The generalization only works if it's weak enough to be meaningless. Thus, the generalization is bad. Examples don't make it useful.
"People act to their own benefit" is an empty generalization that adds no useful information by itself and free of context like that only serves to mislead people. It's only true if "benefit" is explicitly undefined, and only useful if you contextualize it to an specific action and benefit that you can empirically determine it's validity, like in the article.
> I also didn't say anything about pro/antisocial people
The article, and the entire discussion is about pro/antisocial behavior.
I didn't propose it, just clarified what I believe to be their point.
I think it is a useful generalization when you possess a theory of mind, however. In low-trust environments, assuming criminal self-interest is often what keeps people safe... if you're basing your decision on a lack of information, wariness is warranted. Not every social environment is a conversational environment.