A Note on the Epistemology of Disagreement and Politics

Epistemic peerhood in the electorate

Theories of disagreement tell us how we ought to modify our beliefs, if at all, in the face of disagreement from our epistemic peers—people who are (1) roughly as knowledgeable as we are about the matter at dispute, and (2) roughly our intellectual equals. ⁴ If this epistemic peerhood relation does not hold, then conciliationism (or any other theory of disagreement) doesn’t apply.

At first blush, it does not appear that any claim about large-scale epistemic parity within the electorate could be correct. 38 percent of the American electorate knows which party controls the House of Representatives; 44 percent does not know; and 17 percent erroneously believes that the minority party does. Only 36 percent of the electorate can name all three branches of government; 29 percent can name only one or two of them, and 35 percent can name none. ⁵ The literature on the state of political knowledge is extensive and depressing, and it shows—the important thing for our purposes—that the electorate is hetereogenous in this respect. As Scott Althaus summarizes things, roughly a quarter of the electorate is well informed, a quarter is systematically misinformed, and the middle 50 percent suffers serious defects in political knowledge. ⁶

In addition, voters differ wildly in their ability to think rationally, thus violating the second necessary condition for epistemic parity. For example, some voters suffer from serious cognitive biases, such as racism, and others do not. ⁷ Further complicating matters is that sometimes the most knowledgeable voters are the least rational ones. ⁸ In short, when it comes to epistemic peerhood within the electorate, the situation is a mess.

To his credit, Ebeling does not make the implausible claim that voters are generally epistemic peers. He claims instead that they “ought to regard the judgments informing their votes as equally reliable given the relevant threshold of comparison defined by their epistemic role in the decision procedures of modern democracies.” ⁹ Ebeling’s idea is that voters benefit from the work of political parties, which act as “epistemic enrichment facilities,” “significantly eas[ing] . . . cognitive burdens” on voters and ultimately producing coherent and specific ideological platforms. ¹⁰

I am skeptical that parties do such a thing, but let us grant that it is true. It simply does not follow that an individual voter is justified in regarding her political judgments as equally reliable as those of her fellow party members. What follows, instead, is that she is justified in regarding her party’s ideology as more likely to be correct than her own (however we end up defining “correct.”)

As a practical matter, this might mean that an individual voter is rationally justified in voting for her party’s nominee rather than an independent candidate who better reflects her personal preferences. But rationality does not demand that she be indifferent between her candidate and some other independent candidate who is preferred by one or more of her fellow voters. That indifference would follow, however, if she truly regarded her fellow party members’ judgments as equally reliable.

Consider Eddie, an economist, who voted Republican because he believes that free markets best promote the general welfare. Fred, a farmer, voted Republican because he wants the White House free of traitorous Muslims (43 percent of Republicans believe that President Obama is a Muslim.) ¹¹ It may well be that the Republican Party platform is more coherent and specific than Eddie’s is. But even if that is true, and even if Eddie and Fred had equal roles in the process that produced that party platform, Eddie should not regard Fred’s political judgments as equally reliable. Indeed, if Eddie is familiar with the way that Fred forms his judgments, he is rationally required to regard Fred as his epistemic inferior. A rational Republican today is obliged to regard the votes cast by many of his fellow party members as results of an unreliable cognitive process. They are not his peers.

Now, Ebeling would be correct if parties (1) aggregated individual preferences in an epistemically valuable way and (2) voters simply adopted, wholesale, the resulting ideological platform. If that were the case, then each voter would be justified as regarding his fellow party members as peers, for each member’s ideology would be formed in an equally reliable (indeed, identical) way—it would be delivered to him by an epistemic superior (viz., the party). But that does not actually happen. And if it did it would render the whole issue of disagreement moot, for there would be no intraparty disagreement at all.

Note, also, that some epistemic enrichment processes are better than others. No matter how coherent and specific the Republican Party’s platform is right now, it would be better if it were influenced by more sober economists and fewer irrational racists. (A similar claim can be made for the Democrats, of course.) If Ebeling is right and the party process has the epistemic value that he believes that it does, the challenge is then to make that process as valuable as possible. Those of us inclined to meritocratic political systems believe that the way to do so is to vest more knowledgeable and rational voters with more political power. Rather than strike a new blow for egalitarianism, Ebeling’s argument, if it is sound, only changes the terms of the debate; instead of asking which processes produce optimal political outcomes, we ask which processes produce optimal party platforms. The underlying dispute between egalitarianism and meritocracy remains.

