Published in Synthese 91 (1992), 285-318

(Note: This is a manuscript version. Please quote only the published version.)

GROUP BELIEFS

by
Raimo Tuomela
University of Helsinki


I CAN A GROUP HAVE BELIEFS?

1. The possibility of ascribing goals, beliefs, and actions to collectives relies on the idea that collectives can be taken to resemble persons. I shall here accept this idea, although the analogy does not go very far (partly because collectives are not capable of performing primary actions but only act via their members and representatives). Following common-sense examples, I will accept in this paper that both factual and normative beliefs can be ascribed (somewhat metaphorically) to groups, both formal and informal, structured and unstructured. Let me here present for later reference a list of examples indicating a broad spectrum of cases:
a) The Government believes that war against Iraq will begin soon.
b) The Catholic Church believes that miracles exist.
c) Texaco believes that children should be seen, not hurt.
d) The Communist Party of Ruritania believes that capitalist countries will soon perish (but no one of its members really believes so).
e) Our family believes that the schools in this country are inefficient.
f) The team believes that it will win today's game.
g) The Finns believe that sauna originated in Finland.
h) This mob believes that Smith is a traitor.
i) Europeans believe that face-to-face discussants should keep at least half a meter apart from each other.
I will below develop an account of group beliefs which will accommodate cases a)-i) and which, indeed, is meant to cover all possible cases of group belief. My main account will concern a group's beliefs, viz. cases in which a group as a whole can be taken to believe something, say p. These will be called (proper) group beliefs, and they will be argued to include at least a), b), c), d) (as an extreme case), e),f), and h). There is also another kind of social belief that can be called a group belief and this is beliefs characterizable as shared "we-beliefs". Cases g) and i) seem best analyzed in terms of the we-belief account (although g) might also and in addition qualify as a proper group belief). Indeed, in many cases a proper group belief that p and a shared we-belief that p will both exist. But before being able to see all this in detail, many other things must be discussed.
Social collectives (or groups in the general sense) in the sense chosen for this paper are those which according to common sense are taken to be capable of action in a member-binding sense. Accordingly, such typical kinds of social groups as task-groups, teams, and organizations are understood as something capable of making group-commitments (shared "we-intentions"), and thus having goals, which they are disposed to strive for by means of their actions, performed either by all (or at least many) of their members acting as members of the collective or by some members or other persons acting on behalf of the collective (cf. Tuomela, 1984, Chapter 8, and 1991; also cf. Gilbert, 1989, for a related view1). Classes of people such as red-haired women or something similar of course do not qualify as social collectives in this sense. But if red-haired women organize themselves into a collective, then, as far as conceptual considerations go, that collective can have the capacity to act.
I have elsewhere analyzed actions performed by groups (collectives) as well as group goals (see Tuomela, 1989, 1991). The thesis I defended there is very simple, its core content being this: A collective, say G, performs an action X intentionally in certain "right" social and normative circumstances if and only if there are some "operative" agents for the collective who jointly (unless a single agent only is involved) intentionally do something which in those socially and normatively right conditions brings about X (or, more precisely, the so-called "result" event or state logically presupposed by X). My analysis relies on the notions of an operative (versus non-operative) member and of the socially and normatively right circumstances. It seems plausible that if these notions are required for an adequate analysis of group action, they are also required for adequate analyses of group goals and beliefs. The notions of goal, belief, and action are linked in the case of a group to approximately the same degree as in the individual case. In the latter case their interconnection is well established; given that the person-analogy applies to groups (which I shall accept here in the case of these basic notions), these notions apply to groups as well. More broadly and loosely, a group's having a goal and a relevant means-belief related to it is or involves a disposition to act in the pursuit of the goal; and if action requires the mentioned notion of a socially and normatively right circumstance and the notion of an operative member, then so does a disposition to act, for otherwise group action would not typically be successful.
While the above argumentation perhaps is not very strong, let me next state a stronger argument. It relies on the fact that the notions of the socially and normatively right circumstances and the notion of an operative member are crucial in accounting for the functioning of formal and organized collectives such as corporations. In their case the statutes, by-laws and other relevant rules of the collective can be shown to connect goals (interests, purposes, and whatever subtypes of goals are at stake), beliefs (or views), and actions. To have successful group action we must also require that the right circumstances (as specified or presupposed in the rules) obtain; and group beliefs about those circumstances typically mediate between goals and actions, in analogy with the single-agent case. Indeed, in the case of typical formal collectives, certain position-holders are required by the constitutive rules of the collective to set goals and accept views for the collective. So the main argument here is that because formal collectives like corporations require the previously mentioned technical notions for an adequate account not only of their actions but also their goals and beliefs, we need these technical notions, suitably liberalized, in other cases, too. (In the case of mobs, for instance, all the members may be operative members, and we are dealing with a special case.)
Because of limitations of space and because I have elsewhere (Tuomela, 1989, 1991) discussed the notion of an operative member and the notion of the correct social and normative circumstances, I shall not here go into detail. Let me just give the following synoptic characterizations. The operative members in the cases of group actions, group goals, and group beliefs are those actors, goal-formers, and belief-formers by virtue of whom, respectively, actions, goals and beliefs are attributed to groups. The right social and normative circumstances are those in which the members of collectives act and form views in their positions (and in their right tasks within them). Positions can be regarded basically as collections of tasks (based on social rules) and social roles (based on proper social norms, e.g. conventions). The rules defining the tasks in question can be either formal (resembling laws and statutes) or informal (based on informal group-agreement). I will below speak as if the collectives and positions in them were characterized at least in part in terms of the (formal or informal) rules which are in force in them. However, as will be pointed out at the end of the paper, the present approach can handle group beliefs also in cases of groups without rules, as long as there is an "authority system" (viz. an explicit or implicit authority-incorporating "mechanism" or system) for creating joint decisions and commitments in the group. (See Tuomela, 1991, for a systematic discussion of the above issues, reflecting in part ideas of Hobbes and Rousseau.)
One of the key concepts of my analysis of group beliefs is the notion of accepting a proposition (or, if you prefer to avoid obscure notions, sentences or the like). Accepting a descriptive proposition (or sentence), say p, seems closely related to believing that p is true. It should be emphasized, however, that accepting is an action, something one can do at will, while what can be called "experiential" believing is a non-actional mental state that can be brought about causally (cf. perceptual beliefs). Therefore accepting and believing (in this experiential sense) are different notions. To be sure, it is typically the case that, in the case of a descriptive proposition p, a person believes (in the experiential sense) that p is true (or, given a disquotational account of truth, believes that p) if and only if he accepts p as true. But this is not always so. In some cases a person may, for instance, accept as true that he (or his body) is a probabilistically fluctuating bunch of hadrons and leptons without really believing it to be true in the experiential sense, let alone having that conviction. His acceptance (the action and the mental state created by acceptance) would then be "cognitive" acceptance in the sense that he would be willing to operate on the assumption in question viz. to concretely act on it and to use it as a premiss in his reasoning, and so on. In another case a person might, in contrast, experientially believe and be convinced that, for instance, his view (e.g. a scientific hypothesis) is true without (yet) being able to accept it as true. (See Cohen, 1989, for relevant discussion.2) Thus, while the aforementioned equivalence seems to hold in standard cases, it seems not to be invariably true - which is no wonder as belief as non-actional experiential mental state versus belief as a state based on acceptance (an action) are indeed different notions. I will below speak of experiential belief and belief in the sense of acceptance as true, both regarded as analytical ideal types. (And note that experiential belief also is truth-related.) I will argue that group beliefs and what I will call positional beliefs may involve acceptance in a sense not related to truth and thus may not involve acceptance of propositions as true nor (truth-related) experiential belief. The notion of accepting a proposition as true will be called "acceptance in the narrow sense". Actually the distinction between (narrow) acceptance belief and experiential belief is not central in this paper, in contrast to the distinction between ordinary truth-related belief versus "belief" not so related (wide acceptance must be at stake here).
It should be emphasized that my distinction between acceptance belief and experiential belief is analytical and attempts to capture (vaguely perhaps) the intellectual and experiential dimensions or aspects of beliefs. I do not mean to suggest that these aspects be strictly separable in actually existing personal beliefs. While accepting does not admit of degrees, surely acceptance beliefs can in addition have experiential features (related to a feeling that the believed state of affairs is real and, perhaps, to being convinced about it). The acceptance aspect of beliefs again relates to the rational and cognitive features of beliefs, e.g. to their use as premises in reasoning and argumentation and to their logical properties (such as logical closure under conjunction, with some qualifications perhaps).3
2. Let me start my discussion of the notion of group belief by commenting on the only analysis of it I am aware of. This is the so-called "joint acceptance" account of group belief by Gilbert (1987) (and basically the same analysis is given in her 1989 book). She states her analysis as follows:

