Existing methods of evidential reasoning include narrative, probabilistic and argumentative approaches. Narrative and probabilistic methods often face issues of high formality and arbitrariness. In contrast, argumentative methods are more factually plausible due to their practical and appropriately formalized advantages. With the growing popularity of argumentation diagramming tools in recent years, these tools use visual interfaces to present premises, conclusions, and sequences of arguments, aiding users in identifying and analyzing arguments. Diagramming, as a reasoning system based on inductive logic, provides a valuable knowledge resource for the evidentiary reasoning process in AI legal systems. Contemporary argumentation methods primarily use diagrammatic forms to illustrate the process of evidential reasoning. These including Wigmore’s diagrammatic method, the improved version of Twining et al.’s diagrammatic method, and Walton’s diagrammatic method. However, these existing diagrammatic methods require further refinement due to their inherent complexity and bias, in order to adapt to the increasing complexity of AI legal systems and social structures. Based on this, we propose a path for modeling evidential reasoning through a three-dimensional argumentation progression of analysis, dialogue, and assessment. This new argumentation method, as a systematic practical approach to evidential reasoning, can substantiate complex facts in challenging cases. It can also transform the substantiation process into a multi- inheritance reasoning system within AI legal systems, which is crucial for the future development of AI legal systems in fact-finding.