Epistemocracy

Epistocracy and Epistemocracy don’t appear to have set meanings. In The Black Swan, Nassim Nicolas Taleb uses the latter ― epistemocracy ― to define a trait of leadership where the leader has epistemic humility (i.e. is comfortable with saying I don’t know). That’s not necessarily a system of government, as it’s compatible with democracy and autocracy. However, the complete opposite meaning is also used: a leader or system of leadership that claims its own infallible revealed wisdom: theocracies, ideologies and cults. Epistemocracy, then, can be summarised as any leadership type where relationship to knowledge ― whether it be certainty or humility ― is defining.

If you were to Google “epistocracy”, the first thing you would get back would be ‘noocracy’, (sounds like ‘no-ocracy’). This is to be ruled by the wise or ― as Plato originally specified ― philosophers (sounds like ‘people like me’, conveniently). There are so many problems with this that it’s difficult to know where to begin. Starting with the question “who is to say what is wise or sufficiently verbose to be called a philosopher?” we can get a naive and a cynical answer.

The naive person might assume that discussion about what wisdom is would essentially evolve into a sort of democracy, as the people decide who is wise in the areas that are important to them. The cynic might assume that wisdom will be defined to suit the people already in power; I mean, they must be wise, they’re already in power. Isn’t that how a meritocracy works?

In Plato’s noocracy, it is difficult to know how you can choose the genuinely wise. Much like the UK, where we can basically pick from Eton alumni and Oxbridge graduates, any selection process for the ‘wise’ will almost definitely end up looking like a democratic plutocracy (choose your rich person to rule you) or oligarchy (choose from this other narrowly defined group to rule you).

Epistocracy, as I mean it, is a form of democracy. It is democracy as articulated by Jason Brennan, in a podcast (because I’ve not read the book). The summary is this: the electorate is not actually the public, but a statistically idealised version of the public. The methodology is like this: you vote, as normal, but you also fill out a demographic survey and a political quiz. This data can then be used to weight each demographic: in what direction do their votes tend as they score better in the political quiz? Each demographic is then ‘idealised’; it is assumed everyone scored 100% in the quiz and the votes are tallied as if this were the case. In short (and we’ll come back to this) it is stating how everyone would have voted, if you waved a magic wand to make them “perfectly” informed.

It’s an epistocracy, because it creates a knowledgeable electorate (epist-) and chooses a leader according to that (-ocracy). I think the principle here is great, but it’s practically unworkable. And it’s worth having a look at why. But, before we do, it is also worth looking at why it’s worth having a discussion about refining democracy at all. What’s wrong with democracy?

Democracy is not a singular thing. In the UK you only vote for the legislature (Parliament, the US equivalent is voting only for Congress). The party that gets the most seats in the Legislature then gets to appoint the executive (Government). In the US, you vote for both. In the EU, the electorate (public) vote for the Legislature (EU Parliament) and there is a European Council composed of Heads of State of the various members (so they are also elected); these two bodies then appoint a European Commission. So, there’s a number of ways it can work, but it is defined by a body that legitimately has some power being accountable to the people it governs. What’s wrong with that?

Jason Brennan and Nick Cohen open with perhaps the most well known: an ignorant electorate. That’s not meant as a dig or a way of delegitimising election results I don’t like. It is actually an observation of the longstanding failure of the education system, media outlets and politicians to make accurate representations.

Brennan says that research has consistently found, since the late ‘50s, that the average voter: tends not to know who their representative is in the Legislature; in the US voters tend not to know which party controls the Legislature; has approximately a 6 month memory of economic performance and no way of knowing who’s responsible for that; will defend or denounce policies based on which party they think has put it forward (which shows both a tribalistic ‘my team’ mentality and an inability to distinguish between Right and Left philosophies). Even better informed voters can still be very partisan.

A study [1] by Kahan et al found that people who demonstrated they could understand or solve a particular maths problem still engaged in motivated reasoning and failed to demonstrate that same mathematical ability when the problem was presented as a question about gun control. These findings undermine the view of democracy, that rational voters vote for their interests with an understanding and that pressure forces parties to cater to what we want. What we actually have is voters engaged in motivated reasoning who aren’t equipped with a lot of understanding.