A group G believes that p if and only if it is common knowledge in G that the individual members of G have openly expressed their willingness to let p stand as the view of G.
This analysis is partly circular. Gilbert improves it and arrives at the following final account:
i) A group G believes that p if and only if the members of G jointly accept that p.
ii) Members of a group G jointly accept that p if and only if it is common knowledge in G that the individual members of G have openly expressed a conditional commitment jointly to accept that p together with the other members of G.

I find the basic idea in the above account attractive, but there are some serious problems involved.4 Here is a list of these difficulties:
1) The requirement of conditional commitment seems mistaken in the following two senses. First, I take it that in Gilbert's analysis each agent must be taken to presuppose that the others are committed to accepting the proposition p. It is not a contingent condition for the agent's commitment to accept that the others are committed to accepting p. But the notion of conditional commitment - as we typically understand the notion - relies on the condition's in question being contingent and unsure. (See Tuomela and Miller, 1988, for relevant discussion.) What should instead be required in the joint acceptance model in the case of each agent is absolute commitment together with the presupposed belief that the others have committed themselves to the acceptance of p. Secondly, in Gilbert's analysis the analysandum in ii) is a notion of unconditional (joint) acceptance, whereas the analysans operates with conditional acceptance (or, more precisely, commitment or intention to accept). What is missing is an account of how conditional acceptance can result in unconditional acceptance here.
2) It seems too strong to require openly expressed conditional commitment in ii). Common knowledge about commitment jointly to accept p seems to suffice, independently of what each agent's ground for believing in the others' commitment is. (Openly expressed commitment seems not to be needed always in cases in which the agents are well familiar with how the others operate.)
3) Gilbert does not in her analysis explicitly specify the nature of p. She does not either clarify the notions of acceptance and joint acceptance. She claims that believing that p is true cannot be a general requirement in the case of joint acceptance. While this seems to be right - and also my account will have this consequence, it does not become clear whether she has in mind acceptance in the narrow sense or in the wide sense. Furthermore, it should be noted that in Gilbert's above analysis the notion of joint acceptance appears on the right-hand side of ii). Thus ii) is circular and fails to give an analysis of this notion. One surely can speak of acceptance in many senses. For instance, besides accpeting something as a group's view, one can accept a proposition because it is useful (has some pragmatic virtue or other) or because one wants to entertain the thought that it is true although one believes it is false; or one can accept it because doing so expresses the idea that something is desirable or ought to be done, and so on. It can, however, be pointed out that even if Gilbert's schema does not clarify joint acceptance, she says in the text that it relates to the parties being committed as a body to doing such things as saying that p and refraining from denying that p in public. This I find plausible as such, but one may ask how much of a noncircular analysis it gives. (My account below will assume that joint acceptance in this context involves joint agreement to accept a view.)
I shall below discuss the important philosophical question - which Gilbert neglects - whether collectives can believe only in the wide acceptance sense or also in the standard truth-related sense. This is an important problem: my view in this paper that groups can not at least in all cases have normal truth-related beliefs indicates that on this score (among many others) the analogy between groups and natural persons breaks down.
4) Even if it were otherwise unproblematic, the above joint acceptance account does not work without modifications in the case of structured groups having representatives. For instance, consider Gilbert's own example: the United States believes that the invasion of Afghanistan was an unconscionable act. Here a belief is ascribed to the whole nation even if perhaps only its government believes or the ministers jointly believe so and have accepted it. A joint acceptance by all (or even most) Americans is not involved (or at any rate need not be involved) here. Gilbert now says that "we presume that the citizens have endorsed the idea that they may be regarded as jointly accepting whatever propositions the government itself accepts." But this seems to be too strong a requirement - tacit acceptance or, better, an obligation for such tacit acceptance in the case of nonoperative members seems to suffice (cf. Chapters 4 and 5). Gilbert's analysis does not make systematic use of what we have called the operative-nonoperative distinction, and she does not analyze the crucial normative and social circumstances involved.
3. As said earlier, my account will rely on concepts related to social positions and tasks based on the rules and proper social norms in force in the collective. I will thus emphasize positional beliefs (or positional views), although this should not be taken to imply that such institutional and other positional constraints always are present. These are beliefs or views that a position-holder has qua a position-holder or has internalized and accepted as a basis of his performances of aforementioned kinds of social tasks. I will assume that the correct social and normative circumstances C will implicitly partially define positional beliefs. Accordingly, I will explicitly speak of positional beliefs as distinguished from mere personal beliefs. A position-holder can naturally have his own personal beliefs at the same time with his positional beliefs, some or even all of which may contradict his personal ones at least in the sense that he can personally believe something p without having the positional belief (relative to his position P in G - or, alternatively, the circumstances C) that p, and conversely. To take a fictitious example, we may think that the Flat Earth Society secretary has the positional rule-based belief that the earth is flat. Her having that belief is based on the general presupposition underlying the existence and characteristic or constitutive rules of that society, viz. the assumption that the earth is flat. But her personal belief might be different. She might doubt that presupposition and perhaps have no definite belief of the matter at all. Generally speaking, positional beliefs rely on the presuppositions embodied in the statutes, by-laws, and other constitutive rules as well as (possibly) on various informal rules and proper social norms characterizing a collective.
Now we come to the crucial question: Are positional beliefs (positional views) related to narrow or to wide acceptance? One might try to reason that they are related to narrow acceptance by arguing that the notion of truth is viewpoint-dependent and by further arguing that in the case of positional beliefs truth is relative to the viewpoint of the collective in question, viz. its characteristic presuppositions. But this approach, it seems, does not work. Even if we accept that truth is viewpoint-dependent, the notion of truth which a normal personal belief presupposes relies on a global point of view, one taking into account all relevant things and one that the believer cannot - at that moment - transcend. If a person believes that (it is true that) the earth is flat, or whatever, such a belief is for him absolute in the sense that he cannot knowingly have left out of consideration central assumptions he personally believes to be true. The notion of truth required here is truth-all-relevant-things-considered (by the agent).
In the case of the secretary of the Flat Earth Society, her positional belief does not rely on truth all-relevant-things-considered. She is well able to transcend the flat earth -viewpoint and question the presupposition that the earth is flat. And this observation seems true of positional beliefs in general. It might be possible to construct a robot which is the "Compleat Position-Holder" in the sense of possessing only the assumptions and knowledge embodied in the rules of the collective. Such a robot could have beliefs in the narrow acceptance sense (if it would otherwise satisfy the conditions required of believers). In the robot's case, positional beliefs (positional views) and personal beliefs (in the acceptance sense) would be identical. In contrast, for normal agents of the Homo sapiens kind, that would not be the case: personal beliefs would always (or typically) be more wide-ranging and would transcend the viewpoints presupposed by positional beliefs. The conclusion of this argument then is that a positional belief that p need not be related to the truth of p.5 It may often be connected only to acceptance in a wider sense. Accordingly, the beliefs of a collective need not be proper truth-related beliefs - at least as long as they are analyzed in terms of the positional beliefs of their members. But when the positions in a collective do not prevent the position-holders from accepting views in the broad all-things-considered sense a collective can have proper truth-related beliefs. Note here that it would perhaps be more appropriate to speak of a positional view than a positional belief and to emphasize that a person's accepting or having a positional view basically amounts to his being disposed to act (and his acting) in relevant ways when performing the social tasks belonging to his position. This acting will include also his reasoning and thinking relevant to his performing his positional tasks, although it must be remembered that the positional viewpoint can never be a global one in the sense referred to earlier. Accordingly, acceptance of p in the wide sense will be taken below to be acceptance as a basis of all "positional" thinking and acting in a relevant "local" sense (depending on the group in question).



