causal

Algorithms Trained On Biased Data Reproduce Bias

Because algorithms are trained on historical data, they incorporate existing human biases—e.g., Amazon's hiring algorithm learned to weed out women because women hadn't been successfully hired before—and these biased systems are now deployed in criminal justice and hiring as opaque black boxes whose decisions humans can't scrutinize.

causalpending

Speaker

Kate Darling

Evidence Quote

“Amazon did this recently and got in trouble for it, because it turns out, if you train an algorithm on historical data, which is what we do with all the algorithms, it will incorporate a lot of biases”

Source

46 | Kate Darling on Our Connections with Robots— Sean Carroll's Mindscape
Created: 6/13/2026, 12:19:58 AM

Your notes

Loading notes...