deliberative democracy and the discursive dilemma0

8/4/2019 Deliberative Democracy and the Discursive Dilemma0

1/36

For Ph ilosophical Issues (Sup p. N ous) Vol 11, 2001

Deliberative Democracy and the D iscursive Dilemm a

Philip Pettit

Taken as a m odel for how group s should make collective jud gments and

decisions, the ideal of deliberative democracy is inherently ambiguou s. Consider the

idealised case where it is agreed on all sides that a certain conclusion shou ld be

endorsed if and on ly if certain premises are admitted. Does deliberative dem ocracy

recomm end that members of the group d ebate the premises and then individually

vote, in the light of that debate, on wh ether or not to supp ort the conclusion? Or does

it recomm end that m embers individually vote on the p remises, and then let their

comm itment to the conclusion be settled by w hether or not the group endorses therequired p remises? Is deliberative dem ocracy to enforce the discipline of reason at

the individual level, as in the first possibility, or at the collective level, as in the

second?

Deliberative-democratic theory has not addressed th is issue, perhap s because

of an implicit assumpt ion that it does not matter whether th e discipline of reason is

imp osed at the individua l or at the collective level. But that assumption is false and

there is no excuse for neglecting the issue raised. The d iscursive dilemma of my title

a generalisation of the d octrinal parad ox that has recently received attention in

analytical jurispru dence shows that the p rocedures distnguished can come apart.

Thus deliberative democrats must m ake up th eir minds on w here they stand in

relation to the issue; they cannot sit on th e fence.

This paper is an attempt to add ress the issue and look at the grounds on

which it may be resolved. In the first section, I give a brief account of the ideal of

deliberative dem ocracy, as I un derstand it. In the second , I introd uce the d iscursive

dilemma w ith the help of some stylised examples and then in the third section I show

why the issue that it raises is of relevance, theoretically and practically, to the

deliberative-democratic ideal. How should deliberative dem ocrats resolve that issue?

I argue in the fourth section that the role in wh ich republican theory casts

deliberative dem ocracy argues for preferring th e imposition of reason, where

possible, at the collective level. And then in the final section I argue for the


2/36

2

consistency of that position with th e main sorts of argum ent put forward by others in

defence of the ideal.

1. The id eal of delib erative dem ocracy

There are three issues on which deliberative democrats divide among

them selves. First, the question of how many contexts electoral, parliamentary,

industr ial, edu cational, and so on ough t to be dem ocratised. Second , the question

of how m any issues in any dem ocratised context ought to be und er democratic

control: just the choice of office-hold ers, or also the choice of policy-programs, or

perh aps the choice of some d etailed p olicies. And third , the question of how far a

democratic character serves to justify or legitimate a regime and pat tern of decision-

making, or at least to give them a p resump tive authority: to place the onus of

argum ent on the shoulders of those who wou ld not comp ly.

But there, or thereabouts, disagreement ru ns out. For no matter wh at their

d ifferences on such matters, deliberative democrats do show a remarkable degree of

consensus on how d emocracy shou ld be organised. Wherever dem ocracy is

instituted , however deeply democratic control runs, and w hatever standing it is

taken to give to decision-making, they agree that it ough t to be inclusive, jud gmental

and d ialogical (see the represeentative read ings in Bohm an and Rehg 1997; Elster

1998). These constraints spell out w hat I shall describe here as th e deliberative-

democratic ideal.

The inclusive constraint: all mem bers are equally entitled to vote on how toresolve relevant collective issues, or bu nd les of issues, with something less than a

unanimous vote being sufficient to determine the outcome.

The jud gmental constraint : before voting, members should deliberate on the basisof presumptively comm on concerns about w hich resolut ion is to be preferred.

The dialogical constraint: they shou ld conduct this deliberation in open andunforced dialogue with one another, whether in a centralised forum or in various

decentralised contexts.

The inclusive constraint means that deliberative democracy is to be contrasted

with elitist or authoritar ian schemes, even ones in which deliberation and dialogue

have an imp ortan t place. It will be satisfied in any context by having a representative


3/36

3

democracy, if dem ocratic control only run s to th e choice of office-holders, but the

general assum pt ion is that wh ere direct par ticipation by all mem bers is feasible

and is consistent w ith the general case for deliberative democracy it w ill be

preferred to indirect representation. The constraint includes the stipu lation that

un animity is not required for the determina tion of an ou tcome by voting, since a

combination of inclusiveness and u nanimity wou ld lead to a group s being unable to

reach a common view on most significant issues; un animity is probably achievable,

at best, only on very abstract constitutional m atters (Buchanan and Tullock 1962).

The jud gmental constraint has got two sides to it. First, it requ ires voters to

deliberate or reason about how they shou ld vote, not just vote in an u nreflective or

spontaneous or reflex manner. And second , it requires voters to deliberate about how

they shou ld vote on the basis of considerat ions as to what is best for the society as a

whole: what is likely to advan ce those common interests that p eople are capable of

recognising as common interests. This constraint need not itself specify any

par ticular conception of such comm on, perceived interests: that m ay itself be matter

for the sort of deliberation recommended. What it coun sels against is any pattern of

voting in which each ind ividual voter takes account only of what is good for his or

her particular coterie or corner or circle. The model of voting recomm ended und er

this constraint can be described as jud gment-voting rather than p reference-voting(Cohen 1986; Coleman and Ferejohn 1986; Brennan an d Pettit 1990). The idea is that

each voter should m ake up his or her own m ind as to wh at is for the good of the

group in question and should vote on th e basis of that jud gment, not on the basis of

bru te preference (Sunstein 1993) or bargained com prom ise (Elster 1998, 5-8; cf Pettit

1993, Ch.5).

The third , dialogical constraint in the ideal of deliberative democracy marks a

further, important level of differentiation. It ru les out the sort of plebiscitarian

dispensation in wh ich each participant privately forms his or her jud gment about

comm on p erceived interests, rather than doing so in d ialogue w ith others, and then

votes on the basis of that judgment. It is sometimes thought, on the basis of his

remarks about the danger of faction, that Rousseau embraced this plebiscitarian

ideal. According to Rousseau, so it is said, each voter is polled about his

independ ently reached choice, without any group deliberation (Grofman and Feld


4/36

4

1988, 570). But th is interp reta tion is almost certainly mistaken , if only becau se it

makes no sense of Rousseaus requiring th at the p eople come together in an

assembly. What he was anxious to guard against was not d ialogue and debate, with

the formation of ind ividu al jud gment that this fosters, but rather the th reat of some

individuals being so intimidated or imp ressed by others so deferential toward s

them that they vote according to the jud gments of those others, not according to

their own (Estlund and Waldron 1989).

The third constraint r equires that the d ialogue envisaged be open and

un forced, wh ile allowing that it may be centralised or decentralised. It mu st be open

in the sense that each can get a h earing and it mu st be un forced in the sense that no

one need fear to speak their mind ; it must ap proximate the conditions for ideal

speech that Juergen Habermas (1984, 1989) emphasises. Some will insist that

d ialogue mu st be centralised in a single forum, if talk of deliberative democracy is to

be justified (Goodin 1999). But I think th at it is better to leave that qu estion open and

to take the centralised or collective picture of deliberative dem ocracy as a m ore

specific version of a broad er ideal.

2. The discursive dilem ma

So much for the different elements in the ideal of deliberative democracy. In

this and the next section I want to show that the ideal is seriously un derspecified

and , in p articular, that there are tw o quite d ifferent w ays in which it m ay be

understood in any context.

My argu ment in this section derives from the recent discussion in

jur ispru den tial circles of what its analysts have sometimes called the d octrinal

paradox (Kornhauser and Sager 1986; Kornhauser 1992; Kornhauser 1992;

Kornh auser and Sager 1993. See too Chap man 1998a; Chapman 1998b; Brennan

1999). This is a parad ox that arises when a mu lti-mem ber court has to make a

decision on the basis of received d octrine as to the considerat ions that ought to be

taken into accoun t: on the basis of a conceptu al sequencing of the issues to be

decided (Chapm an 1998a) . I describe the problem, in its general form, as a

d iscursive dilemma. I prefer the w ord d iscursive, because the problem in question

is not tied to the acceptance of comm on doctrine, only to the enterp rise of making


5/36

5

group judgments on the basis of reasons. I prefer the w ord d ilemm a, because while

the p roblem generates a choice in which each op tion has its d ifficulties, it does not

constitute a paradox in any strict sense. My analysis of the d iscursive dilemma

derives directly from that jurisprud ential literature, though many of the points I

wan t to make do not appear there; wh ere they do appear, I explicitly acknowledge

them.

