In pol.is, if you get 5,000 people voting exactly the same way, it’s only to be one dot. In the principal component analysis, it has no effect at all. Whereas, in a reply-happy environment, if you get 5,000 people individually replying, then you just marginalize everybody else.