[论文解读] Addressing Intersectionality, Explainability, and Ethics in AI-Driven Diagnostics: A Rebuttal and Call for Transdiciplinary Action

Myles Joshua Toledo Tan, Panayiotis V. Benos|ArXiv.org|Jan 15, 2025

Artificial Intelligence in Healthcare and Education被引用 7

一句话总结

本文驳斥以纯粹准确性为重点的AI诊断观点，并呼吁关注交叉性、隐私、可解释性以及跨学科协作，以确保以伦理、实现公平的AI驱动医疗保健。

ABSTRACT

The increasing integration of artificial intelligence (AI) into medical diagnostics necessitates a critical examination of its ethical and practical implications. While the prioritization of diagnostic accuracy, as advocated by Sabuncu et al. (2025), is essential, this approach risks oversimplifying complex socio-ethical issues, including fairness, privacy, and intersectionality. This rebuttal emphasizes the dangers of reducing multifaceted health disparities to quantifiable metrics and advocates for a more transdisciplinary approach. By incorporating insights from social sciences, ethics, and public health, AI systems can address the compounded effects of intersecting identities and safeguard sensitive data. Additionally, explainability and interpretability must be central to AI design, fostering trust and accountability. This paper calls for a framework that balances accuracy with fairness, privacy, and inclusivity to ensure AI-driven diagnostics serve diverse populations equitably and ethically.

研究动机与目标

• 突出在AI驱动的医疗诊断中仅关注诊断准确性所带来的伦理局限。
• 主张在AI系统中纳入交叉性、健康社会决定因素以及更广泛的决定因素。
• 强调隐私、数据安全与治理，以防止在使用敏感属性时带来的危害。
• 倡导可解释性、透明度，以及参与式、跨学科发展。

提出的方法

• 批判性分析在AI诊断中优先考虑准确性而非公平性的立场。
• 提出一个整合社会科学、伦理学与公共卫生洞见的跨学科框架。
• 概述一个由四大支柱组成的框架：交叉性公平、健康决定因素整合、隐私与安全、跨学科协作。
• 推荐隐私保护型AI的方法学途径（如差分隐私、联邦学习等）。
• 倡导能够反映交叉性与真实世界健康决定因素的指标与评估。

Descriptio 1: Framework for an Equity-Centered and AI-Driven Diagnostics. This framework illustrates the four pillars essential for achieving equitable and inclusive AI-driven diagnostics. These include Intersectional Fairness, Determinants of Health Integration (social; environmental; behavioral an

实验结果

研究问题

• RQ1 交叉性与健康社会决定因素如何影响AI驱动诊断的公平性？
• RQ2 需要何种框架与治理结构以在AI医疗工具中平衡准确性、公平、隐私与包容性？
• RQ3 如何将可解释性和跨学科协作整合到AI开发生命周期以提升信任与公平性？

主要发现

• 认为将健康差距简化为离散的亚人群与指标会使不平等持续存在。
• 强调种族是一个没有一致生物学基础的社会构建，警示不要把种族作为诊断代理。
• 强调在利用敏感属性优化AI性能时的隐私与数据安全风险。
• 主张将生活方式、环境和结构性健康决定因素纳入AI系统，以反映真实世界的复杂性。
• 呼吁透明性、可解释性和治理，以确保对AI驱动诊断的问责与信任。
• 提出一个包含四大支柱的整合框架，以引导公平的AI诊断。