The discursive d ilemm a: the conjunction case

In order to introdu ce the d iscursive d ilemma, consider an issue that m ight

arise in a w orkp lace, among the employees of a company: for simp licity, as we m ay

assum e, a compan y owned by the employees. The issue is whether to forego a pay-

rise in order to spend the money thereby saved on introducing a set of workp lace

safety measures: say, measures to guard against electrocution. Let us supp ose for

convenience that th e employees are to make the decision perhaps because of prior

resolution on the basis of considering two separable issues: first, whether the

danger is serious: say, whether the risk of electrocution is above some tolerable

threshold; and second, whether the pay-sacrifice would make it possible to buy a

safety measu re that w ould be effective in reducing the risk below that threshold , in

the event of its having been above it. If an emp loyee thinks that the d anger is

sufficiently serious, and the safety m easure sufficiently effective, he or she w ill vote

for the p ay-sacrifice; otherwise they will vote against. And so each w ill have to

consider the seriousness issue and the effectiveness issue and then look to w hat

should be conclud ed abou t the pay-sacrifice.

Imagine now that after approp riate dialogue and deliberation the employees

are disposed to vote on the relevant prem ises and conclusion in the p attern

illustrated by the following m atrix for a group of three workers. The letters A, B, and

C represent the three employees and the Yes or No on any row represents thed isposition of the relevant em ployee to adm it or reject the correspond ing premise or

conclusion.

Serious danger? Effective measure? Pay-sacrifice?

A. Yes No No

B. No Yes No

C. Yes Yes Yes


6/36

6

Matrix 1

If this is the pa ttern in w hich the employees are inclined to vote, then what

decision or jud gment w ill be mad e by the grou p und er deliberatively d emocratic

procedures? It tu rns out that a d ifferent d ecision will be made, depending onwh ether the group jud gment is driven by how m embers jud ge on the premises or by

how they jud ge on the conclusion. Looking at the matrix, we can see that thou gh a

majority rejects the pay-sacrifice, a majority supp orts each of the premises. If we

think that the views of the employees on th e conclusion should determine the group-

decision, then w e will say that the grou p-conclusion shou ld be to reject the pay-

sacrifice: there are more Nos than Yess in the final colum n. But if we think that

the views of the employees on the p remises should determine the group-decision,

then w e will say that the grou p-conclusion shou ld be to accept the pay-sacrifice:

there are more Yess than Nos in each of the premise colum ns. The fact that

socially aggregating the conclusion-jud gments gives us a d ifferent result from

socially aggregating the prem ise-jud gmen ts illustrates the d iscursive dilemm a.

It should be clear that the discursive dilemm a will generalise from the three-

person case to a group of any size. All that is required for the paradox to arise is that

a majority in the grou p supp orts the first of the premises, a different majority

supp orts the second B and C rather than A and C and the intersection or

overlap of those majorities C in our example represents only a minority in th e

group as a w hole. The fact that th e intersection represents only a minor ity explains

why there is a majority against the conclusion. But n ot only will the parad ox

generalise to group s of any size. It should equally be clear that it will generalise in

other w ays too: for example, to a case where there are any nu mber of premises, not

just tw o; and to cases that arise in quite d ifferent dom ains.

The structure requ ired for the sort of parad ox illustrated to be capable ofgeneralising can be sum med up in these points:

a. there is a conclusion to be decided am ong a grou p of peop le by reference to a

conjunction of independent or separable prem ises the conclusion w ill be

end orsed if relevant p remises are endorsed, and otherw ise it will be rejected;

b. each m ember of the group forms a jud gment on each of the prem ises and a

correspond ing jud gment on the conclusion;


7/36

7

c. each of the p remises is sup por ted by a majority of members but those majorities

do not coincide w ith one another;

d . the intersection of those majorities will sup por t the conclusion, and the others

reject it, in view of a; and

e. the intersection of the majorities is only a minor ity in the grou p as a w hole.

The discursive dilemm a, as illustrated in ou r workplace examp le, stems from

the fact that the group can m ake its jud gments in the conclusion-driven w ay or in the

prem ise-driven w ay and that in many cases those procedures yield different results.

It constitutes a d ilemma, so far as each op tion has its problems. As the examp le

shows, going the conclusion-driven w ay means adop ting a course that is inconsistent

with the premises endorsed by the group and going the premise-dr iven w ay means

adopting a course that a majority ind ividually reject. Going the first way means

sacrificing collective rationality for the sake of respon siveness to ind ividuals, going

the second means sacrificing responsiveness to ind ividuals for the sake of collective

rationality.

There are familiar p ractices of group d eliberation and decision-making

correspond ing to those different app roaches. Thus the group wou ld go the

conclusion-driven w ay if members entered into d eliberation and dialogue and then

each cast their personal vote on whether to endorse the pay-sacrifice or not; in that

case the decision w ould be against the pay-sacrifice. The grou p w ould go the

prem ise-driven w ay, on the other hand , if there was a chairperson w ho took a vote

on each of the prem ises say, a show of hand s and then let logic decide the

outcome; in th is case the d ecision w ould be in favour of the p ay-sacrifice.

The discursive d ilemm a: the d isjun ction case

But the sort of discursive dilemma illustrated is not the only sort there is.

Another version of the dilemm a arises wh en the issue facing a group has to be

decided , not by reference to a conjunction of premises, but rath er by reference to a

d isjun ction. In this case the conclusion w ill be endorsed if any one of the premises is

endorsed any, not a ll and otherw ise it w ill be rejected. (It might also be

required, of course, that any two of three or more premises wou ld have to be


8/36

8

endorsed or any three of four or more premises, and so on but I shall ignore

that possibility here.)

The second var iety of the d iscursive d ilemma can be illustrated with another

sort of decision that ou r w orkplace group might face. Sup pose that the group has todecide on whether to introdu ce a clocking-in and clocking-out system an d th at by

their shared lights the decision should be m ade on the ground s, and only on the

ground s, either that such a time-check w ould increase produ ctivity or that it wou ld

serve to reassure everybody that others are pu lling their weight in the joint

enterp rise. Let someone d ecide the p roductivity question in the affirmative, or let

them decide the reassurance question in the affirmat ive, and their conclusion will be

that the time-check shou ld be introduced; otherwise they will conclud e that it ough t

not to be introduced. The conclusion will be jud ged by reference to a disjun ction of

those independ ent premises: endorsing either premise on its own will be enough to

sup por t the end orsement of the conclusion; rejecting both of the p remises will be

enough to supp ort the rejection of the conclusion.

Imagine now , as in the conjun ction case, that after appropr iate dialogue and

deliberation the emp loyees are disposed to vote on the relevant premises and

conclusion in the pattern illustrated by th e following matrix for a group of three

workers. The letters A, B, and C represent the th ree emp loyees, as before, and the

Yes or No on an y row represents the d isposition of each to ad mit or reject the

corresponding premise or conclusion.

Prod uctivity issue? Reassurance issue? Time-check?

A. Yes No Yes

B. No Yes Yes

C. No No No

Matrix 2

Where votes follow th is pat tern, then the premise-driven app roach will yield

one group-jud gment, as in the previous sort of case, the conclusion-driven approach

another. Each of the p remises is rejected by a m ajority, as we can see by considering

the p remise columns, so that if we go by the p remises, then w e will take the group-

decision to be aga inst the time-check prop osal. But the conclusion, on the other hand ,

is endorsed by a majority so that if we go by the conclusions the employees


9/36

9

ind ividually reach, then we will take the group-decision to be in favour of the time-

check.

This sort of dilemm a allows of generalisation in just the sam e way as the first.

The salient points to preserve in this case mirror the points mentioned with the other:a. there is a conclusion to be decided am ong a grou p of peop le by reference to a

d isjun ction of independent or separable premises the conclusion will be

end orsed if any of the prem ises is end orsed, and otherw ise it will be rejected;

b. each m ember of the group forms a jud gment on each of the prem ises and a

correspond ing jud gment on the conclusion;

c. each of the premises is sup por ted by a m inority of members bu t those minor ities

do not coincide w ith one another;

d . the union of those minorities will sup por t the conclusion, and the others reject it,

in view of a; and

e. the union of the minorities is a majority in the grou p as a w hole.

