带必要回溯的自然反事实
Natural Counterfactuals With Necessary Backtracking
AI总结：
针对朱迪亚·珀尔反事实生成方法与观测场景偏差过大、可行性不足的问题，提出自然反事实框架与带可控回溯的优化方法，生成更贴合真实数据分布的反事实，经实验验证有效并开源了代码。
摘要：
反事实推理是人类认知的关键环节，在提供解释与决策制定方面尤为重要。朱迪亚·珀尔（Judea Pearl）颇具影响力的方法虽在理论上十分精妙，但我们通过简单示例表明，其生成的反事实场景往往与观测场景偏差过大，难以具备可行性。为缓解这一难题，我们提出natural counterfactuals（自然反事实）框架及一种反事实生成方法，所生成的反事实相对于实际数据分布更具可行性。我们的方法会在必要时引入一定程度的回溯，允许对因果前置变量进行调整，以尽可能缩小与真实场景的偏差。具体而言，我们提出了一种全新的优化框架，该框架允许回溯，同时以自然性准则控制回溯程度。实证实验验证了我们方法的有效性。相关代码可在https://github.com/GuangyuanHao/natural_counterfactuals获取。
英文摘要：
Counterfactual reasoning is pivotal in human cognition and especially important for providing explanations and making decisions. While Judea Pearl's influential approach is theoretically elegant, its generation of a counterfactual scenario often requires too much deviation from the observed scenarios to be feasible, as we show using simple examples. To mitigate this difficulty, we propose a framework of \emph{natural counterfactuals} and a method for generating counterfactuals that are more feasible with respect to the actual data distribution. Our methodology incorporates a certain amount of backtracking when needed, allowing changes in causally preceding variables to minimize deviations from realistic scenarios. Specifically, we introduce a novel optimization framework that permits but also controls the extent of backtracking with a naturalness criterion. Empirical experiments demonstrate the effectiveness of our method. The code is available at https://github.com/GuangyuanHao/natural_counterfactuals.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
