“类案类判” 中人工智能与人类智能之限度与效用 - 以刑事类案智能检索与推送系统为出发点

The identification of case similarities is a critical element in legal practice, often facilitating the determination of legal outcomes. However, in the context of artificial intelligence (AI), the selection of elements for case retrieval in intelligent systems necessitates a careful consideration of the specific characteristics of legal disputes. This paper argues for the preference of legal issues over controversial focus or factual points in the design of intelligent case retrieval systems. Leveraging deep learning techniques, we propose a hybrid model that combines Convolutional Neural Networks (CNN) and dictionary rules for the automatic identification of controversial focus sentences. This approach harnesses the technical utility of AI by utilizing legal provisions as classification units, structurally processing legal issues, and prioritizing retrieval results in a three-dimensional ranking. The envisioned framework comprises the development of a case database and an intelligent recommendation system. Human intelligence plays a pivotal role in supervising the confirmation and selection of legal issues, making final determinations regarding the substantive similarity of cases, and considering other factors that may influence case outcomes. By achieving a complementary synergy between human and artificial intelligence, this approach maximizes the utility of both, ultimately realizing the goal of "similar cases, similar judgments."