[论文解读] Trustworthy AI: From Principles to Practices

Bo Li, Peng Qi+6 more|arXiv (Cornell University)|Oct 4, 2021

Ethics and Social Impacts of AI被引用 24

一句话总结

本综述提出一个系统框架，用于在整个生命周期构建值得信赖的AI，涵盖鲁棒性、泛化、可解释性、透明性、可重复性、公平性、隐私与问责性，以及实用指南和权衡。

ABSTRACT

The rapid development of Artificial Intelligence (AI) technology has enabled the deployment of various systems based on it. However, many current AI systems are found vulnerable to imperceptible attacks, biased against underrepresented groups, lacking in user privacy protection. These shortcomings degrade user experience and erode people's trust in all AI systems. In this review, we provide AI practitioners with a comprehensive guide for building trustworthy AI systems. We first introduce the theoretical framework of important aspects of AI trustworthiness, including robustness, generalization, explainability, transparency, reproducibility, fairness, privacy preservation, and accountability. To unify currently available but fragmented approaches toward trustworthy AI, we organize them in a systematic approach that considers the entire lifecycle of AI systems, ranging from data acquisition to model development, to system development and deployment, finally to continuous monitoring and governance. In this framework, we offer concrete action items for practitioners and societal stakeholders (e.g., researchers, engineers, and regulators) to improve AI trustworthiness. Finally, we identify key opportunities and challenges for the future development of trustworthy AI systems, where we identify the need for a paradigm shift toward comprehensively trustworthy AI systems.

研究动机与目标

• 提出一个从数据到部署到治理的整体生命周期框架，以提升AI的可信度。
• 剖析在实际场景中多种可信性要素如何相互作用与权衡。
• 为研究人员、工程师、监管者和利益相关者提供具体行动项。
• 识别未来值得信赖AI发展的关键挑战与机遇。

提出的方法

• 在从数据采集到部署和监控的生命周期管线中系统性地组织多学科方法。
• 分析鲁棒性、泛化、可解释性、透明性、可重复性、公平性、隐私和问责性之间的关系与互动。
• 调查每个可信性要素的现有技术与评估。
• 提出跨生命周期阶段的持续工作流与反馈整合。

实验结果

研究问题

• RQ1 AI可信性的关键要素有哪些，它们如何相互作用？
• RQ2 如何在AI开发生命周期中系统性地集成可信性？
• RQ3 鲁棒性、泛化、可解释性、透明性、可重复性、公平性、隐私和问责性的当前方法与评估有哪些？
• RQ4 哪些挑战与机遇将塑造未来值得信赖的AI实践？

主要发现

• 需要一个全面的生命周期框架，在数据、模型、部署和治理各阶段提升AI的可信性。
• 在现实系统中，可信性要素之间存在相互提升和权衡。
• 可解释性和透明性需要设计时和事后方法，以及定性与以人为本的评估。
• 可重复性日益受到AI研究与部署社区的关注，涉及数据、方法和实验。
• 公平性需要解决数据、模型和过程偏见，并认识多种公平概念（如独立性、分离性、充足性）。
• 未来值得信赖的AI需要更深的理论理解和跨学科合作。