II THE POSITIONAL ACCOUNT OF GROUP BELIEFS


1. I will now present my (preliminary) analysis of proper group beliefs, viz. beliefs involving the group as a whole, in view of the above discussion. This analysis - termed the positional account of group belief - is closely analogous to the analysis of group goals given in Section I. It basically relates group belief to view-acceptance or belief-acceptance by the members of the group; and I will accordingly speak of "positional believing" in their case. In this analysis p will be a statement or proposition, also to be called a view; but it is not required that p be a descriptive proposition (which allows, for instance, that G might believe that all murderers ought to be decapitated). The content or meaning of p will here be regarded as fixed. (In some contexts this may be a problematic assumption, as conceivably there might not be a prior interpretation independent of what the group is taken to believe.)
My (preliminary) analysis is as follows:

(BC ) G believes that p in the social and normative circumstances C if and only if in C, there are operative members A1,...,Am of G in respective positions P1,...,Pm such that
1) the agents A1,...,Am, when they are performing their social tasks in their positions P1,...,Pm and due to their exercising the relevant authority system of G, (intentionally) jointly accept p as the view of G, and because of this exercise of the authority system they ought to continue to accept and positionally believe it;
2) there is a mutual belief among the operative members A1,...,Am to the effect that 1);
3) because of 1), the (full-fledged and adequately informed) nonoperative members of G tend to tacitly accept - or at least ought to accept - p, as members of G;
4) there is a mutual belief in G to the effect that 3).

As argued above, social tasks here will have to include rule-tasks (also cf. the discussion later in this paper). Put in general terms, the notion of joint acceptance involves the joining of the participants' wills into a group will, and thus a joint commitment. Accordingly, acceptance must in clause 1) be taken to involve agreement-making. Moreover, agreement must always be intentional. Thus we are dealing with intentional acceptance, although such acceptance may not be fully conscious. (Let me note that the concept of tacit acceptance of clause 3) is a much weaker notion - see below.) The notion of joint acceptance of something as G's view in the sense of clause 1) can be spelled out in terms of joint positional beliefs as follows:
(*) Agents A1,...,Am jointly accept p as the view of G in the social and normative circumstances C, if and only if, in C, these agents (in their respective positions) jointly accept that p because of their relevant exercise of the authority system of G.

We are here (and in the analysis to follow) dealing with positional acceptance-belief - a mental state resulting from wide acceptance; and this is what accepting a belief means in (*). Joint acceptance here then comes to mean joint or shared acceptance-belief. As said, joint acceptance must be (at least to some degree) intentional here, and thus a mutual belief (not necessarily fully conscious) in the joint acceptance will exist. An acceptance-belief in this positional sense need not be a belief in the ordinary truth-related sense, but mutual belief can be regarded as a truth-related belief. We then get this as our final analysis:

(BC' ) G believes that p in the social and normative circumstances C if and only if in C, there are operative members A1,...,Am of G in respective positions P1,...,Pm such that 1') the agents A1,...,Am, when they are performing their social tasks in their positions P1,...,Pm and due to their exercising the relevant authority system of G, (intentionally) jointly accept p and because of this exercise of the authority system they ought to continue to accept and positionally believe it;
2') there is a mutual belief among the operative members A1,...,Am to the effect that 1');
3') because of 1'), the (full-fledged and adequately informed) nonoperative members of G tend to tacitly accept - or at least ought to accept - p, as members of G;
4') there is a mutual belief in G to the effect that 3').

