legal studies. In the latter half of the 20th century, American law schools continuously increased collaboration with social sciences, shedding the burden of traditional English legal doctrine. In contrast, German legal doctrine is often criticized for being “circular and absolute,” overly detailed due to excessive investment, stronger in practicality but weaker in theory, leading to high predictability of law but low adaptability, resulting in structural imbalance.
At this point, introducing generative AI into legal doctrine may lead to “disruptive innovation.” Under the influence of generative AI, the practical doctrinal studies of civil law systems are heading towards an end; currently, many large interpretative books published by major German publishers have transitioned to digital and online use. It is not hard to imagine that in the future, through the collaboration of AI experts and senior commentators, relevant content can be gradually generated independently. When resources traditionally invested in practical doctrinal studies are removed, it will guide civil law scholars to invest in theoretical thinking, namely how to reconstruct a more contemporary legal doctrine through a rebalancing of legal principles, even incorporating social science theories, without fundamentally negating existing laws.

Generative AI Aligns with Legal Doctrinal Thinking

The reason for linking generative AI with civil law legal doctrine and believing that it can undertake part of the work in legal doctrine in the future is that there are similarities in their ways of thinking.
First, both generative AI and legal doctrine focus on the generation of new content. Second, the thinking of legal doctrine excels at deconstructing complex issues; for instance, crimes and torts can be broken down into constituent elements, and beneath those elements lie further components, thus presenting a logical diagram formed by multiple sub-judgments. This bears a close resemblance to deep learning AI, whose characteristic is achieving goals through multilayered nesting and phased implementation. Third, neither denies the object of study itself; the main task of legal doctrine is “to start from established norms, confirming the connotation of norms and applying them correctly to social facts as the sole purpose,” while the current AI merely extracts new content from old knowledge without modifying old knowledge or creating new knowledge. Fourth, the basic logical thinking of both is similar; artificial intelligence methods represented by machine learning extract patterns using mathematical algorithms, while legal doctrinal thinking generates more specific doctrines by comparing concepts, statutes, principles, systems, and even similarities and differences between departments. Fifth, AI extracts patterns from massive datasets, and the work of legal doctrine also employs induction; if common components can be distilled from numerous judicial cases and scholarly viewpoints, it can present its relatively subjective objectivity, and the inclusion of AI can allow for an overview of all data relied upon for content generation.

Division Between Practical and Theoretical Doctrines

At this stage, introducing AI into legal doctrinal studies will inevitably be partial rather than comprehensive; the key is how to appropriately segment it. In this regard, we can draw from the “practical/theoretical” dichotomy in German legal doctrine, linking generative AI to practical legal doctrine.
The activities of civil law doctrinal studies can be summarized as “starting from interpretation and ending in systematization.” On the normative side of the law, although systematic research has not yet developed a general methodological framework, once the rule of law reaches a certain level of maturity, the methodology of the interpretative stage is easier to reach a consensus among scholars and judges, thus practical legal doctrine should focus on legal interpretation with relatively simple judgment factors. On the social fact side, most continuously repetitive and similar facts can be placed within the scope of practical legal doctrine, while facts belonging to emerging fields that touch on significant value choices or interest distributions can be excluded. This can delineate the foundational boundaries of practical legal doctrine.
If there is concern that this purely methodological approach is still too abstract, we can also incorporate the “collectivity” hidden in the concept of “practical,” which is AI’s greatest strength. Whether concerning authoritative judicial decisions or non-authoritative purely theoretical analyses, legal professionals will jointly incorporate research from their respective perspectives and accumulate and confirm the degree of consensus through diverse literature tools. Because doctrines and judicial practices pursue balance from the perspectives of jurisprudence and facts, if the threshold for practical legal doctrine is set where both achieve a high level of consensus, concerns about being overly abstract or judicial realism should be eliminated.
In fact, if one observes carefully, the division between theoretical and practical legal doctrines has already been subtly emerging. Taking Germany as an example, its doctrinal literature tools have long presented a division of labor between practical legal doctrine corresponding to detailed interpretative books and theoretical legal doctrine corresponding to specialized books, even before the theoretical foundation for the division existed (journals and textbooks are divided based on their preset author groups). In the future, China’s legal doctrine should also gradually move towards a similar division of labor. Similarly, in legal talent cultivation, the judicial examinations in mainland China and Taiwan have coincidentally divided into two stages: first testing candidates’ basic understanding of law through multiple-choice questions, without involving significant disputes regarding principles or systems; then testing their grasp of legal theories through argumentative questions, even forming the ability to create theories. This can also reflect that the practice of dividing practical legal doctrine and theoretical legal doctrine has already spontaneously existed. Furthermore, I predict that practical legal doctrine will gradually be replaced by AI services. The first to appear will be legal technology organized and invested in on a commercial basis, and only after a considerable period will there be more integrated, even state-supported AI services. Of course, the speed of legal technology development depends on many factors, such as the scale of pre-training legal data corpus, the level of involvement of legal field experts, economic investment in algorithmic computing power development, and the stages of model development from narrow to broad.

