Robustness to Spurious Correlations via Human Annotations

Megha Srivastava, Tatsunori Hashimoto, Percy Liang
Proceedings of the 37th International Conference on Machine Learning, PMLR 119:9109-9119, 2020.

Abstract

The reliability of machine learning systems critically assumes that the associations between features and labels remain similar between training and test distributions. However, unmeasured variables, such as confounders, break this assumption—useful correlations between features and labels at training time can become useless or even harmful at test time. For example, high obesity is generally predictive for heart disease, but this relation may not hold for smokers who generally have lower rates of obesity and higher rates of heart disease. We present a framework for making models robust to spurious correlations by leveraging humans’ common sense knowledge of causality. Specifically, we use human annotation to augment each training example with a potential unmeasured variable (i.e. an underweight patient with heart disease may be a smoker), reducing the problem to a covariate shift problem. We then introduce a new distributionally robust optimization objective over unmeasured variables (UV-DRO) to control the worst-case loss over possible test- time shifts. Empirically, we show improvements of 5–10% on a digit recognition task confounded by rotation, and 1.5–5% on the task of analyzing NYPD Police Stops confounded by location.

Cite this Paper

BibTeX
@InProceedings{pmlr-v119-srivastava20a, title = {Robustness to Spurious Correlations via Human Annotations}, author = {Srivastava, Megha and Hashimoto, Tatsunori and Liang, Percy}, booktitle = {Proceedings of the 37th International Conference on Machine Learning}, pages = {9109--9119}, year = {2020}, editor = {III, Hal Daumé and Singh, Aarti}, volume = {119}, series = {Proceedings of Machine Learning Research}, month = {13--18 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v119/srivastava20a/srivastava20a.pdf}, url = {https://proceedings.mlr.press/v119/srivastava20a.html}, abstract = {The reliability of machine learning systems critically assumes that the associations between features and labels remain similar between training and test distributions. However, unmeasured variables, such as confounders, break this assumption—useful correlations between features and labels at training time can become useless or even harmful at test time. For example, high obesity is generally predictive for heart disease, but this relation may not hold for smokers who generally have lower rates of obesity and higher rates of heart disease. We present a framework for making models robust to spurious correlations by leveraging humans’ common sense knowledge of causality. Specifically, we use human annotation to augment each training example with a potential unmeasured variable (i.e. an underweight patient with heart disease may be a smoker), reducing the problem to a covariate shift problem. We then introduce a new distributionally robust optimization objective over unmeasured variables (UV-DRO) to control the worst-case loss over possible test- time shifts. Empirically, we show improvements of 5–10% on a digit recognition task confounded by rotation, and 1.5–5% on the task of analyzing NYPD Police Stops confounded by location.} }
Endnote
%0 Conference Paper %T Robustness to Spurious Correlations via Human Annotations %A Megha Srivastava %A Tatsunori Hashimoto %A Percy Liang %B Proceedings of the 37th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2020 %E Hal Daumé III %E Aarti Singh %F pmlr-v119-srivastava20a %I PMLR %P 9109--9119 %U https://proceedings.mlr.press/v119/srivastava20a.html %V 119 %X The reliability of machine learning systems critically assumes that the associations between features and labels remain similar between training and test distributions. However, unmeasured variables, such as confounders, break this assumption—useful correlations between features and labels at training time can become useless or even harmful at test time. For example, high obesity is generally predictive for heart disease, but this relation may not hold for smokers who generally have lower rates of obesity and higher rates of heart disease. We present a framework for making models robust to spurious correlations by leveraging humans’ common sense knowledge of causality. Specifically, we use human annotation to augment each training example with a potential unmeasured variable (i.e. an underweight patient with heart disease may be a smoker), reducing the problem to a covariate shift problem. We then introduce a new distributionally robust optimization objective over unmeasured variables (UV-DRO) to control the worst-case loss over possible test- time shifts. Empirically, we show improvements of 5–10% on a digit recognition task confounded by rotation, and 1.5–5% on the task of analyzing NYPD Police Stops confounded by location.
APA
Srivastava, M., Hashimoto, T. & Liang, P.. (2020). Robustness to Spurious Correlations via Human Annotations. Proceedings of the 37th International Conference on Machine Learning, in Proceedings of Machine Learning Research 119:9109-9119 Available from https://proceedings.mlr.press/v119/srivastava20a.html.

Related Material