Let me now make some explicatory remarks on (BC ) and, especially, (BC' ). In the above analyses group-beliefs are treated as socially backed acceptances, which need not be acceptances in the narrow sense. Recall our previous discussion and especially the argument that positional beliefs are concerned with local rather than global points of view and must therefore be beliefs in the sense of wide acceptance. While acceptance and belief need not necessarily covary in the case of non-positional beliefs, we have above - when explicating 1) of (BC ) by 1') of (BC' ) - connected positional beliefs and joint acceptances in terms of truth-equivalence on the grounds mentioned in the previous sentence. (Note that if what will be said below in defense of 1') is correct, (BC' ) is defensible in its own right irrespective of the truth of (*).)
Also note that in those cases where the circumstances C do not put any special restraints on the relevant factors that can be taken into consideration (relative to p) a positional belief reduces to a truth-related belief, viz. to an ordinary belief which involves narrow acceptance (and can of course involve experiential features). On the other hand, some groups at least sometimes have beliefs which are only beliefs in the wide sense of acceptance, because the circumstances C do not (except for the "empty case") allow for beliefs and views all things considered; cf. our examples a)-d).
(BC' ) allows for the possibility that in group G there are some special operative members who set the views and form the opinions of the group. In some groups all the members may be operative members, and then clauses 3') and 4') become vacuously satisfied; see below. It is instructive to think of a formal organization (one basically characterized by formal rules) which has a group-decision-making body for setting the goals and forming the views of the organization. My analysis has been modeled on this idea, but it is also meant to cover informal collectives, as will be seen.
According to clauses 1) and 1') the operative members jointly accept a view (opinion, belief) p for the collective in question and accordingly positionally believe that p. This can be said to reduce in part the macro-notion of a belief attributed to a group to a shared joint acceptance-belief had by some relevant members of the group. 2') involves the idea that there is intersubjective, mutual belief about this matter among them (normally shared awareness of 1') will suffice (cf. Tuomela, 1991, Chapter 1). Clause 2') is entailed by the notion of intentional joint action, and I shall not here argue for it specifically. As noted and as will be seen below, the mutual belief might not be (fully) conscious.
The basic idea underlying 1) and 1') is, generally put, the following. Group properties are "embodied" in group life at the level of individual group members and interactions between them. More specifically, (BC' ) assumes that proper group beliefs (which amount to accepted group views) are functions of group-members agreed-upon views. The authority system of the group creates joint intentions (group commitments) in the group to accept (and continue to accept) views, and these views become jointly accepted (by the operative members in the first place). All the members of the group accordingly ought to positionally believe that p if p has so been jointly accepted by the operative members, given that the members have knowledge about this agreement. The nonoperative members ought to accept (even if only tacitly) those views on the ground that the operative members have so accepted (on the basis of an appropriate exercise of the group's authority system for view formation).
Clauses 1) and 1') in effect say that the operative members ought to positionally accept and continue to accept (and acceptance-believe) what they have jointly accepted as the view of G, and this "ought" just is the objective obligation (or commitment) involved in, and due to, the relevant exercise of the authority system of G. This may be called the "normative" aspect of proper group belief. The factual aspect of proper group belief involves the relevant psychological features the members in fact have and this aspect is dealt with by all the clauses 1')-4'). Let me emphasize that 1') must be regarded as idealized if it is read to entail that all the members of G believe that p (in part due to their having the relevant subjective commitment, viz. due to their believing that they ought to accept and positionally believe that p). The ultimate penalty for not acknowledging and obeying the objective group-commitment when no stronger commitments exist is the loss of membership (in a literal or less literal sense). Let me stress that 1') involves under its "collective" reading that the operative members have as a body jointly accepted the belief that p when the group belief was created; and under this understanding of 1') some exceptions can be tolerated. But obviously a group cannot function properly unless its position-holders think and act in accordance with what the group has agreed upon; and of course the original joint acceptance of p must have involved a sufficient degree of factual participation for the joint agreement simply to have come about. We must allow that also after belief formation exceptions may occur - e.g. some operative members may not originally have participated at all or may have changed their mind as to p; and we still would like to say that the group continues to accept p as its view. I shall not discuss the synchronic situation here but simply accept some exceptions below, without trying to specify how many can be allowed. At the same time I continue to emphasize that the full blown situation is that in which all the operative members positionally believe that p (and obey the obligation involved).
It should be emphasized that in 1') we are dealing with what joint acceptance and the resulting state of acceptance-belief involves, not with how that joint acceptance was arrived at. In some groups there may be an institutionalized procedure (such as voting on the majority principle) for reaching a group result which involves group-authority in the sense that all members ought to comply with such a result (qua members of G, albeit not necessarily in the personal sense) even if the group view differed from all the participants' positional views (cf. cases of compromises between extreme positional views). But we are here dealing with the "post-decision-making" situation, and we do not really need to care about how the group decision came about. (Although there are voting procedures which in some cases do not respect the majority principle, this does not affect our ex post actu analysis . My analysis relies on the idea that the operative agents act in their proper positions, performing their social tasks (this is, in general, guaranteed by the circumstances C, except perhaps in the case of tasks based on direct agreement). They may also have other positions in G, but the notion of socially and normatively correct circumstances is supposed to distinguish between relevant and irrelevant positions (and tasks); e.g. in a business corporation the statutes may specify this.