Guiding the Development of Legal Doctrine as a National Task

The idea of integrating AI into legal doctrine also involves constitutional questions regarding the role of the state in this process. The civil law system advocates a rationalist thinking of the rule of law, intentionally linking the state with a complete legal system, forming a legal will built collectively by the legal community through legal doctrine in a pre-emptive and systematic manner. However, the introduction of AI could potentially transform the writer’s database into a generative repository that replaces the writer, and as technology matures further, it could become desirable, thereby making AI a new protagonist in practical legal doctrine, triggering constitutional controversies. The foremost concern is whether scholars’ freedom of speech and publication rights are restricted. Even after the introduction of AI, the state will not restrict anyone from continuing to write small papers or legal commentaries. However, if the legal technology available on the market can autonomously publish or go online after a new transaction occurs, individual scholars’ publications may also find themselves without a market, just like Kodak film, which seemingly disappeared overnight due to the proliferation of mobile phones a few years ago.
Therefore, whether guiding the development of legal doctrine is a national task is undoubtedly a question worth considering. From the perspective of the rule of law, rather than the common law system’s rule of law, the state has a more active role in implementing the rule of law, i.e., improving the predictability and adaptability of law through knowledge generation activities like legal doctrine. First, regarding the application of generative AI in legal doctrine, the state should not allow legal technology to develop autonomously. It may be considered that the state indirectly subsidizes the legal doctrine industry by purchasing a large number of doctrinal literature through universities, independent research institutions, and government agencies. Secondly, governments that possess complete judgment information can, at the appropriate time, establish a comprehensive public platform representing the legal community or set up an independent legal entity under government departments, using advanced generative AI to handle most content classified as practical legal doctrine while allowing access for all demanders. Finally, as we gradually transfer practical legal doctrine to AI processing, scholars will gradually withdraw from the market for initial doctrine, disrupting the traditional structure of civil law doctrine, where the co-constructive structure of academic theory and court practice will lose balance. At this point, the integrated public platform will need to adopt a “collective survey” approach to periodically collect and record opinions from representative scholars in different fields to fill the gap left by the sharp decline in doctrinal publications.
In addition, the disruptive innovation brought by the introduction of AI into legal studies will also have spillover effects on legal education, the judicial system, and various legal professions. For instance, introducing legal informatics, social science legal studies, and other teaching content into legal education in civil law systems, emphasizing the importance of theoretical innovation in academic evaluations such as discipline assessments and professor promotions; establishing mechanisms similar to Germany’s “periodic judges” to encourage legal researchers to participate in judicial practice, promoting bilateral exchanges between theoretical law and judicial practice; establishing a pyramid structure for trial levels that ensures the first trial clarifies disputed facts, the second trial is for correction, and the final trial focuses on resolving legal disputes, fully utilizing judicial AI in typical cases to reduce the burden of first trials and achieve professional diversion at the first trial level. If responsible parties can plan supporting adjustments early, they can enhance spillover benefits, reduce costs, and systematically respond to the impact of generative AI on legal studies.
In summary, the bottlenecks faced by the development of legal doctrine in civil law systems are already clear. Whether for Germany, where theoretical research is squeezed leading to structural imbalance, or for countries and regions highly dependent on systematic legal inheritance for building the rule of law, the development of generative AI brings eager hope for breakthroughs in the dilemmas of legal doctrine. Perhaps, given time, the introduction of generative AI can overturn the imbalanced structure of legal doctrine in civil law countries or fundamentally improve the inherent shortcomings of legal doctrine in latecomer countries.
(The author is a distinguished professor at National Chengchi University in Taiwan)

Leave a Comment