AI models, when trained on polarized content, will inherit and normalize existing biases, leading to oversimplified and nuance-free information.
"The risk is that these models not only inherit this bias but they also normalize it, smoothing over complexity with confident, context and nuance-free answers."