Clause 3') speaks about the nonoperative members' tendency to conform to what the operative members have accepted as the views of G. This clause presupposes that the nonoperative members must be "adequately informed" and have the right beliefs about what the operative members jointly position-believe. If they do not have any beliefs or only incorrect beliefs, we should not speak of group beliefs in a categorical sense - but at best of a kind of condition-dependent group beliefs. Indeed, such condition-involving (but not in a proper sense conditional) group beliefs can be analyzed by means of the conjunction of 1'), 2'), and 4') of (BC' ) and the following weakened version of 3'):
3'') Because of 1'), the nonoperative members of G tend tacitly to accept - or at least ought so to accept - p, as members of G, if they believe the operative members jointly believe (when performing their social tasks in their positions) that p.
Let me emphasize that if the nonoperative members indeed do tacitly accept what the operative members have chosen as the view of G, this will be in part due to 1'). Thus it will involve their recognition and acceptance of the objective "oughts" created by the exercise of the authority system; briefly, it involves their subjective commitment to those objective oughts. These commitments concern the acceptance of whatever the operative members have agreed upon (in the sense of clause 1')) and thus by implication, given their right belief, of the proposition that p in particular.
Let us consider the acceptability of clause 3'). Basically it is grounded on group agreement. There is supposed to be an authority system at work here, and it results in the formation of a group will - to accept p as the view of G. The notion of an authority system in this case involves that the nonoperative members give up their will with respect to group goal formation and transfer that right to the operative members. When the operative members reach an agreement, it will be a group agreement. An agreement entails that what is agreed upon ought prima facie to be carried out. Clause 1') in conjunction with 2') is understood here to entail that the operative members accept the objective ought and are therefore (subjectively) committed to supporting p. What is more, the whole group is involved in the agreement, and so also the nonoperative members ought to contribute. This ought is an objective ought to contribute to whatever the operative members have agreed upon for the group. By implication there is an objective obligation for the members to contribute specifically towards the acceptance (and satisfaction) of p. If the nonoperative members are adequately informed then they have the right belief about what the operative members have chosen as G's view, and then they will be "subjectively" committed to the tacit acceptance of p, viz. they will believe that they ought to (tacitly) accept p. (Full-blown, "ordinary" acceptance of p is not required.)
Our official reading of (BC') then involves that the group belief in question is normatively both objective and subjective in the case of the operative as well as the nonoperative members. Are there other possibilities? The answer is yes, even if we have, for simplicity, formulated (BC') so that it best fits the aforementioned reading. There are in fact altogether five conceptually possible normative combinations. In addition to the standard case just referred to it is also possible to have a case otherwise similar to the first one but where the nonoperative members have not internalized the objective ought and thus are not subjectively committed to it. This case could come about. for instance, due to their having rather accepted some conflicting commitment. The third possibility, a rather peculiar one, would be like the second case but where also the operative members lack the subjective commitment. A fourth possibility would be the case like the third one but where only the operative members (but not the nonoperative ones) lack the subjective commitment. Finally, the fifth possible case is the one in which there is no objective commitment but where both the operative members and the nonoperative members have the subjective commitment to accept and continue to accept p. This fifth case could come about because of a mistake on the part of the agents concerning what they are doing when exercising the authority system in question. Here we could actually speak of a merely subjective group belief (in contrast the previous, objective cases): the agents thought they were intentionally agreeing to something which agreement, however, objectively did not take place (not with the content to accept p, at any rate). All the five mentioned cases involve either an objective or a subjective ought and all the members of the group are subject to at least one kind of ought. Thus the whole group is always involved. A proper discussion of these various possibilities must, however, be left for another occasion. I will below be explicitly concerned only with the first, standard case.6
(BC' ) is compatible with there being nonoperative members who even qua members of G disagree with the joint positional beliefs of the operative members. We can say that the disagreeing members' part or share mainly consists in that they at least "tacitly accept", or at least ought so to accept,the operative members' positional beliefs (accepted for G). Based on commonsense examples, I suggest that silent or tacit acceptance here is to be understood in a broad sense denying overt opposition which would make it difficult for G to act on the belief set by the operative members of G; and thus a nonoperative member is assumed not strongly overtly to oppose the views accepted by the operative members, not at least without being subject to criticism. Let us consider a simple example. Suppose that the operative members of a country (say, the members of its Parliament) accept a certain law for the country. It might be a law granting subsidies for the farmers. A citizen may disagree with the law and even demonstrate against it. But he is still prima facie obliged to obey the law and, for instance, pay the extra taxes imposed on him due to it. He must in this sense tacitly accept the law.