In this case, as in the last, there are familiar decision-making practices that

differ in implementing either a conclusion-driven mode of group-judgment or a

premise-driven m ode. The workp lace ballot in which each casts his or her vote

privately would implement th e conclusion-driven m ode, whereas the workplace

meeting in wh ich the chair takes a vote on each of the relevant considerations and

then annou nces the significance for the conclusion wou ld imp lement the premise-

driven mode.

3. The relevance of the dilem ma for deliberative d emocracy

The relevance in p rinciple

Und er the regime of deliberative democracy, it is required th at with any

logically connected p ropositions in the d omain of discussion the prem ises and

conclusion of our examples people make up their mind s about the prop ositions in

such a way that reason is satisfied. They do not end orse inconsistent or otherw ise

incoherent sets of propositions; they do n ot fail to derive conclusions that are

supp orted, even saliently sup ported, by wh at they already end orse; and they do not

follow a procedu re say, one of unanimitarian voting that gives them nothing to

say on m ost issues (List and Pettit 2000).


10/36

10

The question raised by the d iscursive dilemm a is wh ether this d iscipline of

reason is meant to apply to each ind ividual, taken singly, or to the group taken as a

whole. Und er the conclusion-driven, majoritarian way of voting the discipline is

app lied to each ind ividual, un der the prem ise-driven, majoritarian way of voting it is

app lied to the group taken as a wh ole. The question is a pressing one, where it arises,

because it represents a hard choice or d ilemm a. Let a group ind ividualise reason, and

it will ensure responsiveness to ind ividuals in its collective view on each issue bu t it

will run the risk that th e views w ill be irrational. Let a grou p collectivise reason, and

it will ensure the ra tionality of the collective views main tained but run the risk of

adop ting a view on one or another issue that is unresponsive to the views of

individuals on that issue.

The choice with w hich groups a re faced is not necessarily a choice between

following a conclusion-driven, majoritarian p rocedure and following a p remise-

driven, majoritarian one. Individualising reason and ensuring responsiveness to

individual views w ill always involve having m embers vote, whether on a

majoritarian or som e other basis, about each of the issues involved , and th en letting

that vote d etermine the collective view; this, in effect, is a generalisation of the

conclusion-driven procedu re. But collectivising reason an d ensur ing collective

rationality will not always involve privileging certain issues in the manner ofpremises and letting the collective views on those issues logically determ ine the

collective view on the conclusion. Consistently with behav ing in a deliberative

democratic manner, the group might let the presum ptive set of collective views be

determined by having m embers vote on each issue and only consider mod erating

those views amend ing one or other of them in the event of the collective views

proving to be irrational in some way. The amendmen t chosen might involve letting

the views w ith the p resump tive status of premises if that can be agreed

determine the view on the conclusion. But it might also involve hold ing onto that

conclusion and revising the collectively endorsed view on one of the premises. In

short, it might involve practising mod us ponens letting the premises dictate the

conclusion or modus tollens keeping the conclusion and revising one or more

of the prem ises.


11/36

11

To sum up , then, the d iscursive dilemm a is relevant in pr inciple to

deliberative dem ocracy so far as it raises the qu estion as to whether deliberative

dem ocrats require people to be d isciplined by reason at the ind ividu al or, where this

is feasible, at the collective level. There is a dilemm a involved , because each of the

options ava ilable has its problems. Let peop le be d isciplined by reason at the

ind ividual level and they may collectively endorse an irrational set of propositions.

Let them be d isciplined by reason at the collective level and they m ay collectively

endorse a conclusion that a m ajority of them ind ividually reject.

The relevance of the dilemm a in p ractice

But is the d iscursive dilemma relevant in p ractice to the way the ideal of

deliberative dem ocracy is specified? Some m ight say that it is not, on the grou nd s

that it is rarely going to be a feasible option for people to d iscipline themselves by

reason at the collective level and that all deliberative democracy can require is that

ind ividuals do this in the formation of their opinions at the individu al level. The

problem raised will be that if people are to discipline themselves collectively by

reason, then they must agree on what are the relevant considerations by reference to

which various issues are to be d ecided . But that agreemen t, so it w ill be said, is rarely

going to be available. Thus sceptics will maintain that th e assump tion that the parties

in my w orkp lace examples agree on the prem ises by reference to which the pay-

sacrifice and time-check issues are to be decided is an id ealisation rarely satisfied in

practice.

This point is well taken for large scale collectivities such as an electorate and it

may well be the case that all deliberative democrats can seek in such contexts is a

d iscipline of reason imp osed at the ind ividual level.1 But the point does not hold

more generally. There are two sorts of groups that rou tinely confront the possibility

of collective self-discipline, and that mu st therefore d ecide on whether to followreason at the individual or at th e collective level. And there is a further , more general

consideration w hy any grou ps that seek to advance a comm on pu rpose, and not just

the tw o sorts ment ioned, are going to confront that p ossibility in the course of their

development.


12/36

12

The first sort of case is that of a grou p th at is charged by an external authority

with making certain d ecisions on the basis of designated considerations, and on that

basis only. Examples of such grou ps are appointment and prom otions committees;

committees charged w ith deciding w ho is to win a certain p rize or contract; tru sts

that have to m ake jud gments on the basis of a trustees instructions; associations or

the executives of associations that have to justify their actions by reference to th e

groups charter; corporations that have to comp ly with policies endorsed by their

shareholders; pu blic bodies, be they bureau cratic committees or app ointed board s,

that have to discharge specific briefs; and governm ents that are more or less boun d to

par ty program s and p rinciples. In all of these cases deliberative dem ocrats will want

their ideal to apply and in each case there will clearly be an issue as to whether the

d iscipline of reason shou ld be imp osed at the individual or at th e collective level.

In the second sort of case, it will be a mat ter of internal aspirat ion in a grou p

that m embers find comm on groun ds by w hich to justify whatever line they

collectively take. Think of the political movem ent that has to w ork ou t a policy

program ; or the association that h as to decide on the terms of its constitut ion; or the

church that has to give an accoun t of itself in the public forum; or the learned

academy that seeks a voice in the larger world of politics and journalism. In such

cases members of the group m ay not have access to an anteceden tly agreed set ofconsiderations on th e basis of which to justify par ticular jud gments. But their

identification w ith one another w ill supp ort a w ish to reach agreement on such a set

of reasons. To the extent th at tha t w ish gets to be satisfied, the d iscursive dilemma

will generate the issue as to whether the group shou ld make its decisions in a

premise-driven or conclusion-driven way.

These two cases are special in a certain w ay. They contrast with the case of a

group that has to advance a pu rpose, as those group s have to do, but that may be

quite happ y to have its members reach a jud gment on an y issue without their

agreeing on th e reasons that supp ort that judgm ent; a majority may vote for the

jud gment bu t, theorising it on ly incompletely (Sun stein 1999), different people

within that m ajority may supp ort it for d ifferent reasons. Will a group of this laid-

back kind ever confront the possibility of imposing reason at a collective level and

ever have to face the discursive dilemma? I believe that it will (Pettit 2000).


13/36

13

Like the other grou ps envisaged , this sort of group is bound to genera te a

history of jud gments that it is on record as making, when it seeks to pursue its

pu rposes. But those past jud gments w ill invariably constrain th e future jud gments

that it can rationally maintain at the collective level: this, in the way that a p olitical

par tys past comm itments on red ucing taxes but increasing expenditure in certain

areas will constrain the other decisions it has to make about expenditure. And so

there will be a serious question for the group as to w hether or not it should impose

reason at the collective level, letting later jud gments be d ictated in a p remise-driven

way by earlier ones or d isowning some of the earlier jud gmen ts; after all, the lesson

of the discursive dilemm a is that thou gh every ind ividu al holds a rational set of

views on the issues involved, voting independ ently on each issue could lead to a

collectively irrational result. That question will arise to the extent that effective, joint

action requ ires collectively rational jud gments. And it will assume a par ticularly

sharp form so far as members or outsiders are ready to charge the group, perhaps

even m ock it, for any failu re of collective reason .

