Aside from algorithmic gender bias and "mathwashing" as mentioned in the article, what other sources need be considered?