Bayesian truth serum
June 16, 2011 13 Comments
Here’s a sneaky trick for extracting the truth from someone even when she’s trying to conceal it from you: Rather than asking her how she thinks or behaves, ask her how she thinks other people think or behave.
MIT professor of psychology and cognitive science Drazen Pelec calls this trick “Bayesian truth serum,” according to Tyler Cowen in Discover Your Inner Economist. The logic behind it is simple: our impressions of “typical” attitudes and behavior are colored by our own attitudes and behavior. And that’s reasonable. You should count yourself as one data point in your sample of “how people think and behave.”
Your own data is likely to influence your sample more strongly than other data points, however, for two reasons. First, because it’s much more salient to you, compared to your data about other people, so you’re more likely to overweight it in your estimation. And second, through a ripple effect — people tend to cluster with other people who think and act similarly to themselves, so however your sample differs from the general population, that’s an indicator of how you yourself differ from the general population.
So, to use Cowen’s example, if you ask a man how many sexual partners he’s had, he might have a strong incentive to lie, either downplaying or exaggerating his history depending on who you are and what he wants you to think of him. But his estimate of a “typical” number will still be influenced by his own, and by that of his friends and acquaintances (who, because of the selection effect, are probably more similar to him than the general population is). “When we talk about other people,” Cowen writes, “we are often talking about ourselves, whether we know it or not.”
Another reason that we tend to overcount our own data is the availability heuristic. A person has a lot more information about eirself than about any other given person, in general.
Yes, for sure. That’s more or less what I meant by “your own data is much more salient to you.”
A cheater is likely to think that everyone cheats, though I’m not sure which way the causality points.
But if someone thinks she’s surrounded by idiots, is she more likely to be an idiot or smart?
To quote Jake Gittes (Jack Nicholson / The Two Jakes): “….In this town I’m the leper with the most fingers.”
Basically, she may not be all that smart, but she may very well be the smartest one in her group.
To quote an adage, of which I don’t know the source: “In a country of blind men, the one-eyed man is king.”
And that’s why you ask one robot what he thinks the other robot will say when you ask which door is safe to open.
So to the clustering. . . What if we ask the girl if she thinks other people are liars, and she says “No”. Should we assume that she is honest and hangs with mostly honest people? Or could we assume she is a liar and hangs with mostly dishonest people, and has lied to us with her response?
And what if we ask the girl if she thinks other people are liars, and she says “Yes”. Should we then assume that she is a liar and hangs with mostly dishonest people, but for some reason has answered truthfully on this occasion? Or should we assume she is honest and hangs with mostly honest people and for some reason has lied to us with her response?
Suddenly my head hurts.
I’m very happy that I have the sort of readers who won’t let a potential paradox slink under the radar!
And what about the girl who stabs logicians? http://xkcd.com/246/
This would work better on people who on some level consider themselves to be near average for the purposes of the specific question. For instance, I spent most of my free time as a child reading, which had the result of having a larger than average collection of words I know how to spell but not pronounce.
So, if asked how often I make verbal mistakes, my response would be something like half again as much as my response if I was asked how often people in general make verbal mistakes.
Also, rudimentary knowledge about the structure of governments and society would tend to skew the results in some cases: if someone doesn’t take drugs, and they don’t think their friends take drugs, they are still unlikely to assume nobody takes drugs, because why would there be so many laws against drugs or a war on drugs or treatment centers for rehabilitation if it was a nonissue?
It’s a good trick, but I think calling it Bayesian is an overkill – this is more of a mind hack and has little to do with belief updating.
Hmmm, this trick may not work for every question, but it might give you greater insight into someone’s character than you might’ve got otherwise. Very interesting.
Pingback: News Bits
Cool.