The upshot, then, is this. Take those groups w here the ideal of deliberative

democracy apparently app lies and w here the collectivisation of reason is not made

infeasible by num bers or d isorganisation. Are such grou ps likely to face the hard

choice betw een ind ividualising an d collectivising deliberation? Yes, they are. Theywill certainly face that choice if they are comm issioned to m ake their d ecisions on the

basis of certain criteria or if they succeed in establishing criteria by w hich to m ake

those decisions. But even if they d o not have such criteria imposed on th em, or

cannot agree on any criteria by w hich to opera te, they will still face the hard choice

represented by the discursive d ilemm a; they will not be able to rest content w ith

their mem bers each voting accord ing to their own particular reasons on the different

issues that come before the group . The reason is that any such group will build up a

history of jud gments over time and those jud gments are bound to provide reason,

now here, now th ere, for taking a certain line on some new issue. Thu s the group will

have to choose either to let their line on this new issue be governed by m ajority vote,

at a r isk of proving inconsistent over time, or to en force collective reason in the event

of such inconsistency, whether by letting the previous jud gments d ictate that line or


14/36

14

by allowing the new vote to stand and by revising one or more of those jud gments.

The discursive dilemm a is, qu ite simp ly, un avoidable.

4. The resolution in repub lican th eory

The republican argum ent for deliberative democracy

Repu blican theory, as I have argu ed elsewhere, pu ts a premiu m on p eoples

enjoying freedom as non -domination: that is, on peoples having a status such that

ideally no one else is able to interfere arbitrarily in th eir lives. If any other person or

agency is able to interfere in their lives, then they m ust be forced to track p eoples

perceived interests in the interference they practise; they must not have a pow er of

arbitrary interference.

That ideal raises the question of how th e state can be blocked from enjoying a

pow er of arbitra ry interference in th e lives of citizens. The state is a necessary

institution and it necessarily has a power of interfering with p eople: it cannot operate

without being able to tax, legislate and penalise the governed . So how is its pow er of

interference to be mad e non-arbitrary? Repu blican constitutional theory is built

around that question and has consistently sought to describe various devices

whereby non-arbitrariness may be fur thered . These includ e familiar mechanisms

such as d emocratic election, rule of law, separation of pow ers, limitation of tenu re,

rotation of office, and so on.

In this republican tradition of constitut ional discussion, one important stran d

has been the idea that if the states pow er of interference is to be rendered n on-

arbitrary th en whatever other devices are in place, peop le must be able to contest the

decisions made by various arms of governm ent. They must have access to the

reasons sup porting those decisions and they must be able to contest the sound ness of

those reasons or the degree of sup por t they offer to the decisions mad e. Moreover

they m ust be in a position, ideally, to expect that such contestations will be heard ,

will be impartially adjudicated and, if necessary, will be imp lemented against those

in government. The general m essage is that so far as a governm ent is effectively

contestable, to that extent it is less likely to enjoy arbitra ry p ower .

The effective promotion of contestability in th e political sphere requ ires a

variety of institutions, especially if it is to guard against unw ieldy levels of


15/36

15

complaint, and I have tried elsewh ere to review som e of them (Pettit 2000). From our

point of view, how ever, only two observations are relevant. The first is that p ut ting

in p lace a regime of participatory or representative d emocracy is essential in

guard ing against certain possibilities of non-constestability and domina tion: the

colonial, the au thor itarian, and so on . And th e second is that any such regime still

leaves striking p ossibilities of non-contestability and domina tion in p lace:

possibilities associated w ith the ty rann y of the democratic majority, in p articular, and

the tyran ny of what we m ight describe as the democratic elite: those in the corridors

of power for examp le, in the bu reaucracy, the cabinet, the courts, the p risons, or

the police force who can impose their own w ill in how they interp ret and

implement democratic policy.

How are people to be emp owered in relation to d emocratic majorities and

democratic elites? How are they to be given a power of contestation against them?

Whatever else is necessary, it seems clear that they m ust be able to ask after the

reasons that supp ort the decisions, they must be able to question the relevance of the

reasons and they m ust be in a p osition to expect a fair hearing. And all of that is

going to be possible, of course, only so far as th e democratic bodies in question

opera te in a deliberative mod e. There must be a dispensation of deliberation in p lace

in the commun ity as a whole, and in the microcosm of par liament, which establishesa currency of considerations that are ad mitted on all sides to be relevant to the

doings of government. There mu st be a commitment in the d ifferent arm s of

governm ent to justifying wh atever decisions are taken by reference to the

considerations that are relevant, by comm on consent, in their case. And it mu st be

possible for private individuals, or perh aps for designated representatives, to

challenge such decisions on the ground s that the reasons quoted are not soun d or do

not offer the requ isite suppor t for the decisions taken. It is only in the event of

democracy having this deliberative cast that contestability, and ultimately non-

arbitrariness, can be fur thered .

This republican argumen t for deliberative democracy applies also, of course,

to forums beyond those of government. Take the workplace comm unity or the

commun ity organisation or ind eed the family. Even if decisions are taken

democratically in such a bod y, there will be little protection against arb itrariness


16/36

16

short of exit unless the democracy in question op erates in deliberative mode,

giving ind ividuals a chance to contest the decisions made. If decisions are made on

the basis of interest-group p olitics, or bargaining from d ifferent levels of power, then

there w ill be no grou nd s on which any contestation can be mad e. Bru te force or

naked preference will ru le.

Cass Sunstein (1993) probably h as th is repu blican case for d eliberative-

dem ocratic procedure in m ind wh en he describes deliberative democracy as a

republic of reasons. Just as people are in a position to kn ow where they stand in

relation to a cour t jud gment, only so far as the jud ges have to provid e statements of

their reasons, so more generally people can know where they stand in relation to

pu blic decisions only if they know w hat the groun ds ad du ced in sup port of those

decisions are. They w ill not be able to take a stand in relation to p ublic decisions, if

those decisions are the outcome of interest-group bargaining or of voting on th e basis

of naked, unargued p reference. Such non-deliberatively generated d ecisions would

have the profile of dictats or fiats from on high , where the prod ucts of deliberative-

dem ocratic procedure wou ld p resent th emselves as reasoned well-reasoned or

bad ly-reasoned jud gments that p eople are in a position to examine, assess and if

necessary, challenge.

App lying the republican argument

The contestability argu men t for deliberative dem ocracy ough t to have

persu asive force, qu ite apart from its connection with republican theory (see Pettit

2000). It is of particular interest here because un usu ally among such argum ents, as

we shall see in the next section it provid es a firm ground for wanting, where that

is feasible, to impose the d iscipline of reason at the collective, not at the ind ividual,

level.

If I am to be able to judge the actions of another ind ividual or agency in

relation to alleged reasons and , ultimately, if I am to be able to contest those

actions in the relevant sense then th e individual or agency mu st prove answerable

to those reasons. The ind ividual or agency must be d isposed to act as the reasons

require or, at the least, to adjust in response to the criticism that the reasons ad duced

in supp ort of something they did are not app ropriate or that what they did was not


17/36

17

supp orted by those reasons. In a word , the ind ividu al or agency must p rove to be

conversable: that is, mu st prove to be an interlocutor w hose words can be taken at

face value and, where ap propr iate, invoked in effective criticism of the things they

later d o.

A grou p th at records its assent to certain considerations on the basis of

majority vote and wh ich then p roceeds to make its jud gments and decisions

accord ing to the p remise-driven stra tegy will prove conversable in this sense. Its

jud gments, decisions and actions will track those reasons in the ord inary case. And

wh ere they do not, or where the reasons add uced are shown to be in some way

inappropr iate, then it can be called effectively to book. The group will acknowledge a

role for the consideration of reasons or p remises in relation to any case; it will decide

the group comm itments in regard to those reasons or premises on the basis of

majority vote, or some such measu re of social aggregation; and it w ill then accept

that how it jud ges or decides on the case in hand should be answ erable to those

comm itments. You can deal with such a group as you might d eal with an individual

person ; indeed the grou p w ill deserve to be regard as a legal or at least cultu ral

person : a un ified, corporate entity.

A group that m akes its jud gments and decisions on the basis of

ind ividualising reason, however, will have a very different p rofile. What the

d iscursive dilemm a reveals is that such a group m ay reach theoretical and p ractical

conclusions that a re completely ou t of kilter w ith the m ajority d isposition of its

members in relation to related p ropositions. There will be little or no poin t in

interrogating th e group for its view, und er majority rule, of the different p remises

that m ight be thought to be relevant, for example, and might be acknowledged as

relevant by members of the group . For it will be a perm anent p ossibility that the

conclusion it reaches is not th e conclusion that those majority views w ould

deliberatively support.

