Levels of explainable artificial intelligence for human-aligned conversational explanations

Current XAI systems focus too narrowly on individual decisions: Most explainable AI applications provide only "low-level" explanations of how a specific decision was made for a particular case, but this limited approach fails to address what people actually need to understand and trust AI systems—including the AI's underlying beliefs, motivations, and reasoning processes.

People need deeper explanations to accept AI decisions: The paper argues that truly trustworthy AI explanations must go beyond technical details to address factors like an AI agent's beliefs and motivations, its understanding of cultural expectations, and how it generated its own explanation—all of which are "essential to providing the explanatory depth that people require."

A multi-level explanation framework is necessary: Rather than single-layer explanations, the paper proposes integrating multiple levels of explanation into conversational systems that can provide more comprehensive, human-aligned responses—moving from narrow, technical explanations toward what the authors call "strong" explanations.

Legislative and public pressure is driving XAI development: The growth in XAI research is being fueled by recent legal requirements, government and industry investment, and increasing public concern about autonomous decisions affecting their lives, indicating that explainability is becoming a regulatory and social priority.