Non-existent markets in everything: anecdotes

post by KatjaGrace · 2015-10-16T04:48:00.000Z · LW · GW · 0 comments

Careful scientific study is widely considered a better basis for one’s views than anecdote. But anecdotes are more fun, so everyone reads about them anyway. Another virtue of anecdote is its low price. To verify scientifically that vitamin E is good for you might take millions of dollars or literally be impossible. To verify anecdotally that your friend credits vitamin E with some kind of impressive life-altering success requires nothing but a moderate number of  epistemically sloppy friends who like to try things. Also some time talking to them. Let’s say a few hundred dollars.

But some free anecdote about your friend’s nutritional revelations is not going to get you far. The really good anecdotes are harder to come by. Here’s a good anecdote:

Bird and Layzell used genetic algorithms to evolve a design for an oscillator, and found that one of the solutions involved repurposing the printed circuit board tracks on the system’s motherboard as a radio, to pick up oscillating signals generated by nearby personal computers.

It vividly makes exactly the point the authors want: AI agents might technically satisfy the tasks they are given, while not doing what their creators expected. It gives the reader a concrete instance of something that might have otherwise seemed unrealistically tied up with control of mythical genies. I have seen this single anecdote used valuably multiple times.

This anecdote was both more expensive than an anecdote about your friend’s success with vitamin E. However it could probably have been much cheaper than a scientific study. It was actually scientific research, but note that the research doesn’t have to be nearly as rigorous  or complete for the purpose of making an anecdote about it.

Bird and Layzell were not trying to make an anecdote at all (as far as I know). They were trying to make genetic algorithms that could produce oscillator designs. It was by pure luck that they produced this useful byproduct.

In a different world, the people doing this study would be the ones who knew there was demand for a nice vivid illustration of this particular idea. They would have thought about what would best paint that picture, and thought of this experiment among a few other things. They would carry them out, looking for one which nicely instantiated the desired idea. The next day they might go out and look for a particularly tragic case where someone waited too long because they were afraid, or bridge that collapsed because an engineer made a particular error.

Could we have a market for anecdotes? I claim anecdotes could be created intentionally, separately from the rest of writing, because they are often modular – so might be outsourced – and are used in multiple pieces of writing – so are especially valuable. They are more valuable again because they are bits that people frequently remember from writing. And arguably changing the image a person associates with an idea can have an especially powerful effect on their view of it. Anecdotes are also especially good for outsourcing if thinking good abstract thoughts is not that well correlated with knowing the best stories to illustrate them.

Do we already have such a market? Not that I know of. If we do, please tell me; I want to buy some anecdotes. If not, why not?

0 comments

Comments sorted by top scores.