On the Behavioral Structure of Legal Reasoning - Based on An Analysis of Two Generations of Legal Artificial Intelligence

试论法律推理的行为结构—基于对两代法律人工智能的分析

Traditional conceptions of legal reasoning fall into the fallacy of treating legal logic as mere formal logic and simplistically equating legal reasoning with mental activity. Legal reasoning should encompass both intrinsic psychological and extrinsic physical dimensions, constituting a legal practice behavior. Traditional legal reasoning concepts are reduced to symbolist legal artificial intelligence, which relies on symbolic representations of rules and precedents to achieve legal analysis through deduction and analogy, reflecting the path of legal justification. The new generation of legal reasoning is reduced to connectionist legal artificial intelligence. Grounded in data-driven approaches, connectionist legal AI simulates human brain information processing through neural networks, embodying the path of legal discovery. The behavioral structural model of legal reasoning treats legal discovery and justification as an inseparable whole, emphasizing that the evaluation of legal reasoning's validity should focus on the external physical component, while the internal psychological component achieves legal reasoning execution by influencing the external physical component. This model offers a fresh perspective on understanding the essence of legal reasoning and provides a theoretical foundation for the development of legal artificial intelligence.

传统法律推理观念存在将法律逻辑仅视为形式逻辑和将法律推理简单等同于思维活动的谬误。法律推理应包含内在心理和外在物理两个方面，是一种法律实践行为。传统法律推理观念被还原到符号主义法律人工智能中，其依赖于规则和案例的符号化表示，通过演绎和类比推理实现法律分析，体现了法的证立路径；新一代法律推理观念被还原到连接主义法律人工智能中，连接主义法律人工智能以数据驱动为基础，通过神经网络模拟人脑的信息处理过程，体现了法的发现路径。法律推理的行为结构模型，将法的发现与证立视为一个不可分割的整体，强调法律推理的有效性评判应关注外在物理部分，而内在心理部分则通过影响外在物理部分来实现法律推理的执行。这一模型为理解法律推理的本质提供了新的视角，并为法律人工智能的发展提供了理论基础。