Consider the workforce in our earlier example. We might discover that the

majority view among its members is that electrocution is a serious danger, and that a

pay-sacrifice would solve the problem and that those premises imply that a pay-

sacrifice would be desirable; and yet it shou ld be no surpr ise to find that the group

comes dow n against a p ay-sacrifice. There is no talking to a group th at operates like


18/36

18

this. It is inherently unconversable. Where the grou p that runs by prem ise-dr iven

jud gment and decision is a un ified, corporate entity, the group envisaged h ere wou ld

be a d isparate, aggregate sort of thing. It would not be one, but m any.

Consider an analogy. Imagine that I am composed of man y differentpersonalities, as in some sense I undou btedly am . There is the ecologically mind ed

self, the econom ically mind ed self, and the socially mind ed self that likes to keep u p

with the neighbours. And now supp ose that I face a decision about w hether to buy a

Volvo. That d ecision, let us say, will be determined by the views I take on tw o issues:

one, whether to get a car; and tw o, whether th e best car for someone like me to buy,

asssuming they d o buy a car, is a Volvo. It is quite possible, that faced w ith a d ecision

like this, my d ifferent selves will go qu ite different ways. My economic self, A, may

say that a car is desirable, but not a Volvo: its too expensive. My ecological self may

say that a car is und esirable but agree that a Volvo is the best sort of car to bu y: it is

environm entally the friend liest kind available. And m y social self may hold that it

wou ld be good both to buy a car, in order to keep up with the neighbours, and that

from th is point of view a Volvo is the best sort there is. This, then, is the pattern of

voting among my selves.

Car desirable? Volvo a desirable car? Get a Volvo?

Self A. Yes No NoSelf B. No Yes No

Self C. Yes Yes Yes

Matrix 3

Constituted though I am in such a m ultiplex mode, I would pass as a un ified,

conversable agent were I to make my jud gments and d ecisions in this and similar

cases on a premise-driven basis. I would decide to buy a Volvo in the case on hand ,

reporting as I presum ably would d o that I think a car desirable and that I view a

Volvo as the m ost desirable car there is; after all, those are the views d efend ed by a

majority of my selves. But imagine that I mad e my jud gments and took my d ecisions

in a way that individua lised reason in my d ifferent selves, rather than collectivising

it in my p erson as a whole. I wou ld d ecide in a case like this not to bu y a Volvo. And

yet, interrogated about the relevant reasons reasons that I myself might d eclare to

be relevant I wou ld be found to defend considerations supp orting the pu rchase of


19/36

19

a Volvo. It shou ld be clear that d id I operate in th is way, then I could not p rove to be

a conversable subject. I would be taken at best as a mu ltitude of disorganised voices,

at w orst as someone without anything recognisable as a mind.

Might the dem and s of contestability be satisfied to the extent that theind ividuals in a group , thou gh not the group itself, are conversable and contestable

agents? Might they be satisfied in the w ay in which they are satisfied by the mu lti-

jud ge cour t? In th is case, the grou p is relatively small and the d iffering ind ividuals or

subgroup s give their differing reasons for taking their d ifferent lines, or ind eed their

d iffering reasons for taking the sam e line. It is not clear how satisfactory a model the

jud icial bench offers, given evidence that mem bers switch votes in ord er to avoid

problems and given recent critiques of the system (Kornh auser an d Sager 1993;

Stearns 1999). But those issues aside, I think that there are good reasons w hy

contestability should be secured at the level of the group , not at the level of the

different individuals who comprise it.

The group is the entity that continu es over time despite changes in its

membership, and that fact alone makes it the salient agency that w e should w ant to

be contestable and conversable. And not only does the group continue in time across

changes in mem bership. It will often have so many members that no one could hope

to secure contestability at the ind ividual level. It w ill often be, not just the salient

agency to m ake contestable, but the on ly feasible one available for that role.

These considerations of salience and feasibility are suppor ted by a thought

that w ould ap ply, even in their absence. It is only und er a regime of group-level

contestability that th e relevant judgm ents will be made in a w ay that answers to

reasons alone and that can be challenged on the basis of reasons alone. Where a

group makes its jud gments in the conclusion-driven way, there are two d istinct

factors that d etermine w hat judgment it makes in a given case: one, the reasons thatreceive majority supp ort among mem bers of the group ; and two, the degree of

overlap betw een the supp orting m ajorities. But the second factor is not relevant from

the point of view of contestability and so it is better for the group to give authority to

the first factor only, as it d oes und er the p remise-driven procedure an d, more

generally, un der any p rocedure that collectivises deliberation.


20/36


21/36

21

There are m any d ifferent argu ments in supp ort of deliberative democracy to

be foun d in the recent literature. The question that we m ust consider, then, is

whether th e position taken in favour of the prem ise-driven version of the ideal

and , more genera lly, in favour of the version that collectivises deliberation runs

seriously coun ter to any of those other argum ents. I devote the final section to that

issue.

Argum ents for making democracy deliberative

Most of the argu men ts in the literatu re focus on the virtue of making

democracy deliberative: that is, on th e benefit to a dem ocratic process or society of

having d eliberation of an inclusive, dialogical kind (see Cohen 1989; Dryzek 1990;

Sunstein 1993; Benhabib 1994; Gutmann and Thompson 1996; Bohman and Rehg

1997; Elster 1998). Some assert that making democracy deliberative should help to

ensure that peoples preferences are reflective and informed , not just the brute

produ ct of their ad apta tion to circum stance (Sun stein 1993); or that it should enable

peop le to do better in reaching beyond the chasms of difference that separa te the

members of certain grou ps, even if it does not bring them into consensus (Youn g

1990); or that it shou ld stretch peoples imagination and emp athy as they are forced

to take a general point of view (Goodin 1999). Withou t alleging any su ch

psychological transformation, other argu ments maintain that m aking democracy

deliberative shou ld a t least have th e effect of screening ou t self-regard ing concerns in

favour of more public-spirited considerations, thereby approximating or advancing

an ideal of pu blic reasoning am ong free and equal participan ts (Habermas 1984,

1989; Elster 1986; Cohen 1989; Elster 1998). And yet a further range of arguments

urge that making dem ocracy deliberative would p romote such effects as legitimising

whatever d ecisions are reached, making th em m ore likely to take accoun t of the

relatively pow erless, increasing transparency among members of the group, or

promoting just outcomes (see Elster 1998, 11, for a summary).

There is no likelihood that any arguments of this kind could be invoked

against the view that deliberative democracy ough t to impose reason at the collective

rather than th e individual level. We do not eliminate d eliberation by m aking

deliberative democracy prem ise-driven, for example, and thereby imposing reason at

the collective level, and benefits of the kind invoked in these argu ments all look to be


22/36

22

consistent w ith such a p rocedure. Consider the case of the workforce deliberating on

whether to introd uce a safety measu re or to establish a time-check. Even if they agree

to reach a decision in a p remise-driven w ay, they will first be required un der th e

deliberative-democratic ideal to d iscuss the matter in pu blic. And th e benefits that

come of making dem ocracy deliberative, accord ing to the argum ents under

consideration, would all seem to be available under the prem ise-driven w ay of doing

things; in p articular, they wou ld seem to be available as read ily and as richly as

under the alternative. The premise-driven procedure m akes ample room for

achieving the psychological and other transformative effects mentioned in the

argum ents in qu estion. And so of course wou ld any other p rocedure that

collectivised reason in a d emocratic way: say, as mentioned earlier, a procedure that

allowed mod us tollens as well as mod us p onenes.

Not only that. Among such benefits there are some that look mu ch more likely

to be realised u nd er a reason-collectivising p rocedure. With many democratic bodies,

it is importan t by our general lights tha t mem bers not be prejud iced in the conclusion

they d raw , for example, and not be lazy about the consideration they give to relevant

premises. But if they are allowed to vote their ind ividual jud gments on the

conclusion to be decided, then there is more room for prejud ice and laziness than

would be available under a collectivising procedure. Not having to give theirjud gment on the relevant premises, mem bers of the body m ay find it easier to hide

the influence of prejud ice on their views. And not having to give their jud gment on

each p remise, members may be temp ted, once they have made a p remise-jud gment

that d etermines their conclusion-vote, not to bother giving consideration to the other

premises relevant.

