William M. Briggs

Statistician to the Stars!

Wisdom Of The Crowds (And Voting)

Ramses III had a petite honker.

Have you heard of Mesd-su-Re? One of the participants of the Great Harem Conspiracy under Ramses III? Probably not. But I need to know the length of his nose right before he died (don’t ask why). I thought I’d invoke the “wisdom of crowds”, i.e. the Internet, and run a reader poll. Send this request to everybody you know, the more the better. Write down what you think the length was (inches or centimeters) and I’ll take the average. Got to be a pretty good guess, right?

Let’s tweak it, make it better. Suppose we run a national campaign to raise awareness about estimating Mesd-su-Re’s nose. TV ads, radio spots, pundits, community organizers, teachers, even bureaucrats all getting the word out about this important subject. That’ll really bring ’em in!

Before we begin, I should tell you that only vile racists guess lengths under four inches. Only those sympathetic with the war on women go short. The right side of history is with the long nose! Elites stand as one. Stars and starlets agree: length matters.

Okay, everybody. Put down your numbers. I’ll wait here.

Argumentum ad populum.

He said what?

Ignore the blatant political prompting and suppose only that I had asked for the length of Mesd-su-Re’s nose. If people had no information, other than the usual olfactory arcana we all possess—e.g., none of us has seen a human nose longer than one meter and noses can’t have negative length—there is no reason to suppose just guessing-and-averaging is helpful. How could it be? Ignorance plus ignorance divided by two is still ignorance. Ignorance-averaging is a fallacy which is actually well known; goes by the name The Chinese Emperor’s Nose. I changed it to an Egyptian Prince here for variety and for a second reason to be revealed below.

The proof the Wisdom of Crowds is the Chinese Emperor’s Nose fallacy is somewhat involved, but here’s a rough sketch. People’s guesses about Mesd-su-Re’s nose will have a minimum, maximum, and some arithmetic mean which lies between (or at one of) these two. If people have no idea about the length (except possibly rough bounds; this is the key) then the mean of guesses is probably near the midpoint, the center of maximum minus minimum. Then regardless of where the real answer lies, the error (distance from guess to real answer) averaged across people will be the same as the error using the mean. In other words, crowds have no wisdom of subjects in which they are ignorant.

Wait! That the Wisdom of Crowds can sometimes provide reasonable predictions is obviously true. It could work in the sense as when an economist throws an equation at a list of stocks which sticks. But that’s (credentialed) luck. Crowd wisdom is also successful when people have some idea, some unbiased idea, of the answer. If individuals in a group had opinions like, “I don’t know exactly what the answer is, but I know or can see it’s X plus-or-minus” then averaging might provide superior guesses to the average individual.

Yet when a crowd is fed biased information the game is off.

Example. You might look at that jar of pennies (long-time reader and contributer DAV reminded me of this example) and know that it can’t contain a million pennies; no, nor a hundred thousand. But we all know pennies and many of us have jars of change, so we could all form a crude but not insane idea of the number. The average of many in this case is likely to be a good guess.

Then imagine a moustachioed slickster standing by the jar whispering, “Psst, buddy. There’s a solid cone of cork in the middle. Only looks like there’s a lot of pennies. Word to the wise.” Finger on the nose and everything. Hey, he might be in on it: could be a hot tip—and if many think so there goes the accuracy of the average (supposing he’s fibbing).

Recapitulation. Wisdom of the crowds isn’t worth squat when individuals are ignorant of the subject matter they’re guessing. Averaging is okay, but only when folks are using unbiased information. The bad news is already well known: spreading misinformation works. People, even groups of them, will come to wrong conclusions conditioning on flawed premises.

So what does this have to do with voting? Well, everything. But that’s for another time.

Meanwhile, here’s the answer. Zero, you racist. Inches or centimeters. The length of Mesd-su-Re’s nose at the end of his life. Ramses had it sliced off for daring to corrupt his (Ramses’s) harem. Ouch. The lesson is: don’t guess unless you have to, and when you do, be less confident.

See also Voting (And Wisdom of the Crowds).

8 Comments

  1. Any kind of politicking dilutes the wisdom of crowds. Even politicking done with good intentions.

  2. Actually seems like zero skill and zero knowledge is the very best we can hope for from voting. We’d have low odds of getting a psychopath. What little skill that exists, however, seems to be in choosing the guy who promises to take stuff from others and give it to me. That means the guy who wins is likely to be the one who can convince others to join his scheme. He is far more likely to be psychopath. In the limit, all politicians that are known beforehand to the voters are almost surely psychopaths.

  3. I feel the same way about the global warming model ensemble average.

    I don’t get why this is a good metric.

    I would equate each model to one persons guess about nose length.

    Some might turn out pretty good and some are just plain awful.

    Why include the really awful ones in the average?

    We could improve on this using the subjective Olympic judging technique – throw out the two highest and two lowers values and then average the rest – that would make just as much sense.

    I like to compare each model on its own to actual observations (like Lucia does) and also to the ensemble (but lets throw some models off the island because they are just so wrong).

    By the way – are you asking for the horizontal length extending above the lip or the hypotenuse length from the top of the nose to the tip of the nose? (grin).

  4. Zero huh? Just because it was, allegedly, sliced off?

    Almost certainly an incorrect answer, zero that is — even if it was sliced off…because there would invariably be some bit of nose left.

    One needs to define what constitutes a “nose” (or nose parts) and what doesn’t. Not to mention (though it was mentioned above) the criteria for measurement.

    This is much like the ancient Norse myth involving Loki & the Dwarves. In some deal going not quite all good, the Dwarves conclude Loki owes them his head, to which Loki agreed. However, just before the payment was served Loki reminds the Dwarves that he only agreed to forfeit his head, but none of his neck. Since the Dwarves & Loki (or anyone else) couldn’t agree on where one ended & the other began, Loki kept his head and neck (though the Dwarves did sew his mouth shut after that).

    If the ancients could understand the concept of indistinct borders, so ought we in fathoming a nose’s boundaries.

  5. Briggs

    January 29, 2014 at 3:38 pm

    Ken,

    It was zero. I checked just yesterday in my time machine.

  6. Detached or attached, the nose belonged to Mesd-su-Re and retained it’s length irrespective of location, assuming no shrinkage due to lack of viability, of course. Your point is confirmed even if your conclusion is not.

  7. This is nothing but a dastardly plan to make me read the federalist papers wherein I will find that the US constitution requires a voting populace to be well informed and guided by Christian ethos.

  8. Add to this the anchoring effect: http://en.wikipedia.org/wiki/Anchoring_effect

    I don’t know whether anyone has done any studies on this, but I assume that this should also have the potential to throw off “wisdom of the crowds” type estimates.

Leave a Reply

Your email address will not be published.

*

© 2016 William M. Briggs

Theme by Anders NorenUp ↑