Obviously and trivially my notion of tacit acceptance excludes revolutionary activities, terrorism, or becoming an outsider in some similar sense. A strongly disagreeing member of course always has open the possibility of deserting the group (at least in thought, if this is not possible overtly); and this act is compatible with the analysans of (BC' ) in the technical sense of making its requirements inapplicable to such a person. (Needless to say, the more cohesive a group is, the more inclined are its members to accept its goals and views.) It is, anyway, worth keeping in mind that clause 3') can theoretically be satisfied even if none of the nonoperative members actually tacitly accepts p even when they believe that p has been jointly accepted by the operative members. In such a theoretically possible case they are still deserving of criticism (in the sense of the ought-norm applying to them) if they are in overt opposition in the above sense, but they don't deserve criticism for minor disagreement.
Leaving clauses 1') and 3'), let us now consider the mutual belief clauses 2') and 4'), which stress the importance of mutual, intersubjective belief that p in the group. Generally speaking, we can say that group action is at stake here: the group will created by the exercise of the authority system obviously concerns action, for will is here regarded as an intention, and intention must (ultimately) concern action. The action at stake here is acceptance (of a view). Now, as intentional action is involved, mutual belief must exist. Thus clause 2') can be regarded as a conceptual consequence of the fact that intentional joint agreement to accept p is involved. In normal cases we can say that without such mutual belief it might come as a surprise to the operative members that other operative members share the view that p and would be odd.7 We can also say that if the operative members did not have the mutual belief in question, they would be having their view solipsistically, so to speak, and then they cannot be said to be having their views as members of the group (but only as private persons). Furthermore, when the collective G acts, its actions will be based on views mutually believed to be views of the group at least by the operative members, otherwise disorderly group action is likely to result.
Clause 4') can be supported in part by reference to the idea that the group as a whole is and must be involved (both in a normative and in a factual sense). It must be normatively involved because of the exercise of its authority system and factually because of the psychological, attitudinal connections between the members. Mutual belief is, so to speak, a kind of social glue which glues the members together into a whole on the factual level.
Let me emphasize that we are in our analysis dealing with positional beliefs. It is even possible that no one in the collective would have the personal belief that p (and on this issue my account agrees with Gilbert's analysis).8 Let us consider our example d), and accordingly suppose that in a collective G, the Communist Party of Ruritania, no one personally believes what the members of the collective acting in their positions accept and thus position-believe, e.g. that capitalistic countries will soon perish. According to my analysis, G can and will then have the belief in question. But intuitively it may seem quite odd that a collective can believe something that none of its members personally believes. We will label this a spurious group belief. I suggest that we recognize the difference between cases where the collective belief has more or less extensive backing from the above extreme case, and we may apply the predicate 'genuine' to group beliefs having the kind of personal backing meant above.
In accordance with what was just said, there can be three kinds of group beliefs: genuine , spurious , and "mixed " ones (falling in between the genuine and the spurious cases). Depending on case, it is often relevant to our assessment of a group's belief (acceptance in the wide sense of a proposition) to know what the members of the group believe about the matter in question and what kind of belief distribution results from that, but - in contradistinction to the positional account - this does not suffice to determine what the group as a whole (or in its totality) believes or accepts as its view nor does it qualify to give "belief-premises" for the practical reasonings which groups engage in and which lead to group action. The distribution of personal beliefs (whether experiential beliefs or acceptance beliefs) defines another dimension in addition to the view-acceptance dimension. This other dimension, so to speak, gives the doxastic climate of the group. How much personal support in this sense should required of genuine group beliefs seems to depend on the case at hand and cannot be captured by a general schema such as our (BC' ).
Spurious beliefs in the above sense tend to be dysfunctional and affect the long-run stability of the group. The analysis (BC' ) gives a kind of snapshot view of a certain point of time. If I were to deal with the dynamic case I would also describe how to reconcile positional beliefs and personal beliefs - at least in the case of functionally central beliefs such as those embodied in the constitution of a state. Here obviously the means of opinion-moulding typically are used. Then, ideology and religion as well as, more generally, any kind of propaganda can be used; power (both social and physical) can be exercised by the governing individuals to make nonoperatives accept (viz. not oppose) the proposition p. We can say that the firmness of a collective's belief is centrally connected to the extent to which the operative members positional beliefs are genuine, viz. personally accepted (and not only accepted in the positional sense). Let me note that in the case of genuine group beliefs, at least the operative members of the group (or most of them - cf. clause 1') above) can be taken to believe in the "we-mode," so that sentences of the form "We believe that p" (viz. I believe that p and believe also that it is a mutual belief among us that p) can be truly attributed to each (or at least most of them. This kind of shared we-belief can be called a group belief in the we-belief sense.
2. Let us now test (BC' ) in a certain type of case not taken up earlier. We consider what happens in the (rather rare) case of a collective in which there are no rules in force and hence no rule-based tasks and positions involving such tasks. Social collectives can be usefully classified by means of the (formal or informal) rules and the proper social norms (covert social norms like conventions) that are in force in them. In extreme cases (such as some mobs and some temporary task-groups) there may, conceivably, not be any rules or covert social norms in force. Such groups may lack rule-based tasks - but it can be claimed that when they do, they do not have beliefs either (in an intuitive sense or in the sense of (BC' )). Let me explain.
