What if you said that differently?: How Explanation Formats Affect Human Feedback Efficacy and User Perception

arXiv:2311.09558v2 Announce Type: replace
Abstract: Eliciting feedback from end users of NLP models can be beneficial for improving models. However, how should we present model responses to users so they are most amenable to be corrected from user feedback? Further, what properties do users value to understand and trust responses? We answer these questions by analyzing the effect of rationales (or explanations) generated by QA models to support their answers. We specifically consider decomposed QA models that first extract an …

abstract arxiv cs.cl end users feedback however human human feedback improving nlp nlp models perception responses said type user feedback