Removing biased data to improve fairness and accuracy

02/05/2021
by   Sahil Verma, et al.
61

Machine learning systems are often trained using data collected from historical decisions. If past decisions were biased, then automated systems that learn from historical data will also be biased. We propose a black-box approach to identify and remove biased training data. Machine learning models trained on such debiased data (a subset of the original training data) have low individual discrimination, often 0 and lower statistical disparity than models trained on the full historical data. We evaluated our methodology in experiments using 6 real-world datasets. Our approach outperformed seven previous approaches in terms of individual discrimination and accuracy.

READ FULL TEXT

page 9

page 10

page 15

page 16

research
02/26/2020

Fairness-Aware Learning with Prejudice Free Representations

Machine learning models are extensively being used to make decisions tha...
research
06/28/2023

Systematic analysis of the impact of label noise correction on ML Fairness

Arbitrary, inconsistent, or faulty decision-making raises serious concer...
research
01/06/2020

Fair Active Learning

Bias in training data and proxy attributes are probably the main reasons...
research
06/22/2020

Improving LIME Robustness with Smarter Locality Sampling

Explainability algorithms such as LIME have enabled machine learning sys...
research
12/05/2019

Perfectly Parallel Fairness Certification of Neural Networks

Recently, there is growing concern that machine-learning models, which c...
research
11/17/2017

Predict Responsibly: Increasing Fairness by Learning To Defer

Machine learning systems, which are often used for high-stakes decisions...

Please sign up or login with your details

Forgot password? Click here to reset