[FRIAM] Amazon.com: The Wisdom of Crowds: Books: James Surowiecki

[Robert Holmes]

So if Surowiecki's thesis is correct, our discussions are almost certainly
more reflective of reality than his because we're a crowd and he's a lone
expert. But then if we decide his thesis is flawed and maybe experts do have
more insight than crowds then - because he's the expert - we should accept
his thesis. But if Surowiecki's thesis is correct...

Robert

On 6/8/06, Bill Eldridge <dcbill at volny.cz> wrote:
>
>
> All good and well, but I think Surowiecki weakens his case
> by mixing examples that don't quite rightly belong together,
> cherry-picking some behavior. It's an attractive populist
> message, which should start up the warning bells to be
> careful at least.
>
> First, his first big example is the crowd "picking" the weight
> of a hog by having the mean almost exact (say 1187 pounds
> to 1188 pounds actual). Uh, okay, that proves what?
> I would guess that if you went from county fair to county
> fair, the mean of the means would deviate quite a bit, but
> the averages of the 10 best hog breeders would be much
> more accurate and deviate less. Just a hunch that the item
> in question was a fortuitous instance. Try it with marbles
> in a very large jar rather than the weight of a hog. By what
> I take is the author's thesis is that we could get 6 billion
> humans to vote on the weight sight unseen and they'd get
> it right because they're a wise crowd, or perhaps I'm not
> reading him right - he doesn't lead me towards understanding
> the wisdom.
>
> The lack of rigor is added to when you hear that a group of
> traders in Iowa pick the Presidents better than the Gallup poll.
> Of course they do, or you wouldn't be hearing about them,
> you'd be hearing about a group of stock car drivers in Alabama
> who pick better than the Gallup poll. Of how many crowds
> trying to be correct, how many show Surowiecki's "wisdom"?
> 1/1000? Now, it might be that Iowa is situated enough between
> Montana myopia on the right and New York myopia on the left
> that it balances to be less deceived than those in other places.
> But meanwhile, a group somewhere else like Texas, just as authoritatively
> a "crowd", would be mixing signals and betting wrong - getting it
> right in the Bush years, getting it wrong in the Clinton years perhaps.
> (And considering that an election might be decided by a few electoral
> votes or a single Supreme Court justice, what really does that prove,
> except that you might be as predictable as a coin toss?)
>
> Now, I thought the American public was very wise at the end of
> the Clinton years when it was not swayed by the Lewinski scandal.
> But if I were Republican, I might think the public very stupid and
> gullible.
> We're into value decisions, and is the public more or less consistent
> on values versus deciding how much something weighs and whether
> the sky is a particular hue?
>
> The Google example is something different. You're able to model
> what people expect by modelling lots of people. Especially if people
> are more satisfied with 20 different answers, you can cross-correlate
> the answers from a million people and make sure the top 20 don't
> resemble each other, and that will be much much more satisfactory
> than if every one points to the same URL. But this only tells us that
> it's useful to model people's expectations if we want to derive an
> answer that people expect (and want). An important result, and I
> like Google, but very different from crowds being inherently "wise".
> Crowds are inherently human, and Google caters to that humanity.
> But its results on WMD's may not satisfy a right-winger or a left-winger
> or neither, though folks in the middle might be happy. I just clicked
> "I Feel Lucky" for "Monica", and there was some singer I'd never heard
> of - I thought the name Monica had been retired with Lewinsky, but
> I'm wrong. So Google is more like having the house working for you
> rather than against you in Vegas (unless you're doing clicks for ads,
> but let's not go there). But that still doesn't address "wisdom", it
> addresses "humanness". We're doing a bell curve on human expectations,
> not on right vs. wrong.
>
> Now what happens if Google aggregates Pakistan with Germany,
> or instead separates Asia from Latin America from North America?
> Is Google deriving wisdom to appeal to all sectors equally, or splitting
> by domain ending/country (certainly I get different answers depending
> on what languages I check, but does it change based on my IP address
> as well?). If I'm an African Googling from New York, will I get the
> same satisfactory results as if I'm Googling from Dakar, or will
> Google miss my demographics?
>
> Some of the mathematical work on deriving a neutral stock portfolio
> is very interesting and very similar in some respects to finding the mean
> "wisdom" of the human crowd (or subsets thereof). You end up with
> different classes of investors with different risk and gain requirements,
> so different appropriate investment tools. And you can track
> performance of various popular stocks, mutual funds, people who bet
> the Dow vs. those who are conservative or risky. If Surowiecki really
> wants to test his hypothesis, he should get together the minimum number
> of people that make a "wise" crowd, get them to choose stocks, and
> then figure out the mean choice, and they all take the plunge.
>
> I apologize if I sound negative, but I think he would have been better
> off trying to model the different agents involve and develop a complex
> theory that might have some simplistic analogies. People with experience
> will on average do better than people with no experience in most fields.
> People with accurate information will do better than those without
> accurate information, most of the time. But there are some types of
> decisions that are simply counter-intuitive, and this might depend on
> culture and time - whether the earth is flat, whether rock 'n roll is
> harmful, whether rent control is healthy, whether gay marriage is
> "okay". Some of these judgments have measures, some not.
> There are some areas where expert knowledge is of no help,
> but that's a far cry from saying lay knowledge in aggregate is
> always more "wise" than expert knowledge to whatever measure.
>
>
> Giles Bowkett wrote:
> > I haven't read it but Surowiecki presented at South By Southwest this
> > year (because of the Web 2.0 connection) and I've listened to the
> > podcast of this presentation a couple times while working out. I
> > definitely recommend it and the book's on my list to read.
> >
> > Judging from that presentation, it sounds as if the book would have
> > probably had a more accurate but less catchy title if he had called it
> > "The Wisdom Of Decision-Making Markets," because what he says in the
> > speech is that it's really about how aggregate decisions, summed from
> > the decisions made by all individuals in a large group of people
> > competing against each other, consistently outperform the decisions of
> > the best-informed experts in those groups.
> >
> >
>
>
> ============================================================
> FRIAM Applied Complexity Group listserv
> Meets Fridays 9a-11:30 at cafe at St. John's College
> lectures, archives, unsubscribe, maps at http://www.friam.org
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: /pipermail/friam_redfish.com/attachments/20060609/6df24a4c/attachment-0001.html