零重要性：为可解释机器学习解耦相关性
Null importance: Disentangling relevance for interpretable machine learning
AI总结：
本文提出零重要性统一框架，区分特征重要性的多种相关性概念，连接科学问题、假设与算法，并通过公平性、基因组扰动及模拟实验验证其理论区分与实际影响。
AI中文摘要：
特征重要性是可解释机器学习的核心，但“重要性”一词涵盖了多种根本不同的相关性概念。我们基于零重要性发展了一个统一视角：即在特定相关性概念下，特征何时不相关的一种总体层面刻画。我们考虑了源于边际和条件统计相关性、预测风险、函数不变性以及因果效应的标准零重要性概念，并展示了这些概念如何回答不同的科学问题。我们在两个区分尤为重要的应用中阐述了该框架：算法公平性，其中常见的公平性标准对应于不同的零重要性概念；以及基因组扰动建模，其中不同的相关性概念会导致关于预测模型学到了什么的不同结论。该框架连接了特征分析的三个方面：定义相关性的科学问题、塑造不同零概念之间关系的数据和模型假设，以及用于评估重要性的方法。我们建立了零概念一致时的充分条件，并给出了当这些条件不成立时它们如何分化的反例。接着，我们刻画了不同方法族所针对的零概念，以及它们的零重要性统计量何时能识别这些目标。最后，涵盖特征依赖性、冗余性、非线性、隐藏特征及其他标准现象的模拟实验，以及对图像和多组学数据的案例研究，为这些理论区分及其实际后果提供了经验证据。综合来看，这些结果为关联科学问题、数据生成假设和算法提供了一种共同的统计语言，并阐明了特征重要性分析所能支持的结论。
英文摘要：
Feature importance is central to interpretable machine learning, but the term "importance" encompasses several fundamentally different notions of relevance. We develop a unified perspective based on null importance: a population-level characterization of when a feature is irrelevant under a specified notion of relevance. We consider standard notions of null importance arising from marginal and conditional statistical relevance, predictive risk, functional invariance, and causal effects, and show how these notions answer different scientific questions. We illustrate the framework in two applications in which the distinction is particularly consequential: algorithmic fairness, where common fairness criteria correspond to different notions of null importance, and genomic perturbation modeling, where different notions of relevance lead to different conclusions about what a prediction model has learned. The framework connects three aspects of feature analysis: the scientific question defining relevance, the data and model assumptions that shape how different null notions relate, and the methods used to assess importance. We establish sufficient conditions under which null notions coincide and give counterexamples showing how they diverge when those conditions fail. We then characterize which nulls different method families target and when their zero-importance statistics identify those targets. Finally, simulations spanning feature dependence, redundancy, nonlinearity, hidden features and other standard phenomena, along with case studies on image and multiomics data, provide empirical evidence for these theoretical distinctions and their practical consequences. Taken together, these results provide a common statistical language for relating scientific questions, data-generating assumptions, and algorithms, and clarify the conclusions that feature-importance analyses can support.