Argum ents for m aking deliberation dem ocratic

But there is another sort of argu men t for the ideal of deliberative dem ocracythat r aises a sharp er challenge to the republican position. This is the type of

argu men t that makes a case, not for having deliberation present in democratic

process, but rather for having d emocracy present in d eliberative process; it argues for

making d eliberation d emocratic, not for m aking d emocracy deliberative. The claim is

that if there are matters of truth involved in p olitical deliberation as in the

question, most abstractly, of whether this or that is in the comm on interest then


23/36

23

the chance of reaching the tru th, or of reaching the tru th accord ing to received views

of reliability, is increased by having in p lace a regime of democratic decision-making.

The argum ent may be recast for goals other than th at of maximising the chance of

reaching the tru th or m inimising the chance of falsehood . It app lies just as read ily to

a more specific goal like that of minimising the chance of a false negative, for

example: say, minimising the chance of find ing an inn ocent d efend ant gu ilty. But I

shall restrict myself here to th e original truth-centred case.

This type of argum ent has been forcefully defended by David Estlund

(1993a;1993b; 1997) but th e most famou s version, and the one I shall consider h ere,

derives from the Marquis de Cond orcets work in d efence of his so-called jury

theorem (Black 1958; Condorcet 1976, 33-70; Estlund 1994). The theorem presup poses

that the voters are independ ent of one another : while they may certainly form their

opinions on the basis of d ialogue, none of them votes in a way that is blindly

deferential to others (Estlund and Waldron 1989; Estlund 1994). It shows that if

voters each have the same, greater than evens chance of being right on some yes-no

issue then, first, their collective resolution of the issue, un der majority ru le, will have

a yet greater chance of being r ight and, second , it w ill have a p rogressively greater

chance of being right as the size of the group increases. Ind eed the latter element is

assured , even if some voters have a less than evens chance of being r ight; it w ill holdso long as the average chance of voters being ind ividually right is greater than evens

(Owen, Grofman and Feld 1989).

Let competence in a person be identified w ith the probability that on being

forced to choose between th e answers, they w ill choose the right one. Sup pose that

the person h as to choose, then, between jud ging that p and jud ging that not p. Their

competence will be given by the p robability associated w ith its being the case: either

that p , and they jud ge that p; or that not p , and they judge that n ot p. The message of

the jury theorem is that if the the competence of individual voters is greater than 1/ 2,

then th e comp etence of the group operating under m ajority rule will be greater still

and will app roach 1.0 as the number of mem bers increases.

We cannot go through the p roof of the jury theorem here for an excellent

presentation, see Estlund (1994) but th e idea is reasonably intu itive. Sup pose that

each of us in a certain grou p has a two th irds chance of being right on some sort of


24/36

24

issue. If I provide an answ er on m y own as that sort of issue comes up instance after

instance, then in th e long run I may expect to be right tw o third s of the time. But if

three of us, A, B and C, decide wh at to say instance by instance, on the basis of a

majority vote, then un der the stand ard probability calculus (Skyrm s 1966, 111-37),

our p erformance ought to be better than mine on my own.

Here, in br ief, is wh y. On any occasion the group will be right so long as at

least two of us so long as three of us or any tw o of us are right. If the chance of

any on e of us being right on a given occasion is two th irds, then the chance of at least

two of us being right is the chance that A and B but not C are right, or that A and C

but n ot B are right, or that B and C but not A are right, or that A and B and C are

right. The chance that A an d B but not C are right assum ing independ ence is

the chance of A being r ight times the chance of B being right times the chance of C

not being right: i.e. 2/ 3 x 2/ 3 x 1/ 3, or 4/ 27; and similarly for the other pairs. The

chance that A and B and C are all right is 2/ 3 x 2/ 3 x 2/ 3, i.e. 8/ 27. And so, the

chance that at least two of us are right is the sum of these four figures: 4/ 27 plus 4/ 27

plus 4/ 27 plus 8/ 27; with d isjuncts that are mu tually incompatible the probability of

the d isjun ction is the sum of the disjun ct-probabilities. Thus the chance that at least

two of us are right on any occasion is 20/ 27; and this is higher than the two th irds

chance 18/ 27 of any on e of us being right. The reasoning continues in the sameway for each increase in the size of the group , sup por ting the Cond orcetian

conclusion that group competence app roaches 1.0 as the size of the grou p increases.

The question now is whether the jury theorem p rovides any reason for

resisting a mod e of democratic jud gment-making that imposes reason at the

collective level. Is there a general argument for th inking that it is better, in

Cond orcetian term s, to collectivise reason rather than ind ividualising it? In

par ticular, to make the issue more specific, is there a general argu ment for going the

prem ise-driven rather than the conclusion-driven w ay? As the answer goes in this

case, so it shou ld go m ore generally.

I raise the question un der the assum ption that a conclusion in the conjun ctive

case will be supp orted if the premises are suppor ted and w ill be rejected otherwise.

This is the assum ption , in effect, that if the premises are p and q, then the

conclusion is just the conjun ction of those p remises. We need only consider the


25/36

25

conjun ctive case, since und er our interpretation the chance of an ind ividual or group

being right abou t a d isjun ction, p or q, is just th e chance of their being righ t about the

conjun ction, not(not p and not q).

The calculation of the relevant probabilities is a complex matter that WlodekRabinowicz and I add ress in an app endix, but th e overall picture suggested by those

calculations suggested, not definitively established is relatively clear. The

prem ise-driven ap proach d oes better, accord ing to that picture, provided we focus

on the case wh ere the individual or group is right about the conclusion through

being right about the premises: provided they are right about the conclusion for the

right reasons. But it is possible for an ind ividual or group to be right about w hether a

conjun ctive conclusion obtains, wh ile being wrong about some or a ll of the conjun ct-

premises: it is possible to be right abou t the conclusion for the w rong reasons.

Sup pose that it is not the case that p and q. I may be right in thinking that it is not the

case that p and q th rough th inking that not p and q, or not p and not q, when it is

rather the case that p and not q; or I may be right through th inking that p and not q,

or not p and not q, when it is rather that case that not p and q; or, finally, I may be

right through thinking that p and not q, or not p and q, when it is rather the case that

not p and not q. It turns out, surp risingly, that if it does not matter what the reasons

are for which an ind ividual or group is right abou t a conclusion, then the conclusion-driven approach does better in Cond orcetian term s: it offers a better chance of the

groups making the correct jud gment on th e conjunctive conclusion.

How should w e respond to this finding? It is not the most w elcome result

possible from the point of view of the argum ent in the last section; the most welcome

result wou ld have been that the p remise-dr iven ap proach is better either way. But it

is much m ore welcome than wou ld have been th e result that the conclusion-driven

app roach is better either w ay. And it is a result that, on reflection, proves to be quite

congenial.

When a person or comm unity makes a correct jud gment that p for the wrong

deliberative reasons, then w e deny th at they un derstand wh y it is the case that p, or

that th ey know that p . But the ideal of deliberative democracy, as that has been

articulated on all sides, is closely bound to the alleged p rospect of an increase of

understand ing and perhaps knowledge on the part of individuals in the comm unity,


26/36

26

and the grou p as a whole; there is no suggestion that it merely increases the

likelihood of serendip itous error. The ideal supp oses that in relying on deliberation

to guide them toward s a collective jud gment, people will be guided by right reasons:

that is, by reasons that are sound as well as supportive.

This line of thou ght suggests that w hen w e try to determine how satisfactory

the p remise-driven app roach is in Cond orcetian term s, the question w e should ask is

wh ether it does better than the conclusion-dr iven ap proach in p romoting the chance

of the groups getting the right result, assuming th at the result is to be reached on the

basis of right reasons. And the answ er to that qu estion, hap pily, does go the way of

the p remise-driven app roach. Not only does the app roach make contestability

possible and so answ er to repu blican concerns. Under the assum ption that

jud gments are to be reached via right r easons, which is common to all theorists of

deliberative dem ocracy, it also holds out the best prospect of enabling the group to

get at the truth .

I said that the p icture d efended in the app endix, on w hich w e have been

relying, is not definitively established there. What is worth not ing, however, is that

the less secure element in tha t picture is the suggestion that the conclusion-driven

app roach w ill do better when it does not matter whether the moving considerations

are the right reasons or not. The claim tha t the prem ise-driven procedure will do

better than the conclusion-driven w hen only right reasons are in play that is, the

claim on w hich the validity of the comments just m ade depends looks to be much

more firmly supp orted.

