中文

面向保护基本权利的AI风险评估基础

人工智能 2025-07-25 v1

摘要

本章节引入一个用于定性风险评估AI的概念框架,特别是在 EU AI Act 的背景下。该 framework 解决了 legal compliance 和 fundamental rights protection 的复杂性,通过 integration definitional balancing 和 defeasible reasoning 来实现。definitional balancing 采用比例原则分析来解决 competing rights 之间的冲突,而 defeasible reasoning 则容纳了 legal decision-making 的 dynamic 性质。我们的 approach 强调需要分析 AI deployment scenarios 并识别潜在的 legal violations 和对 fundamental rights 的 multi-layered 影响。基于此 analysis,我们为 AI risk analysis 提供了哲学基础。特别是,我们考虑了 for conceptually grasping AI deployment scenarios 与 fundamental rights 之间 interaction 的 basic building blocks,incorporating in defeasible reasoning definitional balancing 和关于 rights 在 context 中 promotion 或 demotion 的 arguments。这种 layered approach 使得对 high-risk AI systems 和 General Purpose AI (GPAI) systems 的 assessment 模型更具可操作性,强调了后者的更广泛适用性。未来工作旨在 develop a formal model and effective algorithms 来 enhance AI risk assessment,桥接 theoretical insights 与 practical applications 以 support responsible AI governance。

关键词

引用

@article{arxiv.2507.18290,
  title  = {Foundations for Risk Assessment of AI in Protecting Fundamental Rights},
  author = {Antonino Rotolo and Beatrice Ferrigno and Jose Miguel Angel Garcia Godinez and Claudio Novelli and Giovanni Sartor},
  journal= {arXiv preprint arXiv:2507.18290},
  year   = {2025}
}

备注

24 pages, 1 figure. To be published in: The Philosophical Foundations of Information Technology Law. Oxford University Press, Oxford