The basic philosophical point here is the following. Rules for a group can be created only by the help of some "authority system" - a system for creating shared we-intentions, indeed group wills. Often group discussion and negotiation is involved here. The system may involve a set of group members, viz. the operative members, representing the others. What underlies the existence of such authority and the involved possibility of representation is that the group members be capable of - to speak in Rousseau's terms - giving up their will (with respect to some issues) and thus their "original" authority and transferring it to the group-authority system, which pools the wills into a group will (Gilbert, 1989, p. 189, also speaks of "pools of wills" in a related sense). The authority system need not be explicit but can operate implicitly or covertly as often happens when we speak of joint acceptance (merely). Without a "group-decision making mechanism" of this kind the making of rules is not possible. Indeed, the exercise of authority in this sense can be regarded as both necessary and sufficient for group-decisions, group-agreements, etc. to give rules (e.g. of the form "Everyone ought to do his part of X") as result.
What, then, follows for our (BC' ) from these considerations? The existence of rule-tasks is guaranteed by the existence and operation of an authority system in a group. (Also externally imposed rule-tasks can exist, of course, but they must be accepted as rule-tasks by the agents.) Groups without rule-tasks seem very rare, and, indeed, the thesis can be defended that all properly functioning social groups have a-systems that they have exercised (these groups are group-commitment systems called "f-groups"), and this entails that they have rule-tasks of their own (see Tuomela, 1991, Chapter 4, and cf. note 1). Given this, shared or joint belief as used in (BC' ) can be understood to involve the existence and functioning of at least an implicit authority system. Thus (BC' ) comes to cover all (and only) the cases where the group functions as a whole, so to speak. In those cases there will be positions and social tasks (including rule-tasks), as needed by (BC' ). Let me note that even in the simple case where a group has only one group belief and no other specific group-attitude, over and above its group will concerning the acceptance of that group-belief, the joint commitment to act in accordance with that belief will create rule-tasks. Those rule-tasks created simultaneously with the group belief qualify for the analysans of (BC' ).
All the other cases of groups do not involve beliefs that a group has as a whole, and in such cases the best we can do is to investigate the doxastic climate of the group at the level of its individual members and aggregate the personal beliefs. These group beliefs will, more precisely, be aggregated or shared we-beliefs in the sense defined earlier - see Section III for a fuller discussion. Thus an informal dyad can believe, for instance, that it will rain tomorrow, if both members believe that and if it is a mutual belief among them that they so believe and if they jointly commit themselves to that belief as expressed, for instance, in their acceptance and readiness to use that premise in their practical reasoning and planning related to their relevant joint actions. If they do not have that readiness, there is no proper group belief but only a group belief in the weaker sense of shared we-belief - see Section III. (Note that in the case of informal groups like the above where no operative-nonoperative distinction exists, clauses 3') and 4') of (BC' ) will be vacuously satisfied.)
Let us next consider our examples a)-i) presented in the beginning of this chapter. Can our approach handle them and show that they will be either proper group beliefs or we-beliefs (or involve both of these)? Examples a) and b) are factual beliefs related to organized groups, and so is d). Example c) is highly similar to them but its content is normative. In a)-d) these clearly are operative members who determine the group beliefs. In a) all the members of group might be involved as operative ones, in cases b)-d) there is a nonempty set of nonoperative members. (If in a) instead of 'the Government' we would have 'USA' the members of the Government might be operative ones while citizens of USA, qua citizens rather qua private persons, would be non-operative members.) Now, if indeed there is an authority system at work in cases a)-d) - and it surely is justified to think so - there seems to be no problem in taking the clauses of (BC' ) to be fulfilled. Thus a)-d) can be regarded as proper group beliefs, at least if there is a considerable amount of shared non-positional we-beliefs (with the same respective content) backing them. However, as said, case d) is peculiar in the sense that it is spurious or empty in precisely this sense. I shall not argue further about d) but, following (BC' ), simply accept it as a group belief in a most extreme, spurious sense. (For relevant remarks on distinct cases of generally held beliefs, cf. Gilbert, 1987, pp. 187-189, and 1989, Chapter 5.)
Examples e)-h) might be understood as cases where all the members are operative ones. Given this assumption, we can ask whether or not an authority system creating group-commitments is at work. But independently of how that question is answered, it can be argued that we should anyhow require of group beliefs here that relevant personal we-beliefs (with the same respective content) are widely shared in the group. Completely empty or spurious cases are still harder to digest here than in the case of examples of the kind d). (I do not regard this as a counterexample to (BC' ), for arguably there are no special positional constraints here, and therefore 1') and 2') will be made true by shared we-belief - and thus we get a personally backed group belief; see result #) of Section III.) In cases e) and f) it is plausible to think that an authority system is at work, but cases g) and h) are perhaps best thought of as not involving an exercise of an authority system (although the groups might have one). Case i) may be a group belief in the we-belief sense (but not in the sense of (BC' ) if in addition the people involved become aware of its content or at least acquire a subconscious mutual belief about it. Otherwise it will not be a group belief. This also indicates that mutual belief at least in the sense of (potential) awareness of the first-order belief must be required in the case of group belief of any kind. (Tacit group beliefs can be discussed on the basis of subconscious beliefs which, however, can be made fully conscious in suitable circumstances.) A related example excluded by the above considerations is this: Suppose a football team is exercising and that all players notice a beautiful girl watching them. The players might not have noticed that the other players also have noticed her. Then there would simply be a shared first-order belief without mutual awareness of it, and thus no group belief even in the we-belief sense would exist.9



