All That’s ‘Human’ Is Not Gold: Evaluating Human Evaluation of Generated Text

Elizabeth Clark, Tal August, Sofia Serrano, Nikita Haduong, Suchin Gururangan, Noah A. Smith


Abstract
Human evaluations are typically considered the gold standard in natural language generation, but as models’ fluency improves, how well can evaluators detect and judge machine-generated text? We run a study assessing non-experts’ ability to distinguish between human- and machine-authored text (GPT2 and GPT3) in three domains (stories, news articles, and recipes). We find that, without training, evaluators distinguished between GPT3- and human-authored text at random chance level. We explore three approaches for quickly training evaluators to better identify GPT3-authored text (detailed instructions, annotated examples, and paired examples) and find that while evaluators’ accuracy improved up to 55%, it did not significantly improve across the three domains. Given the inconsistent results across text domains and the often contradictory reasons evaluators gave for their judgments, we examine the role untrained human evaluations play in NLG evaluation and provide recommendations to NLG researchers for improving human evaluations of text generated from state-of-the-art models.
Anthology ID:
2021.acl-long.565
Volume:
Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers)
Month:
August
Year:
2021
Address:
Online
Venues:
ACL | IJCNLP
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
7282–7296
Language:
URL:
https://aclanthology.org/2021.acl-long.565
DOI:
10.18653/v1/2021.acl-long.565
Award:
 Outstanding Paper
Bibkey:
Copy Citation:
PDF:
https://aclanthology.org/2021.acl-long.565.pdf
Optional supplementary material:
 2021.acl-long.565.OptionalSupplementaryMaterial.zip