A Survey on Bias and Fairness in Machine Learning.¶
Mehrabi, N., Morstatter, F., Saxena, N., Lerman, K., & Galstyan, A. (2021). A Survey on Bias and Fairness in Machine Learning. ACM Computing Surveys, 54(6), 1-35.
Cited by¶
1 citation across 1 artifact.
Each citation links to the sentence it supports in the citing article.
Primes¶
- Bias
- Mehrabi and colleagues (2021) survey how such biases enter learning pipelines at data collection, labeling, and modeling stages and propagate into systematically skewed outputs.
This sourceSurvey of how model and dataset bias enter learning pipelines at data collection, labeling, and modeling stages and propagate into reproducibly skewed outputs, including institutional and algorithmic bias as directional displacement relative to a fairness or representational target.
Supported in partVerified against the work's full text
The survey compiles bias types "that exist in different cycles from data origins to its collection and its processing", reports at the labeling stage that structured-prediction methods "amplify the existing bias in data", and treats modeling choices in fairness evaluation. Its stated aim is to examine domains and subdomains of AI for what researchers have observed with regard to unfair outcomes, which is the pipeline-to-output coverage the sentence describes. The three stages and the propagation into skewed outputs are carried by separate passages rather than by one statement, and the word systematically is the article's.
“In the cases where the underlying training data contains biases, the algorithms trained on them will learn these biases and reflect them into their predictions.”
- Mehrabi and colleagues (2021) survey how such biases enter learning pipelines at data collection, labeling, and modeling stages and propagate into systematically skewed outputs.
Verification¶
Does it exist? Confirmed. This work's DOI resolves to a registered record, which fixes its identity. That is all it fixes.
Does it back the claim? Read against the text for 1 of 1 citation: 1 supported in part. Each verdict is shown under its citation below, with what in the work backs the sentence.
Support is checked per citation rather than per work — the same source can be cited soundly in one article and wrongly in another. Per-citation recording began recently, so a citation with no recorded check is a gap in the record rather than evidence it went unchecked.
See how references were verified.
Registry ID ref:0742328f5c31 · see in the full table