[b1daly]

Well I think Stalin summed up the perils of totally ignoring anecdotes pretty well:

"One death is a tragedy; one million is a statistic."

This topic sort of came up the other day in the thread about the girl losing the iPad software she needs to talk (Silencing Maya). http://news.ycombinator.com/item?id=4103344

Some commenters thought the story should be ignored as a data point about the societal value of patents. I disagree in that case because I don't believe economics or social sciences have anywhere near the amount of rigor and theory to make a good claim on whether the patent system is a net benefit to society. Deciding such a thing is a very old, and classic problem in philosophy. The OP seems to implicitly relying on a variant of utilitarianism, which IMO is wholefully inadequate to rely on for moral decisions.

Since science is so hard, there is a lot of bad science out there. What gets reported in the wider media has all sorts of weird selection bias, never mind what gets picked for publication in journals. Anecdotes are human stories and they are deeply connected to why we care. Statiscal tools can be used in ways that justify harm people in the name of greater good. I agree that scientific medicine provides tools for "real"'medicine that other methods don't. I just think we should remember to real peoples stories, especially if they are true. Anyone reading this probably gets that one can't exactly counter balance the effect of anecdotes within themselves. But it can still be countered some, I think it's worth it to hear stories (that are relaxant to the topic). Now what is the basis of my beliefs? Mostly intuition, not rationality. But I don't think it's possible to undermine my ideas with some kind of experimentally bases argument. There's just way to many variables!

[slurgfest]

I don't understand the part about Stalin's advice. What was Stalin saying would happen if you ignored anecdotes? How do you figure he was talking about the value of anecdotes?

[b1daly]

He was talking about undertaking large activities (war for example) that would kill hundreds of thousands of people could be tolerated politically by a populace in a way that a single murder might cause an outrage. The large numbers make the horror abstract. It's easier for people to relate to a story about a specific individual.

The quote (and variations) are quite famous, it might be apocryphal, but I think it is true. Hearing about hundreds of thousands of people massacred in a foreign land doesn't hit home when you are reading about far away.

That's why reporters (New York Times style)try to weave in illustrations and stories about individuals even when discussing a larger trend.

[gwern]

Funny thing, I always took this quote as Stalin cynically remarking on exactly OP's point: people overweight anecdotes by thousands to millions of times their actual worth as Bayesian evidence.