Anecdote and Evidence

Andrew Gelman has a written a detailed post about Karl Weick's anecdote about the Hungarian soldiers in the Alps. He and I really think alike on this issue. It got me thinking about the epistemological difference between anecdotes and evidence.

While my concept of knowledge does begin with "justified, true belief" it doesn't end there. I've come to believe that knowledge, and especially "academic" knowledge, is much better understood as the ability to converse intelligently about something. Truth is a kind of norm for the things we know; if we claim to know something, convention says we are claiming it is true. In fact, some epistemologists also posit a "knowledge norm" for belief: if I believe something, I'm also claiming to know it. But another norm is that if you know something you know much more as well.

When Weick tells us the story of the soldiers in the Alps, he is, on the surface, claiming that the story is true. "This is an incident that happened," he says. By stepping into the role of the storyteller, he is claiming to know what happened. But consider that detail that Gelman emphasizes at the end of his post. Suppose he tells the story and someone who knows a bit about European geography and politics says: "What were Hungarian soldiers doing on military manoeuvres in Switzerland?" Well, if Weick really knew the story, he would have a good answer. (In fact, he would have known not to include that detail in the story.) The same goes for the possibility Engel raises, i.e., that the leader of the soldiers knew that that the map was wrong but pretended it was right.

But Weick's understanding of the story is limited to exactly the words he uses to tell it. This is the difference between merely knowing an anecdote and actually having evidence for one's beliefs. While it is sometimes suggested that nothing depends on the story being true, it seems clear to me that Weick wants us to believe it is true. I certainly think that many of his readers believe it. But very few people have looked at the evidence.