From Plausible to Actionable: A Position on LLM Self-Explanations

2026-07-17

AI Summary

Current large language models (LLMs) often generate self-explanations that appear plausible on the surface, yet it remains unclear whether these explanations genuinely reflect their internal reasoning processes. Moreover, existing evaluation methods largely overlook practical utility. This work systematically disentangles and interrelates the plausibility, faithfulness, and actionability of LLM self-explanations, proposing an action-oriented evaluation framework that moves beyond reliance on faithfulness alone. Grounded in explainable AI (XAI) theory and integrating qualitative analysis with tailored evaluation criteria, the study redefines the assessment paradigm for LLM explanations by emphasizing their capacity to support informed decision-making among diverse stakeholders. The proposed practical guidelines advance XAI from merely “seeming reasonable” toward being “truly useful,” offering both theoretical grounding and actionable pathways for deploying trustworthy AI systems.

Abstract

Large Language Models (LLMs) can generate natural language explanations that rationalize their own decisions, a phenomenon commonly referred to as self-explanations.Such explanations have emerged as a promising direction for explainable artificial intelligence (XAI), particularly for interpreting LLM behavior.However, while self-explanations often appear plausible, whether they faithfully reflect a model's underlying reasoning process remains an open question. In this opinion paper, we argue that self-explanations can be highly plausible, questionably faithful, and yet highly actionable. From a traditional XAI perspective, we identify the limitations of standard evaluation protocols for LLM-generated self-explanations and propose practical guidelines for assessing their plausibility and faithfulness. Moreover, we argue that evaluation should extend beyond these criteria to actionability, highlighting applications of LLM rationalization capabilities that support informed decision-making and appropriate action across diverse stakeholders.

Problem

• self-explanations
• explainable artificial intelligence
• faithfulness
• plausibility
• actionability

Innovation

• self-explanations
• actionability
• faithfulness
• plausibility
• explainable AI