From Automation to Collaboration: Human-in-the-Loop Methods for Safe and Trustworthy NLP
A new survey paper explores human-in-the-loop methods to enhance the safety and trustworthiness of Natural Language Processing (NLP) models. It highlights how human expertise is crucial for auditing, evaluating robustness, and constructing data for these models, especially in high-stakes applications. The paper identifies current limitations in scalable probing, sustainable benchmarks, and governance for private systems, proposing future research directions for adaptive auditing and accountable deployment. AI
IMPACT Highlights the necessity of human oversight for improving the safety and trustworthiness of deployed NLP models.