The Hierarchy Of Assumptions: Why All Generalizations Are Not Equal

Getting some righteous pushback on my piece about Men’s Rights Advocates, because I accused them of generalizations while generalizing myself.  A valid critique.  Generalizations are generally bad.
But:
There’s a distinct difference between “Generalizing 50% of our species” and “Generalizing a subculture composed of comparatively few people.”
If I say, “All people love Shaun of the Dead,” then chances of me picking a person at random from the seven billion humans who populate this planet who has even seen Shaun of the Dead: Pretty slim.
If I say, “All women love Shaun of the Dead,” then chances of me picking a woman at random from the three-point-five billion women who’s seen it: Also pretty slim.
If I say, “All liberals love Shaun of the Dead,” then I’ve at least chosen a group of people who have, to some extent, self-defined themselves.  But the concept of what a “liberal” is is pretty ill-defined, and even if I did, the chances that one of the millions of liberal-minded people around the world love Shaun of the Dead?  The odds are better than “All Women” or “All People,” but still kinda sketchy.
If I said “All moviegoers love Shaun of the Dead,” well, again – we’re at least generalizing by an activity that people have chosen to do, and a very well-defined one, which considerably ups the chance that they would have at least seen Shaun of the Dead.  And that they liked it.  But it’s probably not all that accurate a generalization even then.  Yet it’s definitely more of a defensible generalization than the first three.
If I said, “All cinema nerds love Shaun of the Dead,” then we have narrowed the range considerably: not just people who go to the movies, but people who have chosen to be nerdy (and probably somewhat obsessive) about movies.  Chances of my picking a cinema nerd at random (however one reasonably defines it) who has both seen Shaun of the Dead and likes it are approaching maybe one in four, or maybe one in three.
If I said, “All zombie movie fans love Shaun of the Dead,” well, now we’re talking.  That’s still a generalization, but people who have self-identified as zombie movie lovers have a better-than-even chance of having seen Shaun of the Dead, a high-profile zombie movie.  But even then, a fair number of them have seen it and found its comedy dissatisfying, or just haven’t gotten around to it.
If I said, “Everyone who watches movies at Gini and Ferrett’s house loves Shaun of the Dead,” well, you’d be batting about 66%.  And there’s a reasonable statement to make – not true in a perfect sense (I wouldn’t use it as an algorithm) but true in the sense that if you picked someone at random you’d have a good chance of being correct.
My point of this ludicrous thought-exercise is that when I say, “All Men’s Rights Advocates, a very narrow subculture on the Internet who have chosen to self-identify with a set of beliefs about the damaging things women are doing to men” act this way, it’s not entirely correct.  (Even if every dude I’ve encountered who identified as an MRA did act that way, I’m certainly not going to assume that every one is.)  But it’s definitely more correct as a generalization that says “All women act this way,” which a very significant number of Men’s Rights Advocates do say, repeatedly, in essays.
So yeah.  Am I right to make that generalization?  Not entirely.  But not all generalization is bad.  If I say, “Tea Partiers are against taxes,” well, there’s doubtlessly Phyllis the Tax-Lovin’ Republican, but if I picked a Tea Partier at random I’d probably be in the ballpark.  Likewise, if I said, “All of the active LiveJournal bloggers are Russian,” well, it’s not entirely true – the top 10 are English – but LiveJournal is certainly way more popular in Russia than here, these days, so it’s not an entirely inaccurate statement.
When I generalize about MRA dudes, I’m generalizing about a movement that has self-chosen certain beliefs.  Maybe not all MRA guys.
But quite potentially enough.