Quacks who write software make us all look bad

By now I'm sure that many of you have read about the research which claims that people aren't smart enough for Democracy to flourish. This was big news and made the rounds (including here on Reddit). The main researchers listed were Dunning and Kruger and I don't think anyone disputes their credentials.

Except that this is a science article, not a link to the actual research. In particular, I was intrigued by this by a reference to a software simulation validating their results. That piqued my curiosity.

Mato Nagel, a sociologist in Germany, recently implemented Dunning and Kruger's theories by computer-simulating a democratic election ... some text omitted ... When such an election was simulated, candidates whose leadership skills were only slightly better than average always won.

Who the heck is Mato Nagel? As far as I can tell he's a medical doctor and not a sociologist (the address on that site is the same as the one listed in Mato Nagel's paper below). However, being a software engineer, I wanted to see that computer simulation (I haven't found it yet), but I found Dr. Nagel's paper and that's what I'm asking about. It screams BS to me due to its dumbing down of a complex problem space. In particular:

  • It assumes that leader ships skills have a normal distribution (evidence, please)
  • It assumes that people's inability to judge a leader's ability translates to not voting for that leader (again, no evidence is offered)
  • It ignores how elections are actually conducted (such as voters preferring a particular party)
  • It makes multiple references to "Fauceir theory" (see below)
  • It cites as an illustration a non-democratic society not speaking the French language of their leaders (how is this relevant?)

Regarding faucier assumptions, the paper repeats the term multiple times but doesn't define it. I assumed that this was simply a well-known term, but I had trouble finding it, until I found fauceir-evolution.org. The domain is registered to Dr. Nagel and the bits of material I found about this are all published by him.

The paper also had this:

First, we define a scale of leadership qualities in analogy to intelligence. We assume that for each individual a Capability Quotient (CQ) can be determined. That is, the sum of all an individual's leadership qualities divided by the average sum of such qualities. To better handle the figures, this quotient is multiplied by 100. Then the better-than-average leader possesses a CQ greater than 100 while people with a CQ below 100 are not recommended to become leader at all. Next we assume that the distribution of CQ is the same as Intelligence Quotient (IQ). Then the normal distribution's mean is 100 and the standard deviation 15 ...

There are no sources cited for any of this, nor do they explain why CQ should have the same distribution as IQ (which is in itself a problematic construct). I admit, of course, that like Dr. Nagel, I am not a sociologist, so perhaps these are perfectly appropriate assumptions and I simply don't have the background to judge them.

Dr. Nagel's paper was published at Maxwell Science Organization (MSO) and I was curious about them, too. and And while hardly a smoking gun, I'm finding multiple sites listing MSO as not requiring peer review and generally producing very low-quality papers. In fact, several sites claimed that MSO is a fake science publishing organization.

So is Dr. Nagel a quack? There's a strong whiff of BS about all of this and it seems to have attached itself to the very prominent (and respectable) Kruger and Dunning (though I don't think this is their fault).

5 Comments

It is possible that Dr Mato Nagel is a polymath and has doctorates in both medicine and sociology, though in this case I think your explanation is more likely.

Heh, I happened to see the headline of your reddit post about 5 seconds before skimming b.p.o, and went "whoa, everybody's talking about this guy today" :)

in reference to leadership distribution, totally anecdotal I guess but if you look at military history you will notice that a 'squad' (or whatever the smallest command unit is) is generally 8-14 soldiers plus a squad leader. I've notice that teams once they are more that 6 or 7 in size seem to require a significant leadership presence or they become ineffective. Teams under this size can be self organizing as long as the people in it are reasonably peaceful with each other. Also, I note that the mentality of the group seems to change once the size reaches above that 6 to 7 number. Larger groups seem to need some organizing principal to keep them aimed in the same direction. This has lead to me believe that human social organization seems to require one decent leader for every 8-15 people. Given that and what we know of evolution its not beyond belief that leadership skills could be distributed evenly in a population based on our mental and social needs. But am not a social scientist, just have interacted with a lot of groups both large and small and spent a lot of time thinking about this.

IMHO for democracy to function we don't need to always choose the best possible leaders, just ones that are good enough and in the presence of a strong political and legal framework to guide them and keep them in line. As a US citizen I personally believe in our Constitutional framework that divides power in such a way as to account for human greed and weakness rather than to hope that leaders are perfect. Software projects follow a similar trend, which is why I notice startups with strong technical skills but generally weak leadership skills do best when the team size is small, and then when they get bigger you need a regime change generally to account for the need of more 'corporate style' leadership otherwise the company tends to drop productivity wise.

I didn't bother to check out any of the links you provided. Dr. Nagel's publications are probably not worth the effort. But I can provide one piece of evidence you asked for: The "leadership capabilities" have a normal distribution per definitionem. The CQ is constructed like the well known IQ, which is calibrated to have a normal distribution.

Thus, you select a function cqraw(...) that combines an weights all factors that you want to contribute to you CQ into a single number Craw. You must make sure that it is a (quasi) continuous function with no theoretical min and max (or a theoretical min and max that are sufficiently unlikely to be reached). You apply this function to a large number of people, your "population". This gives an arbitrary distribution of Craw. Then you define a function cqcalib() with CQ = cqcalib(cqraw(...))_ such that cqcalib(...) < 100 for 50% of your Craw distribution, etc.

@ Ovid, You are probably aware that your article is merely an ad hominem (http://en.wikipedia.org/wiki/Ad_hominem), so I ask why. What are the reasons to employ an ad hominem in scientific discussions?

Let me help you to better understand the scientific method. A model, any model, makes assumptions and based on these assumptions predictions are calculated. These assumptions more or less mirror the real word.

For instance, you probably recall your physics lessons. The so important Newtonian calculations about free fall are based on two assumptions that never really happen on Earth (1) the mass is shrunk to a single point (http://en.wiktionary.org/wiki/point_mass) and (2) vacuum. If not in vacuum you cannot approximate even remotely the free fall of a feather Newtonianly.

That said, you have (put aside logical fallacies such as an ad hominem) two ways to refute a model calculation scientifically:

  1. You can reject the basic assumptions as not real, outlandish, or not from this world.
  2. You can prove the calculations wrong.

If you succeeded in the former the research would be irrelevant. If in the latter, it would be outright crap. Right.

Though you allege the second, your argument just tried the first, but failed even there. You probably don't realize that you have the burden of proof. You cannot ask for evidence. Please remember Newton. There is no evidence that free fall on Earth happens in vacuum (on the contrary there is strong evidence against it) and the assumption of a point mass is even outlandish, still his calculations had a great bearing on classical mechanics.

Finally, I'd like to demonstrate that you didn't even read the paper carefully. You write:

“It assumes that leader ships skills have a normal distribution (evidence, please)”

If you had read the paper carefully, you would have noticed that this assumption isn't even necessary. It is just chosen for convenience. The model calculations work with any distribution, as it is demonstrated in this very paper when calculating second step elections among the elected.

Last but not least I hope you won't take offense and my comment will help you to improve your scientific methodology.

About Ovid

user-pic Freelance Perl/Testing/Agile consultant and trainer. See http://www.allaroundtheworld.fr/ for our services. If you have a problem with Perl, we will solve it for you. And don't forget to buy my book! http://www.amazon.com/Beginning-Perl-Curtis-Poe/dp/1118013840/