Evaluating Explainability in Machine Learning Predictions Through Explainer-Agnostic Metrics
Keywords: Explainable AI, Explainability Metrics, Machine Learning, Fairness
Abstract: Artificial intelligence (AI) continues to transform industries and research at an accelerated pace, bringing forth numerous challenges related to transparency and accountability in AI-driven decision-making. Decision-makers and stakeholders require not only a clear understanding of how these systems generate predictions but also assurance that these processes are conducted ethically and responsibly. These challenges highlight a critical need for effective tools to evaluate and enhance the interpretability of AI models. To address this gap, we propose a new set of explainer-sensitive metrics aimed at evaluating the interpretability of AI models in the context of specific explainers. By focusing on global and local feature importance, as well as surrogate models, our metrics capture key elements such as feature stability, fluctuations in prediction behavior, and contrasts in feature relevance across conditional subsets. By quantifying these complex dynamics as clear scalar measures, we offer a structured framework for assessing model transparency, fairness, and robustness. We demonstrate the practical utility of our approach through case studies on a set of benchmark datasets, revealing deeper insights into model interpretability that facilitate more informed decision-making among AI developers and stakeholders. Ultimately, our work aims to foster AI systems that are not only technically reliable but also transparent, fair, and accountable, thereby advancing the development of ethical AI practices.
Primary Area: interpretability and explainable AI
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Resubmission: No
Student Author: No
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Large Language Models: Yes, at the sentence level (e.g., fixing grammar, re-wording sentences)
Submission Number: 12656
Loading