I conclude, to return to the general line, that the various argu men ts in the

literatu re for deliberative dem ocracy are consistent with th e republican w ay of

resolving the question that the discursive dilemma raises. The republican concern

with contestability gives pow erful reason for wanting democratic procedure to bedeliberative, where possible, in way that imposes the discipline of reason at the

collective level. And the other considera tions that are more generally invoked to

support deliberative democracy do not represent countervailing forces.3

Research School of Social Sciences,Australian National University,

Canberra, ACT 0200


27/36

27

Philip.Pettit@anu .edu.au .References

Benhabib, S. (1994). Deliberat ive rationality and m odels of democratic legitimacy.

Constellations 1: 26-52.

Black, D. (1958). The Theory of Committees and Elections. Cambridge, Cambridge

University Press.

Brennan, G. (1999). Collective Irrat ionality and Belief. Canber ra, Research School of

Social Sciences, Australian National University.

Brennan, G. and P. Pettit (1990). Unveiling the Vote. British Journal of Political

Science 20: 311-33.

Buchanan, J. and G. Tullock (1962). The Calculus of Consent. Ann Arbor, University

of Michigan Press.

Chap man, B. (1998). Law, Incomm ensurability, and Conceptually Sequenced

Argument. Un iversity of Pennsylvania Law Review 146: 1487-1582.

Chapman, B. (1998). More Easily Done than Said: Rules, Reason and Rational Social

Choice. Oxford Journal of Legal Studies 18: 293-329.

Cohen, J. (1986). An Epistemic Conception of Democracy. Ethics 97: 26-38.

Cohen , J. (1989). Deliberation and Democratic Legitimacy. The Good Polity. A.

Hamlin and P. Pettit: 17-34.

Coleman, J. and J. Ferejohn (1986). Democracy and Social Choice. Ethics 97: 6-25.

Cond orcet, M. d. (1976). Condorcet: Selected Writings. Indianop olis, Bobbs-Merr ill.

Dryzek, J. S. (1990). Discursive Dem ocracy. Cambridge, Cambridge University Press.

Estlund , D. (1993). Making Tru th Safe for Dem ocracy. The Idea of Democracy. D.

Copp , J. Hampton and J. E.Roemer. Cambr idge, Cambrid ge University Press:

71-100.

Estlund , D. (1993). Who's Afra id of Deliberative Dem ocracy? Texas Law Review

71: 1437-77.


28/36

28

Estlund , D. (1994). Opinion Leaders, Ind epend ence, and Cond orcet's Jury

Theorem . Theory and Decision 36: 131-62.

Estlund , D. (1997). Beyond Fairness and Deliberation: The Epistemic Dimension of

Democratic Auth ority. Deliberative Democracy: Essays on Reason andPolitics. J. Bohm an and W. Rehg. Cambridge, Mass., MIT Press.

Estlund , D. and J. Waldron (1989). Democratic Theory and the Pu blic Interest:

Condorcet and Rousseau Revisited. American Political Science Review 83:

1317-1328.

Good in, R. E. (1999). Democratic Deliberation Within. Canberra, Research School of

Social Sciences, Australian National University.

Grofman, B. N. and S. Feld (1988). Rousseau 's General Will: A Condorcetian

Perspective. American Political Science Review 82: 567-76.

Gutm ann , A. and D. Thom pson (1996). Democracy and Disagreement. Cambridge,

Mass, Harvard University Press.

Haberm as, J. (1984, 1989). A Theory of Communicative Action, Vols1 and 2.

Cambridge, Polity Press.

Kornh auser, L. A. (1992). Modelling Collegial Courts. I. Path-Depend ence.

Internat ional Review of Law and Econom ics 12: 169-85.

Kornh auser, L. A. (1992). Mod elling Collegial Courts. II. Legal Doctrine. Journal of

Law, Economics and Organiza tion 8: 441-70.

Kornh auser, L. A. and L. G. Sager (1986). Unpacking the Cour t. Yale Law Journal

82.

Kornh auser, L. A. and L. G. Sager (1993). The One and the Many: Adjud ication in

Collegial Courts. Ca liforn ia Law Review 81: 1-59.

List, C. and P. Pettit (2000). The Aggregation of Reason: An Im possibility Result.

Canberra, Australian National University.

Ow en, G., B. Grofman , et al. (1989). Proving a d istribu tion-free generalization of the

Cond orcet jury theorem . Math ematical Social Sciences 17: 1-16.


29/36

29

Pettit, P. (1997). Repu blicanism: A Theory of Freedom and Governm ent. Oxford,

Oxford University Press.

Pettit, P. (2000). Democracy, Electoral and Contestatory. Nomos.

Pettit, P. (2000). Groups with Mind s of their Own . Canberra, Australian National

University.

Skyrms, B. (1966). Choice and Chance: An Introd uction to Inductive Logic. Belmon t,

Cal., Dickenson .

Stearns, M. L. (1999). Shou ld Justices Ever Switch Votes? Miller v . Albrigh t in Social

Choice Perspective. Supreme Court Econom ic Review 7: 87-156.

Sun stein, C. (1999). One Case At A Time. Cambr idge, Mass, Harvard University

Press.

Sunstein, C. R. (1993a). The Partial Constitution. Cambridge, Mass., Harvard

University Press.

Young, I. M. (1990). Justice and the Politics of Difference. Princeton, Pr inceton

University Press.

App end ix . The jury theorem and th e discursive dilemm a

Philip Pettit and Wlodek RabinowiczBackground

Let u s assume that in a society, S, everyone has a greater than evens

chance, k, of being right in respect of each of two u nrelated propositions, an d

. This, to take the case of in illustration, means that for any mem ber, x, the

probability that x jud ges that , and , or that x jud ges that not , and not , is

greater than 1/ 2. If we use R for is right in respect of, i.e., as short for theunderlined p roposition in the p receding sentence, then w e can express the

assum ption by saying that for every x and y in S, P(Rx) = P(Ry) = k > 1/ 2; and

similarly for .

Let us assume, in addition, that th e chance of any individua ls being right

about one of the propositions is ind epend ent of the chance of any other


30/36

30

ind ividuals being right abou t it. Thu s for all x and y in S, P(Rx & Ry) =

P(Rx).P(Ry), and similarly for .

Und er these assum ptions, the Condorcet jury theorem gives rise to an

interesting qu estion. We know from the theorem that given the tw o assump tions,

the p robability of a majoritys being right abou t either proposition is greater than

k and that it ap proaches 1 as the size of the grou p increases. But w hat is the

probability of a m ajoritys being right about the conjunction & ? That

question divides into tw o, in the light of the d iscursive dilemma. First, what is

the p robability of a majoritys being right, if the group follows the p remise-

driven procedure? And second , what is the p robability of a majoritys being

right, if the grou p follows a conclusion-dr iven procedure?

We explore the answers to these questions in this append ix, entering

specific values for the size of the group and for the chance of an individuals

being right about a proposition. In pu rsuing this exploration w e add some

further assump tions, in order to make the questions tractable. We take it, not

only that and are un related p ropositions in respect of content, but also that

the competence of each ind ividual in respect of those prop ositions is un related.

Thus, the probability that the ind ividual correctly jud ges one proposition to be

true or false is independent both of the truth value of the other p roposition andof his jud gment of that p roposition (Independence). We assum e that each

individual is opinionated w ith respect to both and , i.e., the probability that

he judges () to be true or to be false the probability that h e jud ges that it

holds or jud ges that it does not hold equals 1 (Opinionation). We assum e also

that, for each of these propositions, its being tru e is equiprobable w ith its being

false (Equiprobability of Propositions). And we assume that the comp etence of each

individual is evenhanded as far as each p roposition and its negation are

concerned : that he jud ges it to be true and it is tru e is just as p robable as that he

judges it to be false and it is false (Evenhandedness).

A complexity: being right for the w rong reasons

These questions involve a comp lexity that is not at first obvious. If we ask

about someones or ind eed some group s chance of being right abou t a


31/36

31

conjunction like & then w e have to consider the chance of their being right in

each of four (as we assume, equiprobable) scenarios:

1. & 2. not & not . 3. & not 4. not & .

