Slashdot Mirror
Web Geniuses Or Web Dimwits?
ScribeCity writes "The Washington Post has a provocative piece about online experiments at identifying experts. One wonders when someone will come up with a truly effective formula for measuring human intelligence — or take a stab at doing so — that exploits all the stuff people are publishing online." From the article: "This wisdom of the crowd could be outsmarted by what Michael Arrington, editor of the TechCrunch blog, recently dubbed the 'wisdom of the few.' Sites like PicksPal rely on input from the masses chiefly as a venue for auditioning prospective experts, on the theory that these virtuosos could provide even more accurate information and predictions than the crowd. 'If you figure out which ones did the best and get rid of the ones who have no idea, you'd do even better. Distill it down to the people who really know,' Arrington said."
There is a certain logic to this. How many times have "experts" told us screwy nonsense, and had lousy track records, and yet the public at large retained them as experts? Sometimes, the untrained may be able to see things that the supposedly well-trained can't.
;)
Or to put it another way, it all becomes a set of probabilities. If person X has guessed the outcome of something (say, a football game) correctly 80% of the time, then you're safer betting on his predictions than you are betting on expert Y who is only correct 30% of the time. If you aggregate the probabilities and successes, you should be able to develop a model with a high probability of being correct. You'll never be able to gain 100% accuracy, but that's just the nature of the Universe.
Javascript + Nintendo DSi = DSiCade
'If you figure out which ones did the best and get rid of the ones who have no idea, you'd do even better. Distill it down to the people who really know,' Arrington said.
I've always said that elections should qualify each voter's ballot to make sure the decision is made by the people who are best equipped to decide. The first page of a voting ballot should be a questionnaire that asks simple unbiased questions that require the voter to demonstrate knowledge of who or what they are voting on. "What does candidate X say their stance is on abortion?" "When did you first hear about initiative I-456?" "Please specify which political party each candidate below belongs to", etc. The score a voter gets on their questionnaire would then be used as a "weight" factor when counting their ballot, so that people who know the candidates and the issues better get more of a say, which is clearly how things ought to be.
Moderator hint: a comment is neither "Flamebait" nor "Troll" if it is true.
1) Read old threads at +5, new threads at +2
2) If a person has a lot of insightful/informative posts, check their posting history
3) If they are consistently +3/4/5 informative/insightful, add them as a friend
4) add points to friend's posts so they start out +2.
OK, seriously, I don't do that but if I did, I'd see posts of "wise ones" and ignore posts from those that don't make the cut.
Knowledge is how to play a game, intelligence is how to win, wisdom is knowing what game to play.
Places like /. are the basis for these sorts of developments - user-moderated, information-recommendation boards that rank opinion and content based on a number of criteria. Although a number of boards like this one fail or become shills (ePinions), those that survive are models for social recommendation researchers. Discourse analysis is a peculiar human trait, one that computers cannot (yet) accurately perform because our communicative practices are situated in unique, perspective-based contexts, so I'll be interested to see what develops.
"Tu fui, ego eris" - Virgil
My mother is a professor of education at a college in California and this is something that educators have been talking about for a long while. Google for "multiple intelligences" for a lot more information, but basically there's a theory that says that "intelligence" can be divided up into a number of categories and that people tend to excell in one or two of these areas, but few are outstanding in all of them.
The standard breakdown is something like:
Atheletes tend to excell at Spatial and Kinesthetic, while the stereotypical geek is strongest in Logical-mathematical and weakest in Interpersonal.
I'm not sure I completely agree that this is the end-all-be-all for understanding intelligence, but it does provide an interesting look into ways to classify people who might not be "book smart." For instance, a terrific ballerina might not have excellent Interpersonal or Linguistic intelligence, but she certainly has some special "intelligence" that allows her to excell in an area where I would certainly be an abject failure.I encourage anyone interested in this idea of multiple intelligences to poke around and do some research. Again, it may not be the final answer, but it provides an interesting framework for thinking about the topic.
Interested in a Flash-based MAME front end? Visit mame.danzbb.com
Notice also the similarity to the fallacy underlying the articles one sees occasionally, along the lines of "Man Wins Lottery Twice Against Astronomical Odds". (The odds against a particular person winning twice are not the same as the odds against anyone winning twice.)
There is a correct way to do this forecasting, combining multiple inputs. One needs a measure of the track record of each forecaster, and a measure of the degree to which the input forecasts, and their accuracies, are correlated. (The statistical method is formally similar to Generalized Least Squares.) It's a useful technique, which I have seen used successfully in real financial markets. But there are no fancy "experts" and no fawning newspaper articles.
You may joke, but these days anyone who questions the current pseudoscience-dogma-of-the-month tends to get modded -1 when they interject facts into the discussion, so you're not that far off.
You know, that used to be more true. Then Digg came along and took away all the morons. Go check out some of their flamebait stories (politics would be a good start). If you don't echo the group view, you will be modded into oblivion. However, statements like "Bush is teh stupid!" will actually get modded up. Modding is definitely done by sentiment more than any actual insight.
So thanks Digg, for making slashdot better!
This quote, headlined today on google, is instructive: It has always been the prerogative of children and half-wits to point out that the emperor has no clothes. But the half-wit remains a half-wit, and the emperor remains an emperor. - Neil Gaiman I find that 'Experts' are largely chosen based on qualities other than their expertise. Usually they have good personalities and make friends easily, especially with leadership. The leadership prefers to pick and talk with experts who generally agree with their views and look good. So, it is very difficult for me to look at a given expert and think 'Hrmmm...this guy must really know something about X and that is why he is on CNN'. More likely he was friends with a CNN producer.
Experts are great, and their knowledge is valuable. But in making certain kinds of decisions, it is better to tap into the Wisdom of Crowds.
Funtime Candy Wow! - my plan for eventually conquering Japan.
I didn't send you there expecting to trust their assertions. I sent you there because they compile all the research there and source it, which you can independently verify, such as the Malkiel's A Random Walk Down Wall Street. I'm really not asserting a bold thesis here; any independent source on the matter not after your money (consumer advisor Clark Howard, the various academic researchers they list) and some that aren't independent (Warren Buffet and Peter Lynch) will tell you the same thing.
Or, go to MarketWatch and run your favorite mutual fund against the S&P 500 for its life.
Yes, IFA explains the superiority of index funds, but their arguments apply to any index funds, not just theirs.
The person who you shouldn't trust is the John Edward-type mutual fund salesman who will erase the fund family's misses, show you the hits, and say, "See the ones that outperformed? That was because of great management, really! Now, cough up the cash."
Apology to Ubuntu forum.
Given a collection of experts, there are algorithms for judging the reliability of individual experts on the basis of a set of trial tests, and combining the expert opinions so your error rate is guaranteed to approach that of the best-performing expert (roughly 2x worse, worst-case, asymptotically). This is related to the statistical learning technique of boosting. See the discussion here for more technical information, and a link to a paper.
Doesn't that just lead to group-think?
Part of why I read slashdot is for the (slightly) alternate viewpoints.
Awesome furniture, accessories and cabinetry in Santa Rosa, CA: http://humanity-home.com/