III WE-BELIEFS AND GROUP BELIEFS - A COMPARISON OF TWO APPROACHES


1. In this section I will discuss in some more detail the approach to group beliefs in terms of shared we-beliefs (based on ordinary personal beliefs) and compare it with the analysis of proper group beliefs in the sense of (BC' ). The we-belief approach that also may be called the "statistical" or "aggregative" approach. I will here discuss a certain version of this aggregative approach that may be called the simple or naive we-belief approach (cf. Miller, 1990, for a richer and more sophisticated version, and see also Gilbert, 1987, pp. 187-188 on aggregative approaches). It has much intuitive appeal but leads to some conclusions that conflict with what the positional account gives. My discussion of the simple we-belief approach - and recall that I have already accepted it to account for those group beliefs which fall beyond the scope of (BC' ) - will be concise and conducted in part with use of logical symbolism, which also will serve further to clarify my positional account. My main thesis is that group beliefs require a combination of the positional and the aggregative approach. Proper group beliefs are best handled in terms of the positional approach conjoined with the we-belief analysis to account for the degree of "genuineness" of proper group beliefs. Other group beliefs are to be handled in terms of the aggregative approach, which, however, is not capable alone of dealing with proper group beliefs.
We can take a "we-belief" to be of the form "We believe that p," attributed to an agent a, to have the form Ba(p & Bmp), where B is an (unanalyzed) belief operator, and Bm stands for mutual belief in the collective in question (cf. Miller, 1990). Analyzed thus, a we-belief has two components that can be read off its "content radical" p & Bmp: the agent a believes that p, and he also believes that it is a mutual belief in the group that p. The first requirement seems obvious. But the second is perhaps not obvious. Yet, surely the agent a must be taken to believe that at least most of the other agents in the group believe that p, at least if the group is unstructured. Some kind of common belief must thus be believed by the agents to obtain. But why require mutual belief? Why does not belief in shared belief suffice? We need an argument here. One is that if the agents believed only that shared belief in p exists but no shared awareness of this shared belief, these agents might be surprised to learn about their belief being commonly shared in the group (and would thus not be prepared to take the sharedness aspect into account when planning their own action). This argument gives us a second order belief, and - when needed - the consideration can be iterated to generate higher-order mutual beliefs (see Tuomela, 1984, Chapter 7, and 1991, Chapter 1).
Consider now a belief that a group has, viz. a group belief. In the simple we-belief account (the only kind of aggregative account I will consider) a group belief will amount to shared we-belief in the group. To discuss this idea in more precise terms, I shall use some symbolism. Let again 'B' be a sentential operator standing for beliefs. A group belief that p is expressible by "We believe that p" - here applicable to all the members of the group G; it will be symbolized by BGp. BGp is true according to the aggregative account if and only if the members share the we-belief that p, viz. if and only if
M) for every a in G, a believes that p and a also believes that Bmp.
As a matter of fact we can speak of degrees of group belief - depending on how wide spread in the group is the belief that p. (That seems possible, in a somewhat contrived sense, also in the positional account, for the operative members might not all believe that p, but only a certain, sufficiently large percentage of them.)
We recall from our earlier discussion that there are, so to speak, two dimensions involved in group-beliefs: there is the level of accepting something as a group's view, and that involves an action (and an acceptance-state resulting from it). The other dimension is what the group members believe to be true about the subject matter in question. This belief can be an experiential belief or an acceptance belief (or a mixture of them - if these types of beliefs are regarded as a kind of ideal types). M) emphasizes personal belief which typically is experiential, but the more important feature of M) is that the group as a whole does not really figure in it: the statistical distribution of we-beliefs is the only element in group belief.
2. How does M) relate to the analysis (BC' )? To answer this, we first need to know what a positional belief in the sense used in (BC' ) amounts to in the present terminology. We may take a positional belief that p of a position-holder A in collective G to be the belief that (it is true that) p is believed (and has been accepted) by A in the social-normative circumstances C in question (assumed to specify the right circumstances for the task-performances related to that position). Below I shall ignore C and assume that it is satisfied (or does not pose any restrictions).
Let me state (BC' ) in symbolic terms (and in a somewhat simplified way). Let G+ be the set of operative members of G, and G- its nonoperative ones. Clause 1') of (BC' ) amounts to this (if we - for simplicity - omit the part about the "oughts"):
1*) For every a in G+, a believes (accepts) that p,
and 2') becomes:
2*) it is a mutual belief in G+ that for every a in G+, a believes that p.
Let us denote belief by B and mutual belief by Bm and assume that B means at least acceptance in the wide sense (but may in special cases also involve acceptