In Defence of Post-hoc Explainability

2024-12-23

arXiv.org

AI Summary

This paper addresses the epistemic legitimacy crisis confronting post-hoc explanation methods in scientific AI, arising from the fundamental tension between model complexity and human comprehensibility. Method: It introduces “Computational Interpretivism” (CI)—a novel philosophical framework that analogizes post-hoc rationalizations to expert human cognition, grounding scientific explanation in the normative concepts of *mediated understanding* and *bounded truthfulness*. Integrating analysis from philosophy of science, epistemic modeling, and empirical validation, the paper develops operational criteria for evaluating post-hoc explanations. Contribution/Results: It demonstrates that empirically warranted post-hoc explanations—without requiring full mechanistic transparency—can yield epistemically legitimate scientific insights, thereby bridging the gap between black-box models and scientific understanding. The core innovation lies in providing both a systematic philosophical justification for post-hoc interpretability and a practical, empirically grounded assessment pathway.

Technology Category

• Philosophy and Ethics of AI: Accountability, Interpretability & Explainability
• Humans and AI: Explainable AI (XAI) for Human Understanding
• Computer Vision: Interpretability, Explainability, and Transparency

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

The widespread adoption of machine learning in scientific research has created a fundamental tension between model opacity and scientific understanding. Whilst some advocate for intrinsically interpretable models, we introduce Computational Interpretabilism (CI) as a philosophical framework for post-hoc interpretability in scientific AI. Drawing parallels with human expertise, where post-hoc rationalisation coexists with reliable performance, CI establishes that scientific knowledge emerges through structured model interpretation when properly bounded by empirical validation. Through mediated understanding and bounded factivity, we demonstrate how post-hoc methods achieve epistemically justified insights without requiring complete mechanical transparency, resolving tensions between model complexity and scientific comprehension.

Problem

• Defending post-hoc explainability methods for scientific knowledge production
• Developing philosophical framework addressing reliability and epistemic criticisms
• Demonstrating how proper integration generates hypotheses and advances understanding

Innovation

• Defends post-hoc explainability for scientific knowledge
• Develops philosophical framework with mediated understanding
• Validates explanations through empirical testing methods