When Democracy Meets Pluralism: Landemore's Epistemic Argument for Democracy and the Problem of Value Diversity

ABSTRACT

In Democratic Reason, Hélène Landemore makes an epistemic argument for democracy. She contends that, due to their greater cognitive diversity, democratic groups will engage in superior deliberation and information aggregation than will groups of experts; consequently, the quality of their policies will be better. But the introduction of value diversity into Landemore's model—which is necessary if the argument is to apply to the real world—undermines her argument for the epistemic superiority of democratic deliberation. First, the existence of value diversity threatens to stop deliberation prematurely. This has the effect of making the outcome of group deliberation more dependent on individual ability, which gives groups of experts a distinct advantage. Second, the introduction of value diversity raises the question of how to understand the standard of correctness of an epistemic argument, which Landemore does not adequately answer.

Notes

1. If there is more than one evaluative dimension then there will also be a weighting function between the evaluative dimensions, unless they are perfectly correlated.

2. Landemore, following Page, also includes interpretations and predictive models. We will not be concerned with these here because they are relevant to voting but not to deliberation, and my objections to Landemore's epistemic argument are about deliberation only.

3. Recall that landscapes are functions of perspectives and evaluative dimensions. Thus, a group of people with identical perspectives and heuristics can be cognitively diverse if they have different evaluative dimensions. For now, I assume that people have the same evaluative dimensions. I will revisit this point.

4. Measuring diversity is difficult. For an interesting discussion on this, see Page 2011, ch. 2.

5. Here it is illustrated just by diversity of perspectives. However, a similar illustration can be made with diversity of heuristics.

6. In order for diversity to trump ability, the group of non-experts must not hit the global optimum either; they just need to hit a higher optimum than the experts.

7. This is not to say that, for similar reasons adduced in the Theorem, the deliberative group will not outperform experts: It could be that all of the solutions proposed at the end of deliberation are better than the experts’ local optimum.

8. This does not rule out the possibility of a hybrid system in which experts deliberate and a diverse group of voters selects from among their options if no unique one is chosen. It is unclear whether this should be classified as a democracy.

9. The effectiveness of employing such a set of evaluative dimensions, should it exist at all, will be challenged below.

10. Deliberators plausibly can disagree about the value of proposals for reasons other than value diversity. There are at least two such reasons, both of which are important, but neither of which is my focus here. If I am right about this then the scope of my general argument would increase. The first is factual disagreement. Deliberators with the same values can disagree about the value of a policy because they disagree about the policy's effects. The important question here is whether the Diversity Trumps Ability Theorem applies in the face of factual disagreement. The short answer is that it depends. Note first that factual disagreement entails factual error: if two people agree about a matter of fact then at least one of them is wrong. With this in mind the Diversity Trumps Ability Theorem can apply in the face of factual error, but only when the error is not very large. An important factor is the extent to which deliberators can diminish their errors by learning from the process of deliberation. I am indebted to Scott Page and Paul Quirk for helpful discussions of these issues.

The second kind of disagreement is ontological disagreement. If factual disagreement is disagreement about how the world works, ontological disagreement is disagreement about what kinds of things are in the world. Two people who agree about all aspects of the physical world, and who have the same values, could still have policy disagreements because they disagree about what kinds of things the world is made of. Abortion is a possible example. People with the same values and who agree about all of the relevant physical facts—for example, that a fetus feels pain after so many weeks—can still disagree about whether abortion ought to be legally permissable because they disagree about what kind of thing a fetus is. They could disagree, for instance, about whether the fetus is the kind of thing that has rights.

In Landemore's account, ontological disagreement is importantly different from factual disagreement. This is because Landemore (214) takes her epistemic argument to be able to resolve questions of fact, which she understands as things that can be “verified and checked against reality” by our senses. Whether a fetus is a rights-bearer cannot be checked against reality by our senses. Thus, even if her argument can successfully address factual disagreement, it does not follow that it can successfully address ontological disagreement. I hope to be able to address both of these issues in the future.

11. It is not necessary that deliberation stop here. By introducing point C, Bev could induce Doug to see a further point P that both can (correctly) agree has a higher value. But this is not necessary either: It could also be the case that, after becoming aware of point C, point B is the only point that Doug believes is an improvement.

12. Note that this point is much higher than point A, which Bev and Doug started with. This reflects the fact that experts have higher local optima than non-experts, as we saw above.

