On Hate Scaling Laws for Data-Swamps

What it is

Birhane and colleagues show that training on a larger scrape makes hateful content and racist misclassification worse, not better — direct evidence against "more data fixes it".

Why I recommend it

Useful whenever someone argues scale solves bias. Free in full on arXiv.

Topics

#ai ethics #ai-ethics #dataset-audit #free #reading #research #technology-and-ethics

Added Sep 19, 2026 · 0 opens