Selecting Leaders with Random Sampling and Standardized Testing

One thing you learn when investigating methods of selecting leaders is that many of them don’t involve voting at all. Hereditary monarchy is an example of this. As we have better institutions today, it’s easy to dismiss but “let’s put his first-born son in charge” is a much better selection procedure than selection via internecine warfare. And modern liberal democracy was an advancement over hereditary monarchy, in many respects. The incomparable Scott Alexander makes this point well:

A democracy provides a Schelling point, … an option which might or might not be the best, but which is not too bad and which everyone agrees on in order to stop fighting. … In the six hundred fifty years between the Norman Conquest and the neutering of the English monarchy, Wikipedia lists about twenty revolts and civil wars. … In the three hundred years since the neutering of the English monarchy and the switch to a more Parliamentary system, there have been exactly zero. … Democracy doesn’t always perform optimally, but it always performs fairly, … and that is enough to prevent people from starting civil wars.

Hereditary monarchy provided a Schelling point, too, but it was not a very good one. The nature of paternity means there were always potential rival claims, some good, some-just-good-enough, others completely fraudulent, each providing an excuse for an economically disastrous civil war.

And even if we grant that the characteristics that make a good leader are to some extent heritable, regression to the mean results in the unfortunate tendency of the great to sire not-quite-as-great offspring. One can speculate about forms of hereditary monarchy which mitigate these problems somewhat. Maybe one in which the monarch is cloned every 30 years, their genome a state secret. A premise, perhaps, for a science fiction story, but not a serious proposal. Monarchy’s time has passed. But thinking of it as just another selection procedure gives us a bit of parallax. If we improved upon hereditary monarchy with representative democracy how might we improve upon representative democracy?

When baselessly speculating, it’s often best to start simple. What’s the simplest possible means of selecting a representative? Random selection, also known as sortition, is a good candidate. Unfortunately, it’s likely to work very poorly if applied naively to a modern society.

Due to the fact that the average person is mediocre, naive sortition will always be less selective than current representative democracy, and if implemented the quality of the nation’s representatives would certainly decline, both because of a lack of experience and a decrease in median cognitive ability. The economist Garett Jones’s describes well the extent to which the quality of a nation’s institutions is limited by the average IQ of its citizens. By removing all filters for intelligence and experience (almost any competitive selection procedure selects at least partially for cognitive ability) naive sortition may well vastly amplify this effect, and perhaps disastrously so.

If we ignore this potential flaw for a moment, it is worth noting that sortition does have some interesting properties that most other systems do not. For one, it provides some resistance to corruption, at least before selection. In a discussion on HackerNews, I explained this with a very Silicon-Valley analogy, and hinted at a possible solution to the mediocrity problem:

Suppose a billionaire is looking for someone to marry. They want a partner who is interested in them for more than just a paycheque. Say gold-diggers represent just a very small fraction of the population but are extremely good at feigning fascination in pursuit of their goals and finding themselves in the company of billionaires. In fact, they are so good at ingratiating themselves with billionaires that any potential partner this billionaire picks from the people she or he knows is likely to be a gold digger. What is a good but non-optimal solution? The billionaire should pick 1000 random people of the gender he or she prefers and make their selection from this pack. This cuts out the influence of the adverse selection effects the billionaire experiences just from being a billionaire. If all they’re looking for in a partner is someone in at most in the ~99th percentile of people this approach will work with little cost, as candidates can then be selected from this pool using conventional means.

Conventional elections select for those most interested in getting public political power and allow vested interests to influence potential candidates before they are elected. As with naive sortition every adult citizen has an equal probability of being selected as a representative, this influence is watered down to the point of impotence.

So a question arises: Can we preserve the corruption-limiting properties of sortition while retaining, and ideally, vastly increasing selection for intelligence and other perquisites of competence? One reason to be hopeful is sortition can be combined seamlessly with other selection procedures — including conventional elections.

My proposed system combines three forms of selection: standardized testing for general intelligence for every member of a nation’s populace, random sampling of candidates from the 99th percentile, and a conventional election. I do not claim this system is perfect or even particularly likely to work. I do, though, think it is a more interesting alternative/adjunct to democracy than it seems initially, is perhaps worth trying if you ever find yourself founding a new nation, and that many obvious objections to it have equally obvious solutions. I would also claim this is superior to those forms of Epistocracy which rely on knowledge tests rather than reasoning tests. Such tests would likely quickly degenerate into tests of ideological purity.