In each of those scenarios we have to d etermine th e chance of the p ersons being

right for the right reasons, i.e. through being right about each of the conjuncts; or

being right for the wrong reasons, i.e. throu gh being wrong abou t one or even

both of the conjun cts. A person can be right in scenario 1 only through being

right about both conjuncts; in 2 through being right abou t both or wrong about

one and right about the other; in 3 through being right about both, through being

wrong about both, or through being wrong about the first and right about the

second; and in 4 through being right about both, through being w rong about

both, or throu gh being right about the first and wrong about the second .

First question, re premise-driven procedure

Let us now consider the first qu estion. Specifically let u s consider it und er

the assump tion that we h ave a group , S, of just three members and that the

chance, k, of an ind ividuals being right is 2/ 3. Und er the premise-driven

procedure w e first have to d etermine the chance of the groups being right abou t

each of the p ropositions, and . The Cond orcet jury theorem, applied as

illustrated in th e text, tells us tha t this chance is 20/ 27. So wh at, then, is the

chance of the group s being right abou t the conjun ction & ?

The elements required for making th is calculation are presented in this

table, where J ind icates that th e proposition following is jud ged to be tru e by

the relevant subject: in this case, the majority and so the grou p.

Scenarios Ways of being right Chance of being right

1. an d J and J 20/ 27 x 20/ 27

Total: 400/ 729

2. not and not Jnot and Jnot 20/ 27 x 20/ 27

Jnot and J 20/ 27 x 7/ 27

J and Jnot 7/ 27 x 20/ 27


32/36

32

Total: 680/ 729

3. and not J and Jnot 20/ 27 x 20/ 27

Jnot and Jnot 7/ 27 x 20/ 27

Jnot and J 7/ 27 x 7/ 27

Total: 589/ 729

4. not an d Jnot and J 20/ 27 x 20/ 27


J and Jnot 7/ 27 x 7/ 27

Total: 589/ 729

The table displays the fact that it is hard er for the group to be righ t in

scenario 1 about th e conjun ction than in other cases; in those cases the group can

be right about the conjun ction while being wrong about one or both of the

conjun cts. The overall probability of the group s being right abou t the

conjun ction w ill equal the sum of the four totals, divided by 4, since the scenarios

are equiprobable. And that figure comes out as 1129/ 1458. This is the chance of

the groups being right about the conclusion, for right or wrong reasons, un der

the premise-driven procedu re.

But what is the chance of the grou ps being right for the right reasons? The

answ er is given by the sum of the first row totals for the four scenarios, d ivided

by 4. And the figure in th is case is 400/ 729: i.e. 800/ 1458 as against 1129/ 1458.

Second question, re conclusion-driven p rocedure

Now for the second question, which bears on the grou ps chance of being

right in the event of following a conclusion-driven procedure. Here the firstcalculation involves the random ind ividuals chance of being right about the

conjunction, given an ind ependent 2/ 3rds chance of being right about an y

conjun ct. The elements requ ired in this calculation a re given in the following

table:

Scenarios Ways of being right Chance of being right


33/36

33

1. an d J and J 2/ 3 x 2/ 3

Total: 4/ 9

2. not and not Jnot and Jnot 2/ 3 x 2/ 3

Jnot and J 2/ 3 x 1/ 3

J and Jnot 1/ 3 x 2/ 3

Total: 8/ 9

3. and not J and Jnot 2/ 3 x 2/ 3


Jnot and J 1/ 3 x 1/ 3

Total: 7/ 9

4. not an d Jnot and J 2/ 3 x 2/ 3


J and Jnot 1/ 3 x 1/ 3

Total: 7/ 9.

What is the chance of a rand om ind ividuals being right abou t the

conjun ction for any reasons, right or w rong? This is given by the sum of the four

totals, d ivided by 4, which is 13/ 18. And w hat in that case, using Condorcets

theorem , will be the chance of a majority being correct? This will be the su m of

the chances attaching to the p ossibilities that all three members are correct, that

the first two are correct and the third wrong, that the first and third are correct

and the second wrong, and that the second and third are correct and the first

wrong. And that w orks out at 1183/ 1458. This is somewhat greater than the

chance of the groups being right, no matter for what reasons, un der the prem ise-

driven ap proach: 1183/ 1458 as against 1129/ 1458.

What is the chance of the group s being right w hen the ind ividuals are right

for the right reasons? In order to get an ind ividuals chance of being righ t for the

right reasons about the conjunction, we need to go back to the calculation for the

rand om ind ividu al, add up the first row totals for each scenario, and divide by 4.


34/36

34

This gives us a figure of 4/ 9. Using the now familiar Condorcetian calculation,

we then find that the chance of a majority being right abou t the conclusion is

648/ 1458. This is considerably smaller than the chance of the grou ps being right

for the right reasons und er the prem ise-driven app roach: 648/ 1458 as against

800/ 1458.

Does this last result dep end crucially on the fact that the rand om

individuals chance of being right abou t the conjun ction for the r ight reasons is

less than a half, i.e. 4/ 9? We do not think so. If we do the corresponding

calculations where the ind ividuals chance of being righ t about either proposition

is taken to be, say, 8/ 9 and so with his or her chance of being right about the

conjun ction for the right reasons being 64/ 81 and greater than evens we still

find that the group s chance of being right for the right reasons is higher under

the premise-driven approach. It is 495,616/ 531,441 as against 471,040/ 531,441.

Conclusion

These calculations give u s solid if not absolutely certain grou nd s for the

claims made in the text. These are, first, that th e prem ise-driven p rocedure looks

likely to do considerably better than the conclusion-driven in securing the chance

of a majority being right for the righ t reasons about a conjunction; and second,

that the conclusion-driven procedure ap pears to do better than the prem ise-

driven one, though not by the same sort of margin, in securing the chance of a

majority being right for no matter w hat reasons, right or wrong, about a

conjun ction. This conclusion applies to d isjun ctions as well. Any disjunction p or

q can be expressed as a negated conjun ction: not(not p and n otq); and being

right abou t a conjunction is equivalent, un der ou r assumpt ions, to being right

about its negation.


35/36

35

1 This means that there wou ld be nothing to be gained by having electoral dem ocracy

imp ose the d iscipline of reason at the collective level. But it does not mean that w e

should abandon the ambition, comm on to m ost deliberative democrats, to make

electoral democracy and p olitics as deliberative as possible. There may be reasons of

a non-republican kind for seeking a deliberative dispensation, as we shall see in the

fourth section. But in any event the republican rationale introdu ced in the third

section itself suggests an argu ment in sup por t of electoral d iscussion and

deliberation. This is that such debate is likely to be imp ortant for a political

democracy in helping it to establish a currency of considerations that can be invoked

to sup por t or challenge decisions of government. It can serve as a filter such that

those considerations that go u nchallenged in the electoral forum have a good claim

to express comm on perceived concerns.

2 Kornhau ser and Sager (1993) suggest that w hile the straightforw ard premise-

driven procedure is generally more satisfactory more satisfactory from the

point of view, roughly, of contestability the court should resort to a higher-

level procedu re in any case, like the one illustrated , where there is any d oubt

about th is. It shou ld take a majority meta-vote on wh ich procedure to employ at

the lower level, und ertak ing to justify its d ecision (31; cf 33-36): tha t is,

undertaking to find agreed premises such that the m eta-vote can be rep resented

as itself dr iven by majority endorsement of those considerat ions in a prem ise-

driven w ay. Thu s the higher-level considerations might be, first, that going along

with the lower-level premise-driven decision in this instance would generate a

certain resu lt and , second , that the cour t ough t to avoid such a result. Such

considerations, if endorsed by a majority, might provide a higher-level, premise-

driven case for going along with a different resu lt from th at wh ich was

supp orted by a lower-level, premise-dr iven procedure.

3I am ind ebted to Geoff Brennan, who first m ade m e aware of the dilemm a. I was

helped by comments received w hen versions were presented at Lund University,

Nov 99; at a conference on Deliberating about Deliberative Democracy, University

of Texas, Austin , Feb 00; at the ann ual Analytic Legal Philosop hers Conferece in

New York, April 00; and at seminar presentations in Johns Hop kins University, New


36/36

36

York University and Princeton University. I owe a part icular d ebt to Wlodek

Rabinow itz for his helpful commen ts and I was also greatly aided by discussions

with a num ber of other p eople: David Estlund , Christian List, Victoria McGeer and

Fred Schick. Bruce Chapman and Lewis Kornh auser sent m e useful written

comm ents on the paper.

deliberative democracy and the discursive dilemma0

Documents