Knowledge Transfer, Knowledge Gaps, and Knowledge Silos in Citation Networks

XAI research shows limited knowledge transfer from foundational fields: The study found that explainable AI research draws surprisingly little from established foundational topics in machine learning, statistics, and psychology, despite these being core to understanding the field. This suggests researchers may be reinventing solutions rather than building on existing knowledge.

Application domains operate in isolation ("knowledge silos"): Certain practical applications of XAI exist as separate research communities with minimal cross-pollination. This fragmentation means solutions developed in one domain (e.g., healthcare) may not reach other domains (e.g., finance) where they could be valuable, potentially leading to duplicated efforts and inconsistent approaches to responsible AI.

Significant gaps exist between related XAI research areas: The framework identified "knowledge gaps" between research topics that should logically connect. These gaps represent missed opportunities where insights from one XAI subfield could improve practices in another, suggesting the field lacks sufficient integration for coherent responsible AI standards.

Interdisciplinary integration remains a challenge in emerging AI fields: The study demonstrates that even in a field explicitly designed to be interdisciplinary, knowledge flows between disciplines are limited and uneven. This has implications for responsible AI development, which requires genuine collaboration across technical and social science perspectives.

Citation network analysis can reveal structural problems in research fields: By mapping how knowledge actually flows (rather than assuming it does), this framework provides a diagnostic tool for identifying where research communities need better communication and integration to advance responsible AI practices.