Total: 1
The field of explainable Automatic Fact-Checking (AFC) aims to enhance the transparency and trustworthiness of automated fact-verification systems by providing clear and comprehensible explanations. However, the effectiveness of these explanations depends on their actionability --their ability to empower users to make informed decisions and mitigate misinformation. Despite actionability being a critical property of high-quality explanations, no prior research has proposed a dedicated method to evaluate it. This paper introduces FinGrAct, a fine-grained evaluation framework that can access the web, and it is designed to assess actionability in AFC explanations through well-defined criteria and an evaluation dataset. FinGrAct surpasses state-of-the-art (SOTA) evaluators, achieving the highest Pearson and Kendall correlation with human judgments while demonstrating the lowest ego-centric bias, making it a more robust evaluation approach for actionability evaluation in AFC.