第14回HIAS Brown Bag Seminar

HIASブラウンバッグセミナーは、一橋大学社会科学高等研究院(HIAS)が主催する新しいセミナーシリーズです。このセミナーシリーズは、HIASの研究者、教員、学生の交流を全学的に促進することを目的としています。11の研究センターを擁するHIASは、今後も全学的な活発な研究協力を促進するハブとして機能していきます。

※ご参加には事前登録が必要です。どなたでもご参加可能です。

<定員に空きがある場合は、当日参加者を受け付けます>

※各自お食事はご準備のうえお持ちください。
<コーヒーとお菓子類をご用意しております>

事前登録期日 : 1月22日(水) 15:00
（定員に達し次第、受付を締め切る場合があります）

—————————————————————————————–

■題目 :「Exploring the structure of legal language using computational techniques: theories, methods and early results」

■講師 :  Simon Deakin （社会科学高等研究院 特任教授／ケンブリッジ大学法学部教授、ビジネス研究センター所長）

■日時： 2024年1月23日(木) 12:40-13:40

■要旨： Advances in machine learning (ML) and natural language processing (NLP) are making it possible to explore the structure of legal language in new ways. An emerging claim associated with these studies is that there is an inherent form or structure to legal language, which statistical techniques are rendering visible in ways which were not possible before. Among the techniques being used are those associated with ‘deep learning’ to study large datasets of legal materials. These methods can be used at scale to indicate what appear to be common structural elements to legal corpora, but interpreting the results they generate is not so straightforward, in part because these applications can be, and in practice often are, run in the absence of a prior causal model. Thus it is not clear that the results they produce are always ‘scientific’ in the generally accepted sense of testing hypotheses generated by prior findings or formal models. The ‘explainability’ problem associated with deep learning, meaning the difficulty in discerning in all cases why a particular reweighting of indicators has been arrived at other than its properties in optimising a given pre-defined function, also makes it hard to draw general conclusions, in the nature of new knowledge claims, from some of these studies. However, the inferential or inductive knowledge they generate may still be useful for testing certain claims concerning law as a societal phenomenon. Moreover, deep learning with large datasets is only one application of ML and NLP currently in use. NLP techniques can be deployed in smaller scale studies to test claims of a social scientific nature, concerning the relationship between legal language, on the one hand, and social, economic and behavioural phenomena, on the other. With smaller scale studies involving human (manual) coding of legal data, explainability, by design, is built in. To illustrate this point, consideration will be given to a recent study using the NLP technique of sentiment analysis to analyse the relationship between changes in legal language in the English poor law (equivalent to today’s labour and social security law) during the transition to an industrial economy between the late seventeenth and early nineteenth centuries (Deakin and Shuku, Journal of Law and Society, forthcoming).

（※）キャンパスマップ⑤の建物です

■言語： 英語

参加登録はこちらから！