Summary of How to Be Fair? a Study Of Label and Selection Bias, by Marco Favier et al.
How to be fair? A study of label and selection bias
by Marco Favier, Toon Calders, Sam Pinxteren, Jonathan Meyer
First submitted to arxiv on: 21 Mar 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Computers and Society (cs.CY)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary A novel study investigates the interplay between biased data, model predictions, and bias mitigation techniques, with a focus on establishing relationships between different types of biases and the effectiveness of various mitigation methods. The research builds upon previous work by Wick et al., which demonstrated that certain bias mitigation techniques can lead to more accurate models when evaluated on unbiased data. However, the study acknowledges the limitations of this approach, as it remains unclear which techniques are effective under what circumstances. To address this problem, the authors propose a theoretical framework for categorizing bias mitigation techniques based on the type of bias they optimize. The paper illustrates this principle by examining label and selection bias, demographic parity, and “We’re All Equal” fairness measures. The results provide insights into when minimizing fairness measures may not necessarily lead to the fairest possible distribution. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary A new study explores how biased data affects machine learning models and tries to fix these problems with special techniques. Researchers found that some of these techniques can make better predictions on fair data, but they don’t know which ones work best in what situations. The goal is to understand why different methods work or not work for different types of biases. This paper shows how certain techniques can be grouped based on the type of bias they try to fix and provides examples of label and selection bias, demographic parity, and “We’re All Equal” fairness measures. |
Keywords
* Artificial intelligence * Machine learning
