[ALittleLight]

Regarding the difference between a blog and reddit I would say that the comment section harbors people with many different views. Some of the comments may be objectionable (again: I didn't see examples) but so what? That's got to be true of everywhere with a comments section. Why make that a focus?

You write "if the claim is indeed true that racists/sexists think the blog is great" - but this isn't even the kind of thing that could be true. Imagine how you would prove this claim, do you go to the National Association of Racists and ask for their stance on SSC? Is there a poll of racists and their opinion on niche blogs somewhere?

"Racists" is ill defined by itself. People aren't racist or not the way they are right or left handed. Racism is a spectrum and much of it is debatable and nuanced and affected by context and all that. This is like saying "Some number of an undefined group like something. Prove me wrong." That's not a legitimate claim and there is no need to try and rebut it.

I have no doubt that some racists like the blog and other racists don't like it. Do more racists (proportionally) like and read SSC or the NYT? I don't think you can answer that.

Regarding the feminist issue I'll have to go back and read that post before writing more.

[MrMan]

Some of our applied social scientists are being allowed to work with very large datasets at places like Facebook and I believe they are increasingly in a position to make quantifiable statements on what racists or any other cluster-able groups of humans like. If by spectrum you mean “distance from a centroid” it might be more precise.

Marketing is essentially the working, reproducible arm of the social sciences, and marketers know a lot about human preferences and how to link them to motivation.

[disgruntledphd2]

> Some of our applied social scientists are being allowed to work with very large datasets at places like Facebook and I believe they are increasingly in a position to make quantifiable statements on what racists or any other cluster-able groups of humans like. If by spectrum you mean “distance from a centroid” it might be more precise.

Just because you have more data, it doesn't mean that you can identify constructs like racism from this loads of data. You'd need some kind of ground truth mechanism (like an index of behaviour towards different races) which neither Facebook (nor anyone else) has. It's just wildly implausible.

Maybe, in ten years, NLP will be good enough to identify this, but I don't think these constructs are easily identifiable from text, certainly not in a public space such as Facebook.

> Marketing is essentially the working, reproducible arm of the social sciences, and marketers know a lot about human preferences and how to link them to motivation.

I get what you're trying to say here, and maybe that works in a small number of places, but having worked with marketers in analytics for the past decade or so, suffice it to say that I rarely praise the standards of experimention and (lack of) rigour employed by them.

[ALittleLight]

That still relies on the scientists defining and labeling racists which will be an arbitrary process. Regardless - do you think anything of the kind was done by the NYT?

[chillacy]

Unsupervised learning doesn’t give us named labels. Someone still has to name cluster 1, 2, 3... etc. So not sure that actually gets around what amounts to a semantic question.