Controllable Guarantees for Fair Outcomes via Contrastive Information Estimation
By Umang Gupta, Aaron Ferber, Bistra N. Dilkina, Greg Ver Steeg. 2021
80 citations recorded by the source corpus.
Controlling bias in training datasets is vital for ensuring equal treatment, or parity, between different groups in downstream applications. A naive solution is to transform the data so that it is statistically independent of group membership, but this may throw away too much information when a reasonable compromise between fairness and accuracy is desired. Another common approach is to limit the ability of a particular adversary who seeks to maximize parity. Unfortunately, representations produced by adversarial approaches may still retain biases as their efficacy is tied to the complexity of the adversary used during training. To this end, we theoretically establish that by limiting the mutual information between representations and protected attributes, we can assuredly control the parity of any downstream classifier. We demonstrate an effective method for controlling parity through mutual information based on contrastive information estimators and show that they outperform approaches that rely on variational bounds based on complex generative models. We test our approach on UCI Adult and Heritage Health datasets and demonstrate that our approach provides more informative representations across a range of desired parity thresholds while providing strong theoretical guarantees on the parity of any downstream algorithm.
Subfields: Artificial Intelligence, Genetics, Statistics and Probability
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
- Aram Galstyan
- Fred Morstatter
- Ninareh Mehrabi
- Lawrence Carin
- Zhe Gan
- Liqun Chen
- Shuyang Gao
- Greg Ver Steeg
In Genetics
In Statistics and Probability
- Zhe Gan
- Aram Galstyan
- Lawrence Carin
- Greg Ver Steeg
- Fred Morstatter
- Shuyang Gao
- Ninareh Mehrabi
- Chunyuan Li
Citations in the corpus
- Attributing Fair Decisions with Attention Interventions (2021, methodology): “We used the same pre-processing and train-test splits as in Gupta et al. (2021).”
- Simpler, Faster, Stronger: Breaking The log-K Curse On Contrastive Learners With FlatNCE (2021, background): “Successful applications have been reported in computer vision [5, 6, 1, 7], natural language processing [8, 9], reinforcement learning [10], fairness [11], amongst many others.”