Summary of Fairness Feedback Loops: Training on Synthetic Data Amplifies Bias, by Sierra Wyllie et al.
Fairness Feedback Loops: Training on Synthetic Data Amplifies Bias
by Sierra Wyllie, Ilia Shumailov, Nicolas Papernot
First submitted to arxiv on: 12 Mar 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: None
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 This research paper presents a framework to track model-induced distribution shifts (MIDS) over generations of machine learning models. MIDS occur when previous model outputs pollute new training sets, leading to performance, fairness, and representation losses in initially unbiased datasets. The authors identify opportunities for positive interventions, introducing the concept of algorithmic reparation (AR), which can improve upon unfairnesses through representative training batches. By simulating AR interventions using stochastic gradient descent, the paper demonstrates how to mitigate MIDS and take accountability for their impact. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This study is about how machine learning models affect the world around us. It shows that when we use these models to make decisions, they can create problems by changing the way data is distributed. This means that models can become unfair or biased over time. The researchers propose a new approach called algorithmic reparation, which helps to fix these problems and make sure that everyone is treated fairly. |
Keywords
* Artificial intelligence * Machine learning * Stochastic gradient descent