Two technical worries

Conciliationism, as its theory currently stands, does not support Ebeling’s conclusions. Moreover, generalizing the theory to handle disagreement among multiple people—a necessary step if we wish to apply it to politics—is not straightforward.

Ebeling says that “we can . . . conceive the median voter as the voter who endorses a judgment that expresses, or is close to, the conciliatory position between disagreeing peers.” ¹² In general, there will not be a congruence. Consider Ebeling’s example:

Tax Rates. Five legislators have to determine the rate of the income tax. We assume that these legislators regard each other as equal epistemic authorities with respect to the justice of income tax rates. They have argued in good faith about the issues for a considerable amount of time but have failed to reach an agreement. After deliberating sincerely about the issue for a considerable amount of time, Legislator 1 prefers a tax rate of 40%, Legislator 2 of 50%, Legislator 3 of 60%, Legislator 4 proposes a tax rate of 70%, and Legislator 5 of 80%. ¹³

Ebeling thinks that the “correct”—that is, rational—post-conciliation belief is that the optimal tax rate is 60 percent. In general, conciliationism will not produce this result.

Theories of disagreement apply to epistemic peers who dispute the truth of an individual proposition. If the five legislators were simply disagreeing about the proposition we should implement a 60 percent tax, and three legislators said “yes” and two said “no,” then the conciliatory belief would accord with the median judgment.

But in Tax Rates, there are at least five propositions in play (viz., the tax rate should be 40 percent, the tax rate should be 50 percent, the tax rate should be 60 percent, the tax rate should be 70 percent, and the tax rate should be 80 percent)—and, therefore, a total of 25 relevant beliefs. We are only given 5 of these, so the scenario is underdescribed. In any plausible real-world example, we will also expect our legislators to hold opinions about other potential tax rates—that is, about propositions like the tax rate should be 75 percent. In addition, a belief in a proposition p is more accurately modeled by a continuum of confidences running from 0 to 1 (inclusive), where 0 represents perfect certainty that p is false and 1 perfect certainty that p is true.

With those considerations in mind, we can see why the results of conciliating and the median judgment need not agree. For example: Suppose that the economy for which these legislators are contemplating tax policy is arranged such that a 60 percent tax would badly distort incentives but that higher and lower rates would not. Legislators 1, 2, 4, and 5 understand this; in addition to their opinions on the ideal rate (40, 50, 70, and 80 percent, respectively), they also believe that it is vitally important that a 60 percent rate not be chosen. In this case, the median judgment is 60 percent, but the conciliatory position lies somewhere else (where exactly will depend on the specifics of the scenario).

The second technical worry is that problems arise when conciliationism is applied to groups of people larger than two. ¹⁴ Ebeling assumes that a rational legislator will believe that 60 percent is the ideal rate, but it is unclear how conciliationism might arrive at that result.

Given current theory, an individual legislator must conciliate with each of his peers in turn. And the order he conciliates in matters. ¹⁵ If Legislator 1 conciliates (2→3→4→5) he will arrive at a different tax rate than if he had conciliated (5→2→3→4). For example, Legislator 3—who held the “correct” belief about the tax rate, 60 percent, before conciliating—may end up endorsing a rate as low as 50 percent or as high as 70 percent, depending on the order in which he conciliates with his peers. And no matter what order they choose, Legislators 2 and 4 cannot arrive at the desired 60 percent.

This is puzzling because it is not clear why the order in which a legislator consults his peers should make any difference to his final beliefs. Yet the order does make a difference.

A natural response here is to say that Legislator 1 ought to conciliate with some abstract entity that represents the aggregation of the other legislators’ preferences. Two problems immediately arise. First, Legislator 1 cannot simply average the other peers’ beliefs and conciliate with that: He (and Legislators 2, 4, and 5, were they to follow the same norm) would arrive at the wrong result (i.e., not 60 percent). Moreover, they would all disagree with each other after conciliating—odd since, ex hypothesi, they are epistemic peers. Second, no matter how the other legislators’ preferences are aggregated, if that aggregation has epistemic value, as we have supposed that it does, then Legislator 1 is no longer disagreeing with an epistemic peer—he is disagreeing with an epstemic superior. And so conciliationism doesn’t apply.

I suspect that some of these issues would be resolved by a more sophisticated, multi-peer theory of conciliationism. Nevertheless, given the many well-known problems related to the aggregation of preferences, and the formal results that support meritocratic rather than democratic decision procedures, we should not be too hasty in thinking that conciliationism and egalitarian politics go hand-in-hand. ¹⁶