We Need to Consider Disagreement in Evaluation
Open accessThe DOI matches its Crossref record; 10 of the 10 references it cites were found in a registry or an index.
Abstract
Evaluation is of paramount importance in datadriven research fields such as Natural Language Processing (NLP) and Computer Vision (CV).But current evaluation practice in NLP, except for end-to-end tasks such as machine translation, spoken dialogue systems, or NLG, largely hinges on the existence of a single "ground truth" against which we can meaningfully compare the prediction of a model.However, this assumption is flawed for two reasons.1) In many cases, more than one answer is correct.2) Even where there is a single answer, disagreement among annotators is ubiquitous, making it difficult to decide on a gold standard.We discuss three sources of disagreement: from the annotator, the data, and the context, and show how this affects even seemingly objective tasks.Current methods of adjudication, agreement, and evaluation ought to be reconsidered at the light of this evidence.Some researchers now propose to address this issue by minimizing disagreement, creating cleaner datasets.We argue that such a simplification is likely to result in oversimplified models just as much as it would do for end-to-end tasks such as machine translation.Instead, we suggest that we need to improve today's evaluation practice to better capture such disagreement.Datasets with multiple annotations are becoming more common, as are methods to integrate disagreement into modeling.The logical next step is to extend this to evaluation.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
LabelMe: A Database and Web-Based Tool for Image Annotation
Russell, B. C., Torralba, A., Murphy, K. P., & Freeman, W. T. (2008) · International Journal of Computer Vision · doi:10.1007/s11263-007-0090-8 · Cited by 4,210
SpecAugment: A Simple Data Augmentation Method for Automatic Speech Recognition
Park, D. S., Chan, W., Zhang, Y., Chiu, C.-C., Zoph, B., Cubuk, E. D., & Le, Q. V. (2019) · Interspeech 2019 · doi:10.21437/interspeech.2019-2680 · Cited by 3,558
Making the V in VQA Matter: Elevating the Role of Image Understanding in Visual Question Answering
Goyal, Y., Khot, T., Summers-Stay, D., Batra, D., & Parikh, D. (2017) · 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2017.670 · Cited by 2,287
Inter-Coder Agreement for Computational Linguistics
Artstein, R., & Poesio, M. (2008) · Computational Linguistics · doi:10.1162/coli.07-034-r2 · Cited by 1,665 · cited 2×
“Disagreement and task dif-ficulty and subjectivity also challenge traditional agreement measures (Artstein and Poesio, 2008).”
BackgroundSelf-citationThe Benefits of a Model of Annotation
Passonneau, R. J., & Carpenter, B. (2014) · Transactions of the Association for Computational Linguistics · doi:10.1162/tacl_a_00185 · Cited by 227
Inherent Disagreements in Human Textual Inferences
Pavlick, E., & Kwiatkowski, T. (2019) · Transactions of the Association for Computational Linguistics · doi:10.1162/tacl_a_00293 · Cited by 214
Best practices for the human evaluation of automatically generated text
van der Lee, C., Gatt, A., van Miltenburg, E., Wubben, S., & Krahmer, E. (2019) · Proceedings of the 12th International Conference on Natural Language Generation · doi:10.18653/v1/w19-8643 · Cited by 205 · cited 1×
“Variability in the annotation is a feature of many such tasks (see, e.g., van der Lee et al. (2019) for agreement issues in generated text evaluation) even though many corpora still may come with single references due to data collection costs.”
BackgroundLinguistically debatable or just plain wrong?
Plank, B., Hovy, D., & Søgaard, A. (2014) · Proceedings of the 52nd Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers) · doi:10.3115/v1/p14-2083 · Cited by 86
Self-citationTowards Measuring Fairness in AI: The Casual Conversations Dataset
Hazirbas, C., Bitton, J., Dolhansky, B., Pan, J., Gordo, A., & Ferrer, C. C. (2022) · IEEE Transactions on Biometrics, Behavior, and Identity Science · doi:10.1109/tbiom.2021.3132237 · Cited by 79
Designing an Online Infrastructure for Collecting AI Data From People With Disabilities
Park, J. S., Bragg, D., Kamar, E., & Morris, M. R. (2021) · Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency · doi:10.1145/3442188.3445870 · Cited by 50
Cited by
The “Problem” of Human Label Variation: On Ground Truth in Data, Modeling and Evaluation
Plank, B. (2022) · Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing · doi:10.18653/v1/2022.emnlp-main.731 · Cited by 121
Self-citationHashemi, H., Eisner, J., Rosset, C., Van Durme, B., & Kedzie, C. (2024) · Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · doi:10.18653/v1/2024.acl-long.745 · Cited by 47
“It is increasingly acknowledged that accounting for subjectivity (as opposed to collapsing or removing disagreements) in NLP evaluation is a key part of evaluation design (Pavlick and Kwiatkowski, 2019; Basile et al., 2021; Uma et al., 2021a,b; Plank, 2022; Plepi et al., 2022; Sandri et al., 2023).”
Human-centered neural reasoning for subjective content processing: Hate speech, emotions, and humor
Kazienko, P., Bielaniewicz, J., Gruza, M., Kanclerz, K., Karanowski, K., Miłkowski, P., & Kocoń, J. (2023) · Information Fusion · doi:10.1016/j.inffus.2023.01.010 · Cited by 38
SafetyKit: First Aid for Measuring Safety in Open-domain Conversational Systems
Dinan, E., Abercrombie, G., Bergman, A., Spruit, S., Hovy, D., Boureau, Y.-L., & Rieser, V. (2022) · Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · doi:10.18653/v1/2022.acl-long.284 · Cited by 32
Self-citationWhy Don’t You Do It Right? Analysing Annotators’ Disagreement in Subjective Tasks
Sandri, M., Leonardelli, E., Tonelli, S., & Jezek, E. (2023) · Proceedings of the 17th Conference of the European Chapter of the Association for Computational Linguistics · doi:10.18653/v1/2023.eacl-main.178 · Cited by 31
Metadata
- Authors
- Basile, V., Fell, M., Fornaciari, T., Hovy, D., Paun, S., Plank, B., Poesio, M., & Uma, A.
- Venue
- Proceedings of the 1st Workshop on Benchmarking: Past, Present and Future
- Published
- 2021
- Pages
- 15-21
- OpenAlex
- W3184904351
- Cited by
- 73
- References
- 10
- FWCI
- 7.40
- Type
- conference paper
- Language
- en
- Open access
- gold
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
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: We Need to Consider Disagreement in Evaluation. https://app.cochat.ai/papers/10.18653/v1/2021.bppf-1.3
- The paper
- Basile, V., Fell, M., Fornaciari, T., Hovy, D., Paun, S., Plank, B., Poesio, M., & Uma, A. (2021). We Need to Consider Disagreement in Evaluation. Proceedings of the 1st Workshop on Benchmarking: Past, Present and Future. https://doi.org/10.18653/v1/2021.bppf-1.3
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.