CoChat

Self-Diagnosis and Self-Debiasing: A Proposal for Reducing Corpus-Based Bias in NLP

Open access
Schick, Timo, Udupa, Sahana, Schütze, Hinrich · 2021 · Transactions of the Association for Computational Linguistics

The DOI matches its Crossref record; 59 of the 59 references it cites were found in a registry or an index.

Abstract

⚠ This paper contains prompts and model outputs that are offensive in nature. When trained on large, unfiltered crawls from the Internet, language models pick up and reproduce all kinds of undesirable biases that can be found in the data: They often generate racist, sexist, violent, or otherwise toxic language. As large models require millions of training examples to achieve good performance, it is difficult to completely prevent them from being exposed to such content. In this paper, we first demonstrate a surprising finding: Pretrained language models recognize, to a considerable degree, their undesirable biases and the toxicity of the content they produce. We refer to this capability as self-diagnosis. Based on this finding, we then propose a decoding algorithm that, given only a textual description of the undesired behavior, reduces the probability of a language model producing problematic text. We refer to this approach as self-debiasing. Self-debiasing does not rely on manually curated word lists, nor does it require any training data or changes to the model’s parameters. While we by no means eliminate the issue of language models generating biased text, we believe our approach to be an important step in this direction.1

Overview

Integrity signals

Registry

Crossref · 10.1162/tacl_a_00434

Cites

59 of 59 references verified0 citation errors · 0 not checkable

Statistics

Recomputing p-values needs the paper’s full text.

Verification

Claims the paper makes

We couldn’t retrieve this paper’s full text, so its claims and self-check weren’t run. The registry and citation checks still apply.

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Citations

Cites

Research works printed but not linked: 1

  1. Wang, 2019, BERT has a mouth, and it must speak: BERT as a Markov random field language model, 30, 10.18653/v1/W19-2304 · doi:10.18653/v1/w19-2304

Cited by

See all 237

Metadata

Authors
Schick, T., Udupa, S., & Schütze, H.
Affiliations
Ludwig-Maximilians-Universität München (DE)
Corresponding author
  • Schick, Timo
  • Udupa, Sahana
  • Schütze, Hinrich
Venue
Transactions of the Association for Computational Linguistics
Publisher
MIT Press
Published
2021
Volume
9
Pages
1408-1424
ISSN
2307-387X (ISSN-L)
OpenAlex
W3134354193
Cited by
227
References
58
Citations by year
2026: 29 · 2025: 40 · 2024: 36 · 2023: 70 · 2022: 36 · 2021: 16
FWCI
13.96
Type
journal article
Language
en
Open access
Locations
  • Transactions of the Association for Computational Linguistics (publishedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
  • Open access LMU (Ludwid Maxmilian's Universitat Munchen) (submittedVersion, cc-by)
  • Universitätsbibliothek der LMU
Topics
  • Physical Sciences › Computer Science › Artificial Intelligence › Hate Speech and Cyberbullying Detection
  • Physical Sciences › Computer Science › Artificial Intelligence › Ethics and Social Impacts of AI
  • Physical Sciences › Computer Science › Artificial Intelligence › Topic Modeling
Keywords
self-diagnosis, bias in language models, toxicity, bias detection, decoding algorithms, large language models

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.