Democracy, Current Events
Is Democracy Going Out of Fashion? Are All Heads Together Always Smarter than Just Some Heads?
I’m among his targets, though, like many critics, he confuses epistocracy with technocracy. Still, his main counterargument is this:
Which brings me to the core of the case for democracy. It’s not that the voters are always right; it’s that, in aggregate, they’re wiser than the experts. Or, to put it another way, democracy may not be perfect, but it is preferable to oligarchy.
I agree that democracy is better than oligarchy. But does the rule of all beat the rule of the many but not quite all? That’s the interesting question.
The best epistemic argument for the rule of the many is the Hong-Page Theorem. As Ilya Somin summarizes the theorem:
Some scholars argue that aggregation can work especially well if participants have diverse views and abilities. When a large and diverse group seeks a solution to a problem, it can often make better decisions than a smaller, more expert group because it can pool its diverse collective knowledge which, in the aggregate, is greater than that of the smaller group.
Landemore’s recent book Democratic Reason tries to use the Hong-Page Theorem to show that democracies are smart. She has an ambitious thesis. She intends to show that democracy outsmarts epistocracy, that the rule of the dumb many usually beats the rule of the smart few. As she summarizes her thesis, “for most political problems, and under conditions conducive to proper deliberation and proper use of majority rule, a democratic procedure is likely to be a better decision procedure than any nondemocratic procedure, such as a council of experts or a benevolent dictator”. By “better,” here, she means that democracies are likely to outperform non-democracies, producing better outcomes, where such outcomes are measured independently of the procedure itself. Notice also that Landemore says “any” non-democratic procedure—this is what makes her thesis so ambitious. She’s arguing that even mild forms of epistocracy, say, an epistocracy that excluded only the bottom 5% of citizens from voting, perform worse than full democracies with universal suffrage.
I think Landemore, and others who make use the Hong-Page Theorem, tend to misapply it. In Chapter 7 of Against Democracy, I spend a great deal of time examining possible ways in which democracies could or do make competent decisions despite the average, median, and modal voter being incompetent. I include a lengthy critique of Landemore’s (and others’) application of the Hong-Page Theorem. Today, I’ll discuss just one small part of this critique.
One worry, among many, is that Landemore never really even attempts to show that the Hong-Page theorem calls for letting everyone vote, rather than letting many people, but not quite everyone vote.
Two heads are sometimes better than one, but that does not mean that all the heads are always better than some of the heads. This seems to be Landemore’s essential problem. As far as I can tell, Landemore is much more optimistic about the Hong-Page Theorem’s ability to defend democracy than Scott Page himself is. That doesn’t necessarily mean that Landemore is wrong. Sometimes people who devise a theorem don’t recognize the real power of that theorem. However, when we see why Page didn’t himself draw the same conclusions as Landemore, we’ll see some grounds to suspect she’s overextending the theorem.
Page says there is value in cognitive diversity. “Cognitive diversity” means include diverse perspectives (“ways of representing situations and problems”), diverse interpretations (“ways of categorizing or partitioning perspectives”), diverse heuristics (“ways of generating solutions to problems”) and diverse predictive models (“ways of inferring cause and effect”).[i] The Hong-Page theorem says that when it comes to making accurate predictions, increasing the amount of cognitive diversity among decision makers is as important as increasing the predictive power of any subset of them.[ii] That is, sophistication and cognitive diversity are equally good.[iii]
But Page himself explains that crowds are not always wise. Crowds can make bad, even mad, decisions, either when there are systematic biases or when a tendency toward conformity in deliberation leads to less accuracy and diversity. So, for instance, Page says that if individuals are unduly influenced by others who are charismatic but whose ideas are inaccurate, then group accuracy will be poor.[iv] We should thus ask, are real-life voters are influenced by charisma and political showmanship, or whether they are instead dispassionate, rational truth-seekers not so easily beguiled?
Page holds that increasing diversity can be a bad thing when people’s predictive powers are poor. Page says that for the Hong-Page theorem to take hold, the individual decision-makers must be fairly sophisticated, if not as sophisticated as experts. Page’s modest conclusion is that having many diverse and good predictors tends to be more successful than having just a few excellent predictors.[v] Page says in a lecture, “If we don’t get collective wisdom, it’s going to be because either people lack sophistication—that’s the garbage in, garbage out—or they lack diversity.” He adds that people need not just diverse information, but diverse and good “models” or methods to interpret that information.[vi] He writes: “For democracy to work, people need good predictive models. And often, the problems may be too difficult or too complex for that to be the case.”[vii] Page doesn’t argue that having many diverse but stupid predictors always works better than having fewer, smarter, but less diverse predictors. On Page’s account, highly unsophisticated but diverse crowds do not make good predictions.
It is thus important for Landemore to try to prove that the average or typical citizen is sufficiently sophisticated about politics. But, as we saw in chapter two, the evidence tends to show that most citizens are highly unsophisticated about politics, possessing little of what Page or Landemore would call a mental model. Many of them are Hobbits.
It’s a puzzle why Landemore interprets the Hong-Page Theorem as implying it’s best to have all adult citizens participate. The Hong-Page Theorem tells us that diversity is good, but it doesn’t imply that it’s literally best to have every single citizen vote, or even to have most of then vote. The theorem instead says that two heads are often better than one, that 5 million are often better than two, but sometimes 200 million are much worse than 5 million, or 195 million.
As far as I can tell, Landemore never actually tries to show that democracy beats all forms of epistocracy. Instead, at most she tries to show that democracy with universal suffrage beats forms of epistocracy in which only a tiny number of citizens are allowed to vote. But that isn’t enough to generate her conclusion. Landemore never seriously considers whether a limited form of epistocracy, say one in which the most ignorant 5% of citizens are excluded from voting, would outperform democracy with universal voting. There’s nothing in the Hong-Page Theorem that says universal participation always beats more limited participation.
An epistocrat can (and should!) accept the Hong-Page Theorem, but conclude not that we should have democracy instead of epistocracy, but instead have an epistocracy with a large and diverse group of epistocratic voters. Indeed, the Hong-Page Theorem is one reason why I would favor having larger rather than smaller epistocratic bodies.
Many heads are sometimes better than fewer heads, but that doesn’t mean that many heads are always better than fewer heads. To return to my earlier point: A collection of third-year economics Ph.D. students might collectively know more about economics than one star professor. But that star professor might easily know more about economics than an entire high school. The American public as a whole makes systematic mistakes about economics, including most of the mistakes Adam Smith warned us not to make back in 1776.
Landemore says, on the basis of the Hong-Page theorem, that she prefers the rule of the many over the rule of the few. But that’s a misleading way of putting it. Most epistocrats also want the rule of the many. What Landemore really prefers is the rule of everybody over the rule of the many-but-not-quite-everybody.
[i] Page 2007, 7.
[ii] One problem with Page’s work is that he tends to treat experts as non-diverse, as if they all have the same models of the world. But perhaps Page’s work makes a better argument for having many diverse experts make decisions rather than for having many diverse non-experts make decisions.
[iii] Page’s models work best for cases where issues are easily quantified or where qualitative answers to questions can be easily separated into distinct categories. It’s not as clear how they apply other kinds of issues. Note also that Page does not mean, e.g., that including more people from different vocations or different races tends to lead to group wisdom. Rather, what he means is that having many people with diverse sophisticated models of the world tends to lead to group wisdom. Also, insofar as uneducated people tend to have simplistic, unsophisticated models of the world, their input into collective decision-making tends to lead to less accuracy. Page seems to recognize this at times, but then often appears to overreach in how well his models of diversity apply to actual democratic decision-making. See Tetlock 2007 for a quick but sharp criticism of Page on this point.
[iv] Page 2007, 212-214, 391-391; Page and Lamberson 2009.
[v] Page 2007, 346-347. Page2007, 147) says, “The best problem solvers tend to be similar; therefore, a collection of the best problem solvers performs little better than any one of them individually. A collection of random, but intelligent, problem solvers tends to be diverse. This diversity allows them to be collectively better. Or to put it more provocatively, diversity trumps ability.”
[vi] Page 2012.
[vii] Page 2007, 345