CoChat

The Disagreement Deconvolution: Bringing Machine Learning Performance Metrics In Line With Reality

Gordon, Mitchell L., Zhou, Kaitlyn, Patel, Kayur, Hashimoto, Tatsunori, Bernstein, Michael S. · 2021 · Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems

The DOI matches its Crossref record; 73 of the 73 references it cites were found in a registry or an index.

Abstract

Machine learning classifiers for human-facing tasks such as comment toxicity and misinformation often score highly on metrics such as ROC AUC but are received poorly in practice. Why this gap? Today, metrics such as ROC AUC, precision, and recall are used to measure technical performance; however, human-computer interaction observes that evaluation of human-facing systems should account for people’s reactions to the system. In this paper, we introduce a transformation that more closely aligns machine learning classification metrics with the values and methods of user-facing performance measures. The disagreement deconvolution takes in any multi-annotator (e.g., crowdsourced) dataset, disentangles stable opinions from noise by estimating intra-annotator consistency, and compares each test set prediction to the individual stable opinions from each annotator. Applying the disagreement deconvolution to existing social computing datasets, we find that current metrics dramatically overstate the performance of many human-facing machine learning tasks: for example, performance on a comment toxicity task is corrected from .95 to .73 ROC AUC.

Overview

Integrity signals

Registry

Crossref · 10.1145/3411764.3445423

Cites

73 of 73 references verified0 citation errors · 0 not checkable

Statistics

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

Verification

Claims the paper makes

This paper’s full text isn’t openly available, so its claims and self-check weren’t run. The registry and citation checks still apply.

Citations

Cites

Research works printed but not linked: 4

  1. Paul M Barrett. 2020. Who Moderates the Social Media Giants?Center for Business (2020).
  2. Geoffery C. Bowker and Susan Leigh Star. 2000. Sorting Things out: Classification and Its Consequences. MIT Press, Cambridge, MA, USA.
  3. Matt Gardner Yoav Artzi Victoria Basmova Jonathan Berant Ben Bogin Sihao Chen Pradeep Dasigi Dheeru Dua Yanai Elazar Ananth Gottumukkala 2020. Evaluating nlp models via contrast sets. (2020). arXiv:2004.02709 · arXiv
  4. Danny Pfeffermann and Calyampudi Radhakrishna Rao. 2009. Sample surveys: design, methods and applications. Elsevier.

Cited by

  • Explanations Can Reduce Overreliance on AI Systems During Decision-Making

    Vasconcelos, H., Jörke, M., Grunde-McLaughlin, M., Gerstenberg, T., Bernstein, M. S., & Krishna, R. (2023) · Proceedings of the ACM on Human-Computer Interaction · doi:10.1145/3579605 · Cited by 402

    “(4) Uncertain environments: Some decision-making tasks, such as toxicity prediction, lack a “correct” answer [29].”
    BackgroundSelf-citation
  • Dealing with Disagreements: Looking Beyond the Majority Vote in Subjective Annotations

    Davani, A. M., Díaz, M., & Prabhakaran, V. (2022) · Transactions of the Association for Computational Linguistics · doi:10.1162/tacl_a_00449 · Cited by 219

    “Gordon et al. (2021) explore multiple annotators judgements to disentangle stable opinions from noise by estimating intra-annotator consistency.”
    Background
  • Trauma-Informed Computing: Towards Safer Technology Experiences for All

    Chen, J. X., McDonald, A., Zou, Y., Tseng, E., Roundy, K. A., Tamersoy, A., Schaub, F., Ristenpart, T., & Dell, N. (2022) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3491102.3517475 · Cited by 163

    “[64] developed the disagreement deconvolution, a method that aligns ML classification metrics with “the proportion of the overall population that would agree with each classification decision,” thereby providing a more human-centered approach to ML classifiers in line with the collaboration principle.”
    Method
  • Jury Learning: Integrating Dissenting Voices into Machine Learning Models

    Gordon, M. L., Lam, M. S., Park, J. S., Patel, K., Hancock, J., Hashimoto, T., & Bernstein, M. S. (2022) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3491102.3502004 · Cited by 132

    “While this majoritarian [55] procedure succeeded for many early machine learning tasks, it now runs aground on tasks where there is substantial disagreement on what the correct label ought to be [37].”
    BackgroundInfluentialSelf-citation
  • Building Human Values into Recommender Systems: An Interdisciplinary Synthesis

    Stray, J., Halevy, A., Assar, P., Hadfield-Menell, D., Boutilier, C., Ashar, A., Bakalar, C., Beattie, L., Ekstrand, M., Leibowicz, C., et al. (2024) · ACM Transactions on Recommender Systems · doi:10.1145/3632297 · Cited by 122

    “Also, many of these categories are both complex and politically contested (Bozarth et al., 2020; Martinez and Paluck, 2020) and raters often disagree, which complicates the evaluation of model accuracy (Gordon et al., 2021).”
    Background

See all 107

Metadata

Authors
Gordon, M. L., Zhou, K., Patel, K., Hashimoto, T., & Bernstein, M. S.
Affiliations
  • Stanford University (US)
  • Apple (United States) (US)
Venue
Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems
Publisher
ACM
Published
2021-05-06
Pages
1-14
OpenAlex
W3163078977
Cited by
101
References
67
Citations by year
2026: 5 · 2025: 16 · 2024: 19 · 2023: 27 · 2022: 23 · 2021: 11
FWCI
9.24
Type
conference paper
Language
en
Open access
closed
Locations
CHI Conference on Human Factors in Computing Systems (publishedVersion)
Topics
  • Physical Sciences › Computer Science › Artificial Intelligence › Hate Speech and Cyberbullying Detection
  • Physical Sciences › Computer Science › Artificial Intelligence › Machine Learning and Data Classification
  • Physical Sciences › Computer Science › Computer Science Applications › Mobile Crowdsensing and Crowdsourcing
Keywords
classifier evaluation metrics, crowdsourced annotation, misinformation detection

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.