13. Could deliberators keep on deliberating even if they disagree? Possibly. Landemore has suggested in conversation that deliberators could adopt a procedural solution to this problem. For instance, if there is disagreement then they could vote to provisionally select a particular proposal or set of evaluative dimensions, and then keep on deliberating. However, this technique poses two problems: path dependency and complexity. The path-dependency problem is that if they select the wrong proposal then they might end up never finding the global optimum. The complexity problem is that there are potentially many such disagreements; thus, in order to find the global optimum deliberators will have to keep track of all of the locations in which they choose a provisional path. Plausibly this would be unrealistically time- and resource-intensive. A fuller epistemic argument would have to take these factors into consideration.

14. This is not to say that they will always agree. They could have factual or ontological disagreements (see supra note 10). However, value homogeneity plausibly will increase agreement compared to the alternative.

15. Once again, this assumes factual agreement on proposals, an issue we are abstracting from here.

16. A limitation of my application of their study is that, whereas in their study the reference groups for the norms are the same, this is not the case in the situation I am describing. This could affect the applicability of the results.

17. This is not strictly true, for it could be that Proposal A is better than Proposal B on all evaluative dimensions. In this case, we could say that the correct proposal is one that is best (or weakly dominant) on all value sets, and there would be a sense in which the standard of correctness for proposals does not depend on which value set is correct. Perhaps, as Sen (2009) claims, these are the most pressing cases that justice should be concerned with. Nevertheless, it is also true that not all political problems are such that one proposal is the best on all evaluative dimensions.

18. She also considers other options. One is a truth-independent option, which I will not address because it is sufficiently similar to the political conception, and my response in the previous subsection applies to it as well. Another is an option brought forward by Philip Tetlock (2005). Tetlock's standard, though, only concerns facts, and does not address judging values, which is our concern here.

19. This is in line with Rawls's argument that “any sensible scheme of rules will not exceed the capacity of individuals to grasp and follow them with sufficient ease, nor will it burden citizens with requirements of knowledge and foresight that they cannot normally meet” (Rawls 2005, 267-68). The costs for a set of rules that does not meet these conditions, Rawls claims, would be unreasonably high. We might say something similar here: there might be prohibitively high costs in making an unconvincing argument work in practice, so it is highly desirable that the argument be convincing.

20. Perhaps this is a controversial criterion for arguments in political philosophy in general; some philosophers hold that justice is justice, regardless of what people think. This could be granted while preserving the minimal criterion for epistemic arguments. This is because epistemic arguments, or at least those like Landemore's, do not attempt to provide an account of a value like justice. They give instrumental reasons to choose one political system over another. Intuitively, whether an instrumental argument succeeds depends on its ability to realize its stated goal, in this case, the choice of democracy over the rule by experts. Whether people are convinced by the epistemic argument influences how successful the argument can be: If people are not convinced that Political System A is superior to Political System B then they have not been given a reason to choose Political System A, and so there is no reason to believe that they will choose it.

21. More strictly, they will be convinced only if the argument shows that the democracy will better reflect their value set than will a group of experts. However, all things being equal, there is no way to know what value sets the experts have. So it is fair to assume that the experts will have the same value sets as the deliberators. If this is the same value set as the addressees of the argument then democracy will be epistemically superior to rule by experts for the addressees of the argument because the democratic group will hit higher local optima on those evaluative dimensions. If it is a different value set then the addressees will either be indifferent or will prefer rule by experts, depending on how opposed the value sets are. These are issues that a plausible version of the epistemic argument will have to address.

22. It might be responded that if the epistemic argument puts forward an absolutist criterion, then part of the epistemic argument will involve an argument in favor of this criterion, and if part of the epistemic argument involves an argument in favor of this criterion then everyone, upon truly understanding the epistemic argument, will be convinced that the “correct” absolute standard is in fact correct. But this response seems to move the argument back into some kind of ideal theory while part of the point of introducing the assumption of value diversity is to make the epistemic argument more applicable to the real world. If the response to my objection is to resort to a claim that the epistemic argument applies only in highly idealized conditions, then I consider my argument to be largely successful. Moreover, the proposition that everyone would come to endorse a specific thick moral view even in ideal conditions is highly controversial—it is rejected by much liberal political philosophy, for example—so the burden would be on Landemore to provide this argument.