Learning to love diligent trolls: Accounting for rater effects in the dialogue safety task

Michael Ilagan


Abstract
Chatbots have the risk of generating offensive utterances, which must be avoided. Post-deployment, one way for a chatbot to continuously improve is to source utterance/label pairs from feedback by live users. However, among users are trolls, who provide training examples with incorrect labels. To de-troll training data, previous work removed training examples that have high user-aggregated cross-validation (CV) error. However, CV is expensive; and in a coordinated attack, CV may be overwhelmed by trolls in number and in consistency among themselves. In the present work, I address both limitations by proposing a solution inspired by methodology in automated essay scoring (AES): have multiple users rate each utterance, then perform latent class analysis (LCA) to infer correct labels. As it does not require GPU computations, LCA is inexpensive. In experiments, I found that the AES-like solution can infer training labels with high accuracy when trolls are consistent, even when trolls are the majority.
Anthology ID:
2023.findings-emnlp.928
Volume:
Findings of the Association for Computational Linguistics: EMNLP 2023
Month:
December
Year:
2023
Address:
Singapore
Editors:
Houda Bouamor, Juan Pino, Kalika Bali
Venue:
Findings
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
13893–13899
Language:
URL:
https://aclanthology.org/2023.findings-emnlp.928
DOI:
10.18653/v1/2023.findings-emnlp.928
Bibkey:
Cite (ACL):
Michael Ilagan. 2023. Learning to love diligent trolls: Accounting for rater effects in the dialogue safety task. In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 13893–13899, Singapore. Association for Computational Linguistics.
Cite (Informal):
Learning to love diligent trolls: Accounting for rater effects in the dialogue safety task (Ilagan, Findings 2023)
Copy Citation:
PDF:
https://aclanthology.org/2023.findings-emnlp.928.pdf