Generalizing from one example | Hacker News Reader

Generalizing from one example

lesswrong.com

53 pointsjibiki6 comments

I actually think this is one danger of the internet and increasingly specialised communication channels - individuals are never exposed to contrary opinions, and either gradually grow to believe that their opinions are the only valid ones or fail to see legitimate threats to their world view.

The best example I know is politics - so many partisan sites exist that individuals who are firmly within one camp can get all of their news through their preferred filter - unlike MSM which at least pretends to be objective and achieves somewhat middle ground.

For HN, I think it's fair to say there are some characteristics we have in common which are not common in the wider community (say, respect for rationality and intelligence). I hope we're not too insular to only find news and conversation here; if we're are insular, we may wake up one day to discover our values have been even more marginalised in the wider community and we didn't even see it coming.

For those interested, this is basically the theory of law professor Cass Sunstein in his book Republic.com. He calls this problem the "daily me," the idea being that content can become too personalized to one's views, resulting in irrationally extreme perspectives on those views and their correctness.

I would recommend the book, and also the numerous criticisms of his theory (that prompted him to partially retract his argument in the second edition), easily found online.

A book? Yes, a book! Go out and stare at a fixed-content display for a while.

"For HN, I think it's fair to say there are some characteristics we have in common which are not common in the wider community (say, respect for rationality and intelligence)."

I believe that at least a portion of this "respect for rationality and intelligence" comes from a lack of the constant conflict among users that you see on more mainstream websites (because they attract people of a wide variety of different opinions). Even the most imperturbable of us are only capable of maintaining a civil tone to a certain point on more controversial topics. The way I see it, the key to success for maintaining healthy discourse is to achieve a balance between homogeneity and diversity of user beliefs/viewpoints.

Absolutely. I generalized from one datum once, and hoooo boy. I'll never do that again.

I'm being funny, but with a purpose: if acting on one datum was always fallacious, you'd think it would be bred out of us by now. Sometimes you have to act on what you know, and different ages favor different personality types.

For instance, I believe that strong visual people have an advantage in our times. I suspect it's because visual people are able to explore (by quite literally visualizing) much larger solution spaces in shorter periods of time. Like the author I am skeptical of Galton's claim that people who cannot visualize are over-represented in math and science.

> if acting on one datum was always fallacious, you'd think it would be bred out of us by now.

True enough, but you have to look at the context.

LW is sort of obsessed with Aumann's agreement theorem at the moment, which is why it frequently has long back-and-forth threads that don't go anywhere. This post can be seen as a reminder that sometimes you genuinely cannot convince someone of something, because it's all in your head.

> I am skeptical of Galton's claim that people who cannot visualize are over-represented in math and science.

I was about to say that it's impossible to do (algebraic) topology without visualizing... but then I'd be generalizing from one data point.

> if acting on one datum was always fallacious, you'd think > it would be bred out of us by now.

I think that we're set up to predict and act on N data points, and sometimes N happens to be one. The point of the article, to me, was that you have to understand this behavior in yourself and watch out for fallacies.

Just as if you know about confirmation bias, so you keep an eye out for contradictory evidence, if you know about the statistical problem of generalizing from one you can watch out for that, too.

Reply on news.ycombinator.com