Adoption of Explainable Natural Language Processing: Perspectives from Industry and Academia on Practices and Challenges

There's a significant gap between explainability theory and practice: The study found "conceptual gaps" between what academic researchers propose and what industry practitioners actually need, suggesting that current explainability methods may not be designed with real-world deployment in mind.

Practitioners are dissatisfied with existing explainability tools: The research reveals "low satisfaction with current explainability methods," indicating that the transparency solutions available today are not adequately meeting the needs of those trying to deploy NLP systems in practice.

Evaluation of explainability remains a major unsolved challenge: The paper highlights "evaluation challenges" as a key finding, meaning there's currently no clear consensus on how to measure whether an explanation is actually helpful or trustworthy—a critical issue for high-stakes applications like healthcare or criminal justice.

User-centered design is missing from explainability development: The authors emphasize "the need for clear definitions and user-centric frameworks," suggesting that explainability methods have been developed without adequately considering who needs to understand the explanations and what those specific users actually require.

Explainability adoption is motivated by compliance and risk management, not just transparency: While the abstract doesn't detail specific motivations, the emphasis on "high-stakes environments" and "deployment" suggests practitioners adopt explainability primarily to meet regulatory requirements and reduce liability rather than to genuinely understand how their systems work.