The Great Misalignment Problem in Human Evaluation of NLP Methods
This highlights a major issue affecting validity and reproducibility in NLP research, particularly for researchers relying on human evaluations.
The paper identifies the Great Misalignment Problem in NLP, where problem definitions, methods, and human evaluations are often misaligned, and finds that only 1 out of 10 surveyed ACL 2020 papers was fully aligned, with only 2 having evaluations aligned with methods.
We outline the Great Misalignment Problem in natural language processing research, this means simply that the problem definition is not in line with the method proposed and the human evaluation is not in line with the definition nor the method. We study this misalignment problem by surveying 10 randomly sampled papers published in ACL 2020 that report results with human evaluation. Our results show that only one paper was fully in line in terms of problem definition, method and evaluation. Only two papers presented a human evaluation that was in line with what was modeled in the method. These results highlight that the Great Misalignment Problem is a major one and it affects the validity and reproducibility of results obtained by a human evaluation.