A Note on IQ

One of the more unfortunate superstitions of our time is the belief that everyone is born a blank slate of equal intellectual capability, and thus every difference in outcome is the result perverse socialization and lack of education rather than genetic luck. This is false. Denying the validity of IQ and its heritability is about as scientific as denying global warming. Yet due to unfortunate late-century ideological trends in the humanities this denial is the rule rather than the exception. It is important to note that the relevant disciplines have reached a consensus: IQ is a good measure of general cognitive ability, it predicts academic and career success, is greatly influenced by genetics, and is one of the best predictors of an employee’s future productivity. I assume, here, that the scientific consensus is correct. (If you deny the scientific consensus or would just like a good primer on the evidence for the importance, validity, and heritability of the general factor of intelligence, I highly recommend Dr. Stuart Ritchie’s excellent book on the topic Intelligence: All That Matters.)

The First Filter: A Standardized Non-Verbal IQ Test

So how would one go about giving every citizen in a country an IQ test? Rather easily. The vast majority of people in most countries take scholastic aptitude tests in their final year of high school. In my system, we would require all citizens to take such a test at the age of majority. In order to avoid any cultural bias, an entirely symbolic language-free IQ test such as Raven’s Progressive Matrices would be ideal.

One obvious objection: this filter provides great power to whoever writes the test, which would incentivize infiltration of whatever institution designs these tests, leading to tests that select less for IQ and more for ideological conformity. This is the reason that using Raven’s Progressive Matrices is vital. Above is an example of a Raven Progressive Matrices question. The entire test consists of nothing but such questions with no written instructions. So long as a tradition of using nothing but entirely nonverbal questions for this purpose can be enshrined and enforced, it is difficult to see how such a test could be corrupted to select for ideological compliance.

Another check you could add would be to forbid universities and businesses from using any other standardized test for admissions and selecting employees. If the test is used primarily for selecting employees and students and for politics only rarely, corporations (and to a lesser extent universities) would have great incentive to ensure the test remains a useful measure of the intelligence of those who take it. Such powerful institutions should be capable of keeping it from swaying too far from its intended purpose.

A second obvious objection: Even if one agrees that nonverbal IQ tests will be difficult to corrupt, this will not prevent corruption of the grading mechanism. This system incentivizes organizations to infiltrate the grading institution, give their cronies the highest IQ scores in the country, and then take over the government. This would all be perfectly true if I was advocating creating a governing body peopled by whoever got the highest scores. I am not making that proposal, however. An IQ test is only the first filter. The second filter, random sampling, makes such an infiltration highly uneconomical.

The Second Filter: Random Sampling of the Highest-Scoring 1 Percent

About 4 million Americans turned 18 this year. To simplify things, let’s ignore population growth and decline. That’s 80 million people tested over the course of 20 years. So after the first filter, this leaves us with a pool of 800000 people with IQs in the 99th percentile. If a governing body of, let’s say, 535 people is randomly selected from this pool, you would have to falsify the score of an impractically large number of cronies to have a significant chance of taking control of the government.

But can’t you just move the objection up a level? If corrupting the grading mechanism is uneconomical then people will just focus on corrupting the randomization mechanism. One potential solution is to give each individual a number, sort of like a social security number but publicly available from birth, then use an independently verifiable and incorruptible source of randomness to select numbers from this pool of 800000, preferably an observable physical phenomena such as the variations in brightness of a particular section of the sun in a precise time period or the coordinates of the next N type Ia supernovae that are observed.

An exploding star is a hell of a Schelling point.

Another problem is people selected in this manner may not agree to the job. The classic economics solution of offering sufficiently high salaries should help combat this. And imagine, if you were chosen by an exploding star to help rule the nation you love for 4 years (for a great salary with an unbeatable pension) would you say no?

The Final Filter: [Insert Conventional Election Procedure here]

This final filter is another round of selection. You could just allow this body to elect its own leader or use further sortition to elect people from this body into various higher level positions. Or you could allow it to splinter into groups around various candidates, and then have a conventional election from there. For a more parliamentary system, you could combine this with a body selected entirely using sortition, requiring an agreement between a council of the wise and a council of the average to pass laws.

There are many options, all of which have advantages and disadvantages. However, the combination of standardized testing and random sampling seems very powerful. Should you ever found a nation, you should at least consider it as an ingredient.