Guideline Bias in Wizard-of-Oz Dialogues

Victor Petrén Bach Hansen, Anders Søgaard


Abstract
NLP models struggle with generalization due to sampling and annotator bias. This paper focuses on a different kind of bias that has received very little attention: guideline bias, i.e., the bias introduced by how our annotator guidelines are formulated. We examine two recently introduced dialogue datasets, CCPE-M and Taskmaster-1, both collected by trained assistants in a Wizard-of-Oz set-up. For CCPE-M, we show how a simple lexical bias for the word like in the guidelines biases the data collection. This bias, in effect, leads to poor performance on data without this bias: a preference elicitation architecture based on BERT suffers a 5.3% absolute drop in performance, when like is replaced with a synonymous phrase, and a 13.2% drop in performance when evaluated on out-of-sample data. For Taskmaster-1, we show how the order in which instructions are resented, biases the data collection.
Anthology ID:
2021.bppf-1.2
Volume:
Proceedings of the 1st Workshop on Benchmarking: Past, Present and Future
Month:
Aug
Year:
2021
Address:
Online
Venues:
ACL | BPPF | IJCNLP
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
8–14
Language:
URL:
https://aclanthology.org/2021.bppf-1.2
DOI:
10.18653/v1/2021.bppf-1.2
Bibkey:
Copy Citation:
PDF:
https://aclanthology.org/2021.bppf-1.2.pdf
Code
 vpetren/guideline_bias
Data
CCPE-MTaskmaster-1