Response to Comments on Chapter 4

As usual, I’d like to offer some comments in response to points and questions raised in this week’s comments following Ben’s terrific summary of my Chapter 4. Sorry again for the length, but I’ve tried to briefly address most of the points that have come up.

First on Ben’s comments: Ben grants that a random choice procedure is procedurally fair, but denies that it is democratic. I agree that obviously it isn’t democratic. But that is not yet any argument against it, because it may be that the decision procedure needs only to be procedurally fair. So any attempt to justify democracy solely on the ground that it is procedurally fair seems forced to say that a random (and non-democratic) procedure is equally justified, being equally fair. Ben argues that democracy includes some element of what I call aggregativity. I agree. He says that the theorist who wants to ground democracy in fairness might just mean that what’s required is a procedurally fair treatment of everyone’s votes or preferences. I agree. My argument is that this shows that they appeal to something more than mere procedural fairness. They add a principle requiring a link between the output and individual inputs, a principle that is not part of the concept of procedural fairness (as shown by the example of a random choice procedure). So I think there’s nothing in that first comment of Ben’s that we disagree about, but I hope this clarifies what I’m arguing for.

Ben adds a brief remark about how some theorists defend democracy on educative grounds. I take it that even these theorists require at least some aggregativity in the procedure, some even going in a more epistemic direction, such as Mill. So they pose no challenge to my arguments in this chapter. Paul also mentions the educative approach in comment 4, so I might as well say just a little bit about it here. (I can’t recall whether anything along these lines made it into the book.) I wouldn’t deny that democracy can have some beneficial effects on the education and development of individuals participants, and this must count in its favor to some extent. But I agree with Elster (in Sour Grapes) that the educative benefits seem bound to piggy-back on some self-understanding that involves some other account of the value of democracy, so it merely postpones the questions about proceduralism, substance, etc. It is difficult to believe that democratic voters would build the lively curiosity and public spirit emphasized by the educative approach if they thought the only value of democracy was this cultivation itself. If, on the other hand, a theory tried to avoid this move and stick with the educative effects themselves, I would suggest comparing this with a merely educative justification for, say, an institution of faith-healing: “No, there’s no reason to think it does any good for the patient, but the healers become the most energetic and interesting people!”

Ben’s second comment is mainly a question about the development of my thinking since my article, “Beyond Fairness and Deliberation.” In that article I granted that political procedures must be procedurally fair, but argued that they must also include an epistemic dimension. I no longer think that there is any role for a principle requiring procedural fairness, and I make to appeal to that value in my argument. I grant that democratic procedures often would count as procedurally fair in certain respects, but I take this to be of little importance. A random choice of outcomes would also be procedurally fair, and yet none of us think that would give much legitimacy to political decisions. So it seems equally clear that it adds little legitimacy when it is present along with other legitimacy-supporting considerations such as the ones I appeal to in epistemic proceduralism. I would even go slightly further, rejecting any principle requiring procedural fairness as such. The reason is that I believe that if there is a procedure which is the epistemically best available within public reason (beyond qualified disagreement), but was not fair, that would be no objection to it at all. It would not be unfair, a pro tanto disvalue, just non-fair, which is nothing against it. And, in fact, I think it is not at all clear that a comparison between the procedural power I have over law and policy, and the procedural power a senator has, can be shown to be equal, which procedural fairness would seem to require. I conclude that procedural fairness is probably not required. (I briefly mention this point at 221-22.)

In comment 2, I offer a response to Bill’s comments in 1, so I won’t revisit that here.

Jonathan (in comment 3) argues as follows:

I think there are two distinct questions we can ask about a decision procedure. One is whether there is an independently correct answer to the question being decided by the procedure. If we use a decision procedure to answer the question – should abortion be legal? – one issue is whether there is an independently correct answer about the legality of abortion. Let’s call this the substantive question. The second question is, by what standards do we evaluate the decision procedure? For example, do we evaluate the decision procedure on its capacity to promote artistic beauty in the world? Let’s call this the procedural question.

