From ML Predictions to Informed Diagnostic Assistance Using the Toulmin Model of Argumentation

从机器学习预测到基于图尔敏论证模型的辅助诊断

Abstract: To provide a structured and interpretable assessment, we decompose the image-based diagnosis into components following the Toulmin model of argumentation. This model consists of a claim, grounds, warrant, qualifier, rebuttal, and backing.

摘要： 为了提供结构化且可解释的评估，我们遵循图尔敏（Toulmin）论证模型，将基于图像的诊断分解为若干组成部分。该模型包含主张（claim）、根据（grounds）、论据（warrant）、限定词（qualifier）、反驳（rebuttal）和支持（backing）。

Consider a claim generated by a machine learning (ML) model for retinal diagnosis. Rather than accepting this claim at face value, one could either apply explainable AI (XAI) methods or adopt an argumentation-based approach.

以机器学习（ML）模型生成的视网膜诊断主张为例。与其直接接受这一主张，不如应用可解释人工智能（XAI）方法，或采用基于论证的方法。

In our framework, a model specialized in biomarker extraction from images provides the grounds. The warrant—linking the grounds to the claim—is analyzed by an agent equipped with medical knowledge; in our architecture, this role is fulfilled by a MedGemma agent.

在我们的框架中，一个专门用于从图像中提取生物标志物的模型提供了“根据”。连接“根据”与“主张”的“论据”则由具备医学知识的智能体进行分析；在我们的架构中，这一角色由 MedGemma 智能体担任。

The qualifier is determined based on the overall quantitative evaluation of both the warrant and grounds models. Finally, a rebuttal is constructed using image similarity measures computed with MedSigLip. All these components are presented to the human expert, enabling a more informed and critical assessment of the ML-generated diagnosis.

“限定词”是根据对“论据”和“根据”模型进行整体定量评估后确定的。最后，利用 MedSigLip 计算出的图像相似度指标构建“反驳”。所有这些组件都会呈现给人类专家，从而使他们能够对机器学习生成的诊断结果进行更明智、更具批判性的评估。