What Would Be an “Epistocratic Political Culture”?
Some Conjectures
I will participate next week in an interdisciplinary workshop at the University of Paris Dauphine on the use of meaningful, useful, and legitimate information in decision-making. On this occasion, I will present a paper dealing with two main objections addressed to the idea of epistocracy. I call them respectively the “Regressions Objection” and the “Selection Bias Objection.” Both target the core epistocratic conjecture that we can identify a set of “competent” voters that are more likely to vote correctly than less or non-competent voters. The former objection states that to identify competent voters, you need competence criteria which the establishment requires – from an epistocratic point of view – a form of legitimacy itself tied to the possession of particular competencies. This leads to an infinite regress, or to an arbitrary and somehow authoritarian decision about what counts as a competent voter. The latter objection states that even if you can identify competent voters, it is highly likely that competence is statistically correlated with non-epistemic attributes that can have adverse effects on voters’ ability to vote correctly. I want here to briefly discuss this last objection and hint at a possible answer. This forces us to make interesting speculations about what a society where political power is distributed according to knowledge and competence would look like.
The Selection Bias Objection is a generalization of the so-called demographic objection first put forward by the philosopher David Estlund.[1] It is sharply stated and defended in a recent paper by Sean Ingham and David Wiens.[2] The basic idea is pretty simple, but nonetheless potentially devastating for the idea that an epistocratic regime could be justified based on its epistemic superiority (i.e., its better ability to track the truth and to generate good collective decisions) to various forms of democratic regimes. To ground the epistemic superiority of epistocracy, you need to give reasons to believe that the following conjecture is true: a set of competent voters is, all things considered, is more likely to vote correctly (or to make better decisions) than a larger set of voters also including non- or less competent voters. Suppose that society has settled one way or another on competence criteria that are publicly acknowledged as justified – thus, what I’ve called the Regress Objection is solved. The Selection Bias Objection argues that, even in this case, there is no reason to think that the epistemic conjecture is true but of the likely existence of “cofounders” that, while statistically correlated to competence, have epistemically adverse effects. This can be illustrated by a figure used by Ingham and Wiens in their article:
As the caption below the figure explains, a cofounder is a factor that increases the probability of two attributes that have a contrary effect on the value of the relevant variable, in this case, the probability of voting correctly. Therefore, even if competence does indeed increase the probability of voting correctly, the fact that competence is correlated to a factor that in the meantime favors the presence of attributes that reduce the probability of voting correctly invalidates the epistocratic conjecture.
The Selection Bias Objection is really the conjunction of two distinct claims. First, the logical claim that the proposition “were Ann to be competent, she would be more likely to vote correctly than if she were not (or less) competent” is not equivalent to the proposition “the fact that Ann is competent makes it more likely that she votes correctly than a non-competent voter like Bob votes correctly.” The former is a proposition about a causal relation, the latter is a statistical claim about a correlation between two variables in a population. The former does not imply the latter, or as Ingham and Wiens put it, “causality does not imply correlation.” The second claim is empirical. It is that there is a vast array of potential cofounders that can make the second proposition false and that there are strong reasons that these cofounders would be at play if we implemented an epistocratic rule.
Ingham and Wiens identify two general classes of such cofounders. First, cofounders that affect persons’ opportunities to train and acquire competence, second, cofounders that affect persons’ motivation to acquire competence. They give examples of both types that are plausible enough to cast serious doubts on the truth value of the epistocratic conjecture. For instance, the opportunity to acquire competence (through higher education for example) is likely to be correlated to gender, race, or income, all factors that are also susceptible to favor attributes with adverse epistemic effects.
That sounds indeed reasonable. The problem however with these examples is that they are superficial and obviously merely illustrative. They do not give specific reason to think that they are at play, or that they should necessarily be at play in an epistocratic society. More importantly, even if they are right about the general causal structure corresponding to the figure above, it remains to be shown that the adverse epistemic effects of cofounders are stronger than their positive effects through the competence pathway. In other words, the strength of these examples mostly depends on one’s priors. To be fair to the authors, it’s difficult to do more. As we have yet to see an authentic epistocratic regime, we’re short of data to evaluate any statistical claim of that sort. Moreover, such evaluation would suppose that we have a good idea of what it is to vote “correctly,” which obviously we don’t have.[3]
Besides, I think that a more sophisticated objection can be addressed to Ingham and Wiens. This involves the somehow obscure logic of counterfactuals, but I will try to put it as simply as possible. Compare the two following propositions:
(1) “If I arrive late in class tomorrow morning, students will be waiting for me in the classroom.”
(2) “If I were to arrive late in class tomorrow morning, students would be waiting for me in the classroom.”
While they look similar, the grammatical structure of these two propositions is different. Proposition (1) is an indicative conditional while proposition (2) is a subjunctive conditional, also sometimes called a counterfactual.[4] To see the difference, note that (1) is obviously true while (2) may reasonably be false. Why? Because in the case of (2), if the antecedent were to be true, I would probably warn the student so that they come later and don’t wait for me. As David Lewis put it in his seminal logical treatment of counterfactuals, subjunctive conditionals are “variably strict.”[5] To understand what it means exactly, we can follow Lewis and use the semantics of possible worlds. Say A is our actual world. In this world, I always arrive in class (almost) on time. Suppose now that I want to assess the truth value of proposition (2). For that, I need to imagine a possible world P where I’m arriving late in class. The thing is that this is not what has (or will) happened. The fact that I arrived late without warning my students could not happen in A. If it were to happen, that would mean that something on top of the fact of being late has changed compared to A. In other words, P is not “A except that I arrive late.” What we know about the causal structure of the world implies that P is “more different” than that. So, when assessing the truth of (2), what you need to look at is the closest possible world where the antecedent is true. If in this possible world, the consequent is true, then the counterfactual is true, if not, the counterfactual is false. This is why we say that subjunctive conditionals are variably strict, their truth value depends on judgment of similarity between possible worlds. The figure below illustrates the idea:
In this figure, A is the actual world, and P is the possible world under consideration. The spheres contain possible worlds ordered in terms of similarity with respect to A. The worlds within the second sphere are the most similar to A, and the more out the sphere a world belongs to the more dissimilar it is. If x is the antecedent (“If I were to arrive late”) and y is the consequent (“the student would be waiting for me”), then we see that P is the closest possible world where I arrive late. In this possible world, it happens that y is false, thus the above counterfactual is false. It’s not that x and y cannot be true at the same time. There is indeed a possible world belonging to the outermost sphere where both are true, but this is irrelevant to assessing the counterfactual.
What’s the point of this digression? When Ingham and Wiens present their examples, they implicitly assume that they should be evaluated in our actual world. In other words, they are interpreting the epistocratic conjecture as an indicative conditional. But this cannot be correct. We are not living under an epistocratic regime and what should be under evaluation is a counterfactual. The point is therefore that epistocracy cannot merely consist in a change of voting rules. If we were to implement epistocratic rules, our understanding of the world suggests that other elements in the society would have to be changed as well. In other words, it is incorrect to reason everything else equal. In the same way that democracy is more than a set of voting rules but also consists in what John Rawls calls a “public political culture,” an epistocratic regime would also imply forms of life that differ from those that flourish in a democratic society. Epistocrats should indeed grant that implementing epistocracy in our actual world is unlikely to work as expected and would be likely to be just rejected.
The point is therefore that epistocracy is more than a political regime, it is a (set of) social form(s) where citizens value intellectual prowess and agree on the legitimacy conferred by the possession of a certain kind of knowledge. For this to be possible, it probably requires that individuals have a high degree of what I would call “epistemic autonomy.” This cannot happen in a too inegalitarian society where education levels are highly heterogeneous and, more importantly, where opportunities are not fairly distributed. In other words, an epistocratic society would be – counterintuitively maybe – a relatively equal society where the state actively supports people’s education and training and promotes a perfectionist form of life putting epistemic values and autonomy at the forefront. In this world, we may imagine that the kind of cofounders that Ingham and Wiens identify would be far less worrying than in our actual world.
Of course, it could be objected that this “epistocratic ideal” is unlikely to happen and maybe is impossible. That may be true. As I explain however in the paper I’ve referred to at the beginning, I think that the mechanisms promoted by proponents of epistemic democracy (e.g., open-mini jury) are likely to favor similar social forms. A point of convergence between epistemic democracy and epistocracy is that they would make the exercise of one’s political freedom more epistemically demanding. Epistemic democrats and epistocrats agree that political freedom means more than just expressing one’s opinion. It requires that citizens are willing to allocate time and cognitive resources to improve their political judgments. Once this is something widely acknowledged and accepted by the population, we will have made a major move in the direction of the epistocratic ideal.
[1] David M. Estlund, Democratic Authority – A Philosophical Framework (Princeton, NJ: Princeton University Press, 2009).
[2] Sean Ingham and David Wiens, “Demographic Objections to Epistocracy: A Generalization,” Philosophy and Public Affairs 49, no. 4 (2021): 323–49.
[3] Admittedly, this is a problem as much for opponents as for proponents of epistocracy. After all, the fact that we are not able to agree on what it is to vote correctly is an argument against epistocracy – this is basically a version of the Regression Objection.
[4] Strictly speaking, this is incorrect, because a counterfactual is a subjunctive conditional where the antecedent is a past event that will never happen. But this is irrelevant here.
[5] David Lewis, Counterfactuals (Malden, Mass: Wiley-Blackwell, 2000).