This is a fine distinction as far as it goes. The problem with it for my purposes is that it suppresses the question I want to address, which is whether there are other ways to import non-procedural values than by attending to the correct answer to some question. I am not arguing that majority rule smuggles in epistemic concerns, but only that it imports procedure-independent standards of good outcomes. As I argue, “…appeals to fairness strike us as so attractive precisely because they purport to be fully neutral, not carrying any conception of good ends other than the procedure itself. But only full anonymity, which often has little moral value, actually lives up to that.” (81) The term “procedural” isn’t so important to me, but a procedure that insists on sensitivity to the participants’ choices in addition to treating all participants exactly symmetrically (as a random choice procedure would) is steering the outcome in a way the random procedure is not. Once it is granted that any plausible account of political procedures imports values on the basis of which to steer the outcome, then none of them can claim to be completely “hands-off” with respect to steering the outcomes in accordance with certain values. Once that is cleared up, we can argue about which sorts of values are available for this purpose. Having said this, I readily grant that I want to defend appealing to more substance than fairness defenses of majority rule do. That’s because I not only want the outcome steered to reflect the differences between voters (which is already a value that goes beyond procedural fairness), but also (or in order), in a certain way, to steer it toward an independently correct answer to the question put to the voters. Whether my approach is justified depends on the outcome of a debate between this use of substance and fair-proceduralism’s use of a different substantive standard. That’s not a debate between appealing to substantive standards and not appealing to substantive standards.

Paul, in comment 4, argues against my claim that there is a kind of primacy of substance over procedure, so that appeals to procedural values are only warranted by the need (for various reasons) to retreats from substance. He considers the example of a jury trial of a person who is known by everyone to be guilty and to deserve jail, with all this being common knowledge (if I understand Paul’s example correctly). The crucial thing, for Paul’s purposes, is to stipulate that a jury trial, in this case, has no advantages with respect to getting the substantively correct decision. If it did, I would explain it as a very partial retreat: from shared certainty about the right answer, to agreement about a procedure likely to get the right answer. The retreat could stop here, and not go all the way to merely agreement on a procedure that is fair between participants.

Here we need to pause and clarify some different uses of the term “fair procedure.” We often talk about a fair trial, but we are not referring to the kind of procedural fairness of the view I call “fair proceduralism.” So the language is treacherous here. The fair proceduralist view eschews any appeal to a tendency to get a right answer, and substitutes only the appeal to giving each person an equal chance of influencing the procedure in the direction they prefer. This latter approach would clearly be crazy in a criminal justice context. (I believe I make this point in the book, but I can’t find it .) Imagine the view that all the lawyers, the judge, the defendant, and the jury votes, and saying the justification for the outcome is simply that it satisfies a majority, with no reference to any tendency to acquit the innocent or convict the guilty. So that’s not what we mean by a fair trial. This latter thing really has almost nothing to do with procedural fairness in the sense we’re discussing.

This linguistic point isn’t fatal to Paul’s point. He argues that in the common knowledge case we should still have a jury trial, and for purely procedural reasons. This wouldn’t be a problem for me if I could show that the resort to procedural values depended on the unavailability of routes to the substantively best outcome. So Paul points out that the substantively correct answer is as available as can be. If I am understanding the challenge correctly, its weakness seems to be the claim that there remains some significant value in having the jury trial. (Let’s put aside the higher order reasons there be for having an institution that handles such cases by a jury trial without looking into whether the common knowledge of guilt drains such a procedure of any value. That might be warranted on pragmatic grounds that seem irrelevant to our main question.) In the common knowledge case, I can’t see any significant value in having a jury trial. For what it’s worth, actual trial systems will often acknowledge this and forgo the trial by accepting a guilty plea. Sometimes guilty pleas are not accepted, but never, so far as I know, in cases where the defendant is competent and all agree she is guilty.

So, on my analysis, the point of a jury trial is (mainly) that it is a publicly acceptable method for coming to the correct judgment in the absence of agreement about what that correct judgment is. That exhibits what I call the dynamic of retreat. (Of course, there are also other values in the picture, such as favoring the defendant, deterring illegal searches, and so on. I believe my approach can accommodate these, but I won’t pursue that here.)

In comment 5, Paul writes,

On pg. 82, it’s claimed that a coin flip is purely procedural. But on pg. 75-6, it’s claimed that a coin-flip procedure that actually insists that the outcome be a result of the coin flip (rather than merely correlated) is not purely procedural, because it too demands that the coin flip cause the outcome.

My point in that latter passage is precisely to distinguish what is sometimes called “responsiveness” in social choice theory contexts, from the very different idea of a retrospective procedural fairness. Social choice functions are not really about responsiveness in a procedural sense at all, any more than 4 is responsive to the procedure of 2+2. Much confusion has stemmed from conflating abstract aggregation functions with actual temporal procedures for casting and counting votes. I fear Paul’s subsequent discussion stems from a confusion about what I was saying, but if not I’d welcome clarification of the point.

Andrew writes,

… the argument that begins on p.78 aims to establish that procedural fairness consists in full anonymity, which requires that the procedure’s outcome be insensitive to anything about individuals, including the fact that (e.g.) we all prefer X to Y. On this conception of procedural fairness, equal insensitivity and equal sensitivity are not equally procedurally fair; only insensitivity is fair.

As he says, I anticipate this reaction at page 81: “We should not say that procedures, including majority rule, are not fair simply because they are not fully anonymous. Rather, procedures can mix fairness and other principles, remaining fair, among other things.” (81) And yet, he asks, if fairness is full anonymity, aren’t departures from anonymity departures from fairness? In a way, yes. But it might risk confusion to insist on putting it that way, because it’s so easy for all of us to slip into the unwarranted inference from “less fair” to “at least somewhat unfair.” That’s a fallacy, as I argue in my discussion of the “non/un issue.” So I prefer to say that only fully anonymous procedures are free of non-fairness elements. Other procedures needn’t be called less fair, but they are not only fair. So their extra elements, such as aggregativity in many cases, cannot be justified by appeal to the idea of procedural fairness. So I don’t say that majority rule is less fair than a random procedure, just that it includes extra elements as well that can’t be explained by fairness.

I think Andrew’s final paragraph in comment 6 might be a fine way of putting my argument while avoiding linguistic disputes about the term “fairness.” He uses the term, “equality in the procedure” instead. So I’m happy with that, but I still offer my analysis of procedural fairness as having much to be said for it. Remember, as I say at p. 80, if aggregativity were built into the idea of procedural fairness, then it would be a mistake to ever say that a random procedure is a fair procedure. That strikes me as conclusively showing that fairness does not incorporate aggregativity. Then, since a procedure can be fair without aggregativity, the case for aggregativity cannot rest on fairness.

Finally (or up to this point in the comment thread so far), Andrew takes up my argument that there is often a moral reason to do what people say rather than what they want, which would be an objection to views (like Pettit’s, perhaps) that take the latter to be the primary democratic aim. Andrew challenges my claim with the example of strategic voting. It’s certainly true that if what we want is to track preferences we are hampered by having to use votes as surrogates for preferences. But Andrew offers no reason to think that we should, if we could, track preferences rather than votes. Even outside of voting contexts we all often “vote strategically,” or direct people to act in ways that are not actually the way we want them to act. “Please, Professor Rawls, don’t spend much time on this questions I’ve sent you,” or “Come to my birthday party, but don’t bring any gifts” are not always accurate information about the speaker’s preferences. It doesn’t follow that, morally, the thing to do is to guess their preferences and do that instead. I give other examples in the book. The general idea is that in many contexts—and I don’t yet see why strategic voting wouldn’t be one of them (though I’m not committed either way)—we ought to do what the person says even when we know it does not reflect what they want. This might be so even where their misrepresenting themselves in this way is morally dubious, in case Andrew thinks strategic voting is morally dubious. For an example of this last thing, it might be wrong for me to tell you to stop helping with the dishes when I really prefer the help. But if it’s my house, you should probably stop helping, at least if I don’t give in. This doesn’t prove that we should follow votes rather than preferences even in cases of strategic voting, but nor is that a clear example where we shouldn’t.

Be Sociable, Share!
This entry was posted in Posts, Reading Group. Bookmark the permalink.

6 Responses to Response to Comments on Chapter 4

  1. Ben Saunders says:

    Thanks again David for your very detailed reply. It’s certainly helped clarify your position. I would like, however, to press my comments a little further. (I would add that you have no obligation to respond further but, after the Rawls remark in your last paragraph, I’m not sure whether you’ll do as I say!)

    You agree with most of my first point, which is as I expected – it was supposed to be something you could accept, so far as coin-tossing and majority-rule are both fair. That seems to suffice for your purpose, since it establishes that fairness can’t chose between the two. The challenge comes with the thought that there might be some other way of choosing between the two, when equally fair, that doesn’t appeal to substantive outcomes.

    You point out the educative effect is unlikely if people only value majority-rule for that reason, and I agree, but I think there may be a certain delusion going on – people vote seeking correct answers but the real reason to value the procedure is not that it’s any more likely to reach such but it develops capcities for broader consideration and deliberation. (You could perhaps compare this to something like the paradox of hedonism).

    Moreover, this is only one possibility. One could justify democracy on the grounds that one thinks there’s some inherent value in stimulating public reason and debate (I think Richard Vernon argues something like this) or simply say the value of freedom applies to groups, so we should let groups be self-determining (in a way realised by majority-rule but not a coin-flip) for reasons other than that they might thereby produce better outcomes.

    As to my second point, I think it’s clear that fairness has diminished in importance in your account and that this is (partly?) because you think things can be non-fair without being unfair. Really my worry was that I’m not sure the substantive issue is lexically prior to procedural fairness. I think substantive values may be indeterminate between different equally just outcomes, so there’s still some room for fairness in deciding between these. One might also push the issue by asking whether there’s any reason to think of this as a case of non-fairness rather than unfairness.

    On the whole, though, I’m very sympathetic to your claims here. I think democracy requires more than fairness, because it also requires citizen input. I wouldn’t say a procedure needs epistemic value to be democratic, but I’m open to saying it needs such to be justified – though I’m presently more inclined to stressing something like educative effects. Perhaps the more fundamental difference is that I’m less inclined to regard modern representative institutions as particularly democratic.

  2. Paul Gowder says:

    Thanks David.

    With respect to comment four, what would you say to the usual panoply of deontological (all the autonomy-based stuff, consent to be governed, duty to participate, etc. etc.) and non-outcome-oriented consequentialist (like Mill’s school for public spirit) reasons for incorporating public participation into governance in cases like the jury trial? I assume you’re not just rejecting them out of hand.

    My first pass at supplying an answer for you to that question would be to make use again of the distinction between substance and procedure, and to say that an appeal to those reasons isn’t purely procedural. But (see below), I guess I still don’t quite understand what “purely procedural” means. For those values are such that they’re present regardless of what the ultimate outcome of the decision procedure is. We can believe the jury will be right or wrong, and still believe that it ought to be making the decision because we care about educating the public to be better citizens. Isn’t that a perfectly procedural, and perfectly good, reason to prefer the jury even when it’s epistemically dominated by a non-jury procedure?

    With respect to comment five: I may have been confused (indeed, it seems likely, since your clarification bears little relationship to how I understood that section of the chapter). May I request a further clarification? Specifically: I take it that you think that there is a distinction between a procedurally fair standard and a substantively fair standard, and that distinction has something to do with the function that maps acts of decision-making (voting, coin-flipping, the utterance of a monarch’s command etc.) to outcomes (the enactment of a law). There must be such a distinction running underneath the chapter in order for the argument that procedural fairness does no work to fly. Can you say what that distinction is?

  3. David, You want this chapter to show three things:

    1) Majority rule is preferable to decision by lottery;

    2) Majority rule is preferable to decision by lottery on grounds other than fairness;

    3) What makes majority rule preferable to decision by lottery is its epistemic superiority.

    Let’s grant 1) and 2), and in granting 2) grant that aggregativity–what Kenneth May called “positive responsiveness”–is not an element of fairness. May showed in 1952 that in a choice between two alternatives, majority rule uniquely satisfies four attractive conditions (the other three, besides positive responsiveness, he called “decisiveness,” “anonynymity, “and “neutrality”–but I digress). He defined positive responsiveness this way: If a decision is favorable to alternative A, or indifferent between A and B, then if a voter switches to A, the result is favorable to A. Though it satisfies the other three conditions, flipping a coin doesn’t satisfy positive responsiveness. But majority rule satisfies all four.

    So, because May has offered a compelling explanation of 1), the burden of your argument is to show that positive responsiveness–concededly, by 2), not an aspect of fairness–is (or reflects) an epistemic consideration, and not merely that it is a non-fairness value, or merely that it is a “procedure- independent” or a “substantive” value in some sense, or that is not part of what you call “full anonymity.”

    Where is that argument? Or must others shoulder the burden of showing that positive responsiveness is nonepistemic?

  4. Ben Saunders says:


    Firstly, I think it’s wrong to say that May shows majority rule preferable to a lottery. As I recall, his article is entirely definitional, and he does little if anything to establish the attractiveness of his conditions. Certainly at least one is violated by any decision rule that isn’t simple majority rule (e.g. a super majority for constitutional amendment), but that doesn’t mean these alternatives are necessarily normatively unattractive.

    Secondly, I don’t think you’re right to draw the conclusion David has to show positive responsiveness to be epistemic. Positive responsiveness isn’t the only difference between a lottery and majority rule. This, of course, should be obvious as adding positive responsiveness to the other three conditions leaves you with only simple majority rule, whereas there are of course other decision rules that differ from lotteries (and perhaps satisfy weaker conditions, e.g. non-negative responsiveness).

    Finally, if we have any confidence that choice can perform better than simply flipping a coin between alternatives, then I guess the epistemic value of democracy is something to do with its responsiveness (though not necessarily its satisfying May’s positive responsiveness, which it probably doesn’t). I’d guess the random choice between options would be pretty unworkable, because it may end up selecting options no one wants and it’d be pretty hard to define the relevant options to be given equal chances anyway.

  5. Ben, Positive responsiveness is somewhat attractive, anyway, and it does something to explain why majority rule seems preferable to flipping a coin. Maybe there are more attractive conditions and procedures, but the task David set himself in chapter 4 was to show that “epistemicity” is what makes majority rule preferable to randomizing. There may indeed be something epistemic lurking within positive responsiveness: I don’t say there can’t be–but I don’t see that there is, unless the procedure-independent standard of correctness is somehow keyed to positive responsiveness. And I doubt that David would want to go in that direction.

  6. Thanks, Bill. Long answer:

    Bill asks whether I have an argument that majority rule is preferable to random choice on epistemic grounds. That’s a good question. However, before taking it up, let me point out that in Chapter 4 I am not trying to show any of the three things Bill lists. So let me review what I am trying to do.

    I try to refute the suggestion that majority rule is required by procedural fairness. The main idea is that a random procedure would be a fair procedure, and equally so, but fairness can’t choose between two equally fair procedures. If that were all I said, I could be pressed to justify my claim that a random procedure is equally fair. That requires some conception of procedural fairness, and so I offer one. Having defended that conception on independent grounds, I point out that it vindicates my claim that a random procedure is perfectly fair, and so certainly not less fair than majority rule. Majority rule is also fair in a way, but it adds a non-fairness element: aggregativity. That is a procedure-independent standard for outcomes, and not one I raise any objection to. But once we’ve gone beyond procedural fairness in that way, the door is open to other procedure-independent standards. The question is which ones are and which are not available. I point out that elsewhere in the book I argue that another available standard is a formal epistemic one: the claim that certain democratic arrangements will tend to produce substantively just decisions (whatever those might be).

    So in this chapter I have not argued that majority rule is superior to a random choice. I’ve just argued that fairness is indifferent between them.

    I haven’t been able to look back at May’s article, so for now I’ll go on the basis of Bill’s gloss. As he states “positive responsiveness,” I would indeed argue that it is not any part of procedural fairness. By the way, just to be clear, I hope we agree that it is not the same thing as my “aggregativity.” But they both violate anonymity, and so go beyond fairness. And he’s right, of course, that my argument for majority rule would rest on a claim that it is part of an arrangement that has epistemic value (all within public reason). But (and this is nothing to brag about) I don’t argue for that anywhere in the book. Still, nothing in Chapter 4 depends on such an argument.

    But let me say a few things about it. I do give some argument that public deliberation mixed with a dose of countervailing deviations from ideal deliberation will tend to do better than random. So far that doesn’t mention voting or majority rule. But we would need some way to take the measure of whatever epistemic improvements have taken place in the course of deliberation and other political activity. Majority rule is not by any means the only possibility. In fact, democracy is not the only possibility. We might rely on someone who is especially wise and able to learn from the points that are raised in public discussion and then let just this person (or elite) vote. My arguments against epistocracy apply here directly, so let’s count that out.

    This leaves decision procedures that are not premised on the supposed superiority of any specified person or class as compared with another. We could pick a decision randomly, which would be perfectly fair procedurally, but epistemically completely worthless. The epistemic value of political deliberation and activity would not be tapped. Here are some of the main remaining alternatives: majority rule (simple or special), voting by a randomly chosen subset of eligible voters, or decision by a single randomly chosen person (with a long or short term). All of these, by the way, are more or less equally procedurally fair, and all go beyond fairness by including some aggregativity in addition (and so none is purely fair in the sense I define as “full anonymity).

    Any of these would tend to track whatever epistemic progress public deliberation and political activity achieve. Which of them would do best I’m not sure. I often speak of majority rule, but that’s for simplicity, and I wouldn’t, without better warrant, want to be committed to the claim that it is preferable to a random voting sample or a single randomly selected person.

    Would the random sample method be democratic? I discuss this in my remarks on the comments on Chapter 9. Briefly, (with a few new points): It depends on how democracy is defined. I am using the definition of democracy as the actual collective authorization of laws and policies by those subject to them. Does the sampling method count? I’m not sure. I’m not sure it doesn’t. It’s not dictatorial, aristocratic, not oligarchic, or epistocratic. It might seem to fall short of the idea of democracy because not everyone votes, but that’s true under majority rule too, since turnout in large elections is far from complete. We just get a self-selected sample. Indeed, random sampling (with voting, not mere polling) might improve statistical representation if those selected were more likely to actually vote (as I predict they would be). I’m not recommending it. I’m just not sure.

    In any case, if you decide sampling shouldn’t be counted as democratic, but it is the epistemically best arrangement so far as can be explained within public reason, then I’m still comfortable saying it is what political legitimacy and authority require—democratic (technically) or not.

    What deeper principle would prefer majority rule to the sample if the sample were epistemically better? Procedural fairness seems to me to be indifferent. If three of us are deciding which movie to go to, we could choose a movie randomly (dumb), choose by majority rule, or choose one of us randomly as the chooser. All three are basically fair, with the last two including some extra element of aggregativity. Someone might offer this as the deeper principle: each person ought to have not just an equal chance of voting (or having their vote counted), but a vote (or a counted vote). But that’s not a deeper principle, that’s the proposition in question.

    Summarizing: I’m not arguing in Chapter 4 that majority rule is preferable to random choice. But I do think it is better on epistemic grounds (even if also on some other thinner grounds like responsiveness). I’m not sure whether it is better than a lottery to choose multiple voters, or even a single voter. I’m not sure if these would all count as democratic, but I don’t see any clear reason why they shouldn’t. And, in any case, that wouldn’t by itself be any argument for or against them.

Leave a Reply