I feel like the point here is that there is a sort of silly manipulation that OP is actively conducting on initial conversations with people that only take one data point. I am 100% sure that people are much more complex than an Atlas Shrugged test, very similar to the way that prediction algorithms use multiple attributes. One test or attribute is overly simplistic.

I don't think it requires any special insight to understand that a fantastic approach to meeting people is to simply be genuine and honest.

There's a difference in reaction I think. Two people can hear "I love Atlas shrugged" from a new acquaintance and one would start arguing that it's a bad book and anyone who likes it is dumb, etc. Another would say OK, but will think that the speaker is dumb.

Basically there's no way to find if someone would "tolerate ideas they hate" or not if they refuse to engage with the idea altogether. The second guy in my example looks like he would tolerate these ideas but he won't.