Mitigating Source Bias for Fairer Weak Supervision

Weak supervision enables efficient development of training sets by reducing the need for ground truth labels. However, the techniques that make weak supervision attractive -- such as integrating any source of signal to estimate unknown labels -- also entail the danger that the produced pseudolabels are highly biased. Surprisingly, given everyday use and the potential for increased bias, weak supervision has not been studied from the point of view of fairness. We begin such a study, starting with the observation that even when a fair model can be built from a dataset with access to ground-truth labels, the corresponding dataset labeled via weak supervision can be arbitrarily unfair. To address this, we propose and empirically validate a model for source unfairness in weak supervision, then introduce a simple counterfactual fairness-based technique that can mitigate these biases. Theoretically, we show that it is possible for our approach to simultaneously improve both accuracy and fairness -- in contrast to standard fairness approaches that suffer from tradeoffs. Empirically, we show that our technique improves accuracy on weak supervision baselines by as much as 32\% while reducing demographic parity gap by 82.5\%. A simple extension of our method aimed at maximizing performance produces state-of-the-art performance in five out of ten datasets in the WRENCH benchmark.

Paper

References (72)

Scroll for more · 38 remaining

Similar papers

Peer review

Reviewer 6e6W7/10 · confidence 4/52023-06-12

Summary

This work found that unfair LFs in programmatic weak supervision could introduce bias to the resultant training labels, and proposed to address the bias via source bias mitigation and provided theoretical guarantee. Experimental results show that the effectiveness of the approaches in both synthetic and real datasets.

Strengths

1. This work studies an important yet overlooked problem in programmatic weak supervision: the biases induced by unfair LFs 2. The proposed method, compatible with traditional fair ML methods, could mitigate biases and improve the performance at the same time. 3. The theoretical results are convincing, which show that the LF bias could be arbitrary yet can be fixed by the proposed method under some conditions

Weaknesses

I am not aware of any major weakness except that the proposed model, if I understand it correctly, is a new label model built on an existing one, which means users have to use the proposed label model if they want to mitigate the biases. It is unclear how biases should be mitigated if users prefer other choice of label model. In terms of label models that incorporate feature vector, one related work is missing: "Leveraging instance features for label aggregation in programmatic weak supervision"

Questions

It is not intuitive how the improvement of fairness and performance can be achieved at the same time. Usually, there is a trade-off between performance and fairness in ML models, could you explain why and how which is not the case in this work?

Rating

7: Accept: Technically solid paper, with high impact on at least one sub-area, or moderate-to-high impact on more than one areas, with good-to-excellent evaluation, resources, reproducibility, and no unaddressed ethical considerations.

Confidence

4: You are confident in your assessment, but not absolutely certain. It is unlikely, but not impossible, that you did not understand some parts of the submission or that you are unfamiliar with some pieces of related work.

Soundness

4 excellent

Presentation

3 good

Contribution

3 good

Limitations

see weakness above

Reviewer GLUc7/10 · confidence 4/52023-06-30

Summary

*The paper studies the problem of unfairness introduced by weak supervision methods through noisy data augmentation. It proposes a mitigation strategy using counterfactuals that would create a more balanced/unbiased dataset. * Paper shows that Labeling functions can be arbitrarily biased by preferring examples away from the center of the distribution and show theoretically there is no change in sample complexity required to achieve the gains in accuracy and fairness, under strict assumptions. * Empirical results show that by further augmenting data that counterfactually transforms examples across groups, both accuracy and fairness metrics can be improved - with results on synthetic and popular benchmark fairness and weak supervision datasets.

Strengths

* Strong empirical results on synthetic and real datasets * New problem formulation that extends fairness methods to the weak supervision case

Weaknesses

* Theoretical results are under strong distributional assumptions, how they can be relaxed should be better articulated * Standard data augmentation techniques such as Autolabel [1] are missing in evaluation set up. https://ieeexplore.ieee.org/abstract/document/10136178

Questions

(see weaknesses)

Rating

7: Accept: Technically solid paper, with high impact on at least one sub-area, or moderate-to-high impact on more than one areas, with good-to-excellent evaluation, resources, reproducibility, and no unaddressed ethical considerations.

Confidence

4: You are confident in your assessment, but not absolutely certain. It is unlikely, but not impossible, that you did not understand some parts of the submission or that you are unfamiliar with some pieces of related work.

Soundness

3 good

Presentation

3 good

Contribution

3 good

Limitations

This is missing, and should be included

Reviewer UYjh7/10 · confidence 4/52023-07-07

Summary

This paper proposes a novel bias mitigation technique to address the fairness issues in weak supervision settings. The core idea is to use a counterfactual fairness-based correction method. The authors theoretically show that the proposed method can improve both accuracy and fairness and this is also supported by both synthetic and real datasets evaluations. Overall, this paper calls attention to bias and fairness studies in the weak supervision context, which has never been addressed specifically and provides an effective and theoretically sound method. To the best of my knowledge, the method considered in this paper is novel and is a valuable contribution to the community. Based on the above factors, I recommend acceptance.

Strengths

This paper is well-motivated and well-written. The proposed method is intuitively simple yet very effective and theoretically sound. I appreciate the comprehensive evaluation both with synthetic data and with WRENCH.

Weaknesses

This paper focuses on a novel area of weak supervision studies. Though the method proposed is simple, I do believe it has any major weak nesses.

Questions

I welcome the authors to discuss the potential drawbacks of formulating the fairness argument behind counterfactual fairness, particularly in the programmatic weak supervision settings. Given diverse sources, wouldn't the biases be smoothed out?

Rating

7: Accept: Technically solid paper, with high impact on at least one sub-area, or moderate-to-high impact on more than one areas, with good-to-excellent evaluation, resources, reproducibility, and no unaddressed ethical considerations.

Confidence

4: You are confident in your assessment, but not absolutely certain. It is unlikely, but not impossible, that you did not understand some parts of the submission or that you are unfamiliar with some pieces of related work.

Soundness

4 excellent

Presentation

3 good

Contribution

3 good

Limitations

Overall, this paper is well organized and I have not identify any apparent limitations for the current content.

Reviewer wmSM5/10 · confidence 4/52023-07-08

Summary

The paper focuses on the bias issue in weak supervision. The paper shows the labeling functions of weak supervision may produce biased pseudo-labels, which is first empirically shown in the paper. Also, they theoretically show that even though a dataset has fair ground-truth labels, the weak supervision process using labeling functions may produce biased weak labels for the dataset. To mitigate this issue, the paper proposes an optimal transport-based algorithm that modifies the weak labels to be fairer. In experiments, the paper evaluates the proposed algorithm in tabular, NLP, and computer vision datasets.

Strengths

S1. The paper considers an important research problem, the fairness issue in the weak supervision pipeline. It seems this paper is the first work to handle fairness in weak supervision. S2. The paper theoretically shows the labeling functions may produce unfair pseudo-labels even though the underlying label distribution is fair. S3. The proposed method can be used together with the existing fair in-processing algorithms, which is another good aspect of the paper.

Weaknesses

W1. The explanations on the labeling function itself are limited. - For example, can the number of labeling functions affect the fairness performance? The Snorkel original paper [1] mentioned that the number of labeling functions highly affects the labeling performances, but the current paper does not explain how such details may make differences in the fairness scenario. - Also, as shown in the vision dataset experiments, using fairer labeling functions can reduce the effectiveness of the proposed algorithm. Then, why is the proposed transport algorithm better than making labeling functions itself fairer? Currently, the paper seems to use very simple labeling functions (described in the appendix), and it would be helpful if the paper could provide any comparison between the proposed algorithm and another possible direction of making fairer labeling functions. [1] Ratner et al., Snorkel: Rapid Training Data Creation with Weak Supervision, VLDB’18 W2. Experiments show some questionable results. - As several scenarios show large F1 score drops, it is unclear whether it is okay to use the proposed algorithm. I understand the F1 score can be affected by the imbalanced label classes, but the F1 score degradation is still severe. For example, SBM (OT-S) + LIFT case in the Bank Marketing dataset and all SBM results in the Civil dataset show large F1 score drops. It would be helpful if the paper could provide at least some possible way to prevent such F1 score drops. - In the experiments, no error range is provided, which makes the empirical results less convincing.

Questions

Although the paper provides several meaningful discussions and insights, there are some remaining concerns, especially regarding the labeling function and empirical results. The details are in the above weakness section.

Rating

5: Borderline accept: Technically solid paper where reasons to accept outweigh reasons to reject, e.g., limited evaluation. Please use sparingly.

Confidence

4: You are confident in your assessment, but not absolutely certain. It is unlikely, but not impossible, that you did not understand some parts of the submission or that you are unfamiliar with some pieces of related work.

Soundness

2 fair

Presentation

2 fair

Contribution

3 good

Limitations

The paper did not discuss the limitations and potential negative social impact of the work. A possible limitation of this work can be other prominent fairness definitions that this work cannot handle.

Reviewer 6e6W2023-08-10

Post rebuttal

I read the authors' rebuttal and other reviews. The author nicely addressed my questions. I would vote for accept and keep my score.

Authorsrebuttal2023-08-11

Thank you again; further questions?

Dear Reviewers, We thank you again for your feedback, questions, and suggestions! We believe we have answered all of your questions in our responses and the updated draft. If you have additional questions, we would love to answer them! The Authors

Program Chairsdecision2023-09-21

Decision

Accept (poster)

© 2026 NYSGPT2525 LLC