Questions and Discussion
Panelat Wrong at the Root: Racial Bias and the Tension Between Numbers and Words in Non-Internet Data · 2019-06-05..06-07·
summary
Comparative word embeddings are computed from public corpora like Wikipedia, NYT, WSJ, BBC and tracked over 22 time points to measure covariance and similarity; analysis finds a few abrupt shifts (about five notable change points) but no causal claim, only post‑hoc matching to historical events. Validation uses citizen‑science surveys and Census metadata; recommendations include automated flag flawed categories in legacy datasets and noting that the Human Cell Atlas lacks donor‑sampling transparency for downstream fairness work.
AI-written summary · about
chapters (4)
related talks
45mDecoding Topics: A Comparative Analysis of Python’s Leading Topic Modeling Libraries Using Climate C
35mProduct fairness testing for Machine Learning developers
57mDiscover semantic topics in a collection of documents
20mApplying Computational Analysis to Textual Data from the Wild: a Feminist Perspective
54mData doesn't lie — but it can mislead: How to ensure integrity of your Machine Learning applications
31mPanel Discussion : Ethics in Data Science
18mBuilding a Gendered Dictionary
45mData Science: It's Easy as Pyǃ