Epistocracy is a way to fix these problems. A practical example is possible regarding Brexit, using data from Ipsos Mori. Ipsos Mori collected a huge amount of data on people, on questions that had predictive power on the Brexit vote: how much foreign investment comes from China?; how much from the EU?; what percentage of the UK population is made up of EU migrants? etc. What the data finds is that everyone is wildly uninformed about these things (but Brexit voters more so than Remainers), even though self-reported answers to these questions are predictors of voting intentions in Brexit. (Also interesting: everyone ― Remainers and Leavers ― overestimated the negative and underestimated the positive.)

Brennan’s claim is that if he could have waved a magic wand and made everyone better informed about these questions, that the vote would have been overwhelmingly for Remain.

Now, I have my doubts: a correlation between being informed and voting intention isn’t necessarily causative. It could easily be that being misinformed and voting for Brexit are both caused by motivated reasoning. For example, if you could make a Brexit voter believe EU immigrants make up less than 6% of the UK population (not 30%) they might still argue that’s too high, or that it’s increasing. Similarly, if you could convince them that Chinese investment into the UK is 1% of foreign investment (not 50%) they might still argue that it’s too high. If xenophobia is driving the misunderstanding and the voting intention, then correcting the misunderstanding wouldn’t adjust the vote. And, listening to LBC and the callers who just angrily yell things like “what right have they got to be here, anyway?”, you realise that 1 EU immigrant would be too many for some people.

You therefore need a quiz that can correct for motivated reasoning as well as being misinformed. And now you’re in really dangerous territory. Although we can argue that ignorance and irrationality disconnects us from the politicians and policies, so we don’t get outcomes that we want, and so it may well be legitimate to correct for that, xenophobia is different. Brexit is a bad example here because it is so encompassing; perhaps if people understood how their material life was going to be affected they would vote for their job security and opportunities and the economy in spite of their xenophobia. But, in smaller scale questions, can it really be right to ‘correct’ for values?

In fact, follow this through and it gets worse still. If being wrong about questions like EU migrant population size and foreign investment is a consequence of motivated reasoning based in xenophobia, and you correct for being right about those questions, then you are “correcting” out xenophobia of people who would have voted for Brexit even if they knew these things.

To recap where we’re at so far, before moving on:

  • Democracy is basically good, but has significant flaws around the understanding and attitudes voters have around important issues.
  • Epistocracy is a proposed solution to that. It would collect demographic data and administer a political quiz alongside your vote. That data can be used to calculate what the vote would have been in an idealised version of the electorate, where everyone was “perfectly” informed.
    • It is feasible that a statistical method to calculate an idealised electorate will be blind to values an individual has that is the cause of both the misunderstanding and the voting intention, such that the individual’s voting intention would not have changed with better knowledge.

That’s the inherent problem. Now let’s start to look at the practical problems: the level the political quiz should be at; the decisions around which demographic details are relevant; the ability to micro-target your supporters with courses to help them get a higher score in the quiz; a zealot bias.

I try to imagine what the quiz would look like. Brennan seems to be advocating moderately low level questions: who is your representative? Which of these is a law passed in the last 6 months? That sort of thing. But it is easy to imagine a quiz that includes a question like ‘what has some legitimate independent think tank forecast the economic impact of this policy to be?’ which is at a much higher level and is immediately open to abuse. The current UK Government acts as if openly partisan ‘think tanks’ are independent and legitimate (and Trump used to do the same). Also, the ‘economic impact’ might not be the defining issue relating to a policy; what about questions relating to how it might affect employment or immigration; what about essay style questions about soft power or political legitimacy on the world stage?

Even the question “who is your representative” assumes you accept the technical definition of the word “representative”. The Member of Parliament for my constituency can rarely be seen representing my views…

A political party with some statistical talent might notice that there are different ways of dividing up demographics. You could, for example, refuse to put ethnicity on the demographic questionnaire (decrying racism, probably) and that would mute the ethnic minority vote ― which would benefit the Tories, the ones currently in power. Similarly, Labour might refuse income information (because: classism) but that’ll diminish the vote of the wealthy.

Technology like Cambridge Analytica would allow politicians to micro-target their supporters and equip them with the information needed to pass. This will then create a phenomenon where it would appear the idealised population votes for the party that trained their voters. And then there’s the zealots: at a lower level, zealots tend to know more than the average people, even when they are deniers and cranks. If you’ve ever tried to argue with a creationist, the creationist will know more about evolution than a randomly picked person at the pub; the same seems to be true of Leave-voters, they can name a lot of politicians and laws and dates (even if they can’t demonstrate an understanding).

Yes, epistocracy does try to solve real problems. But, before we run into the embrace of it as a new system, it has a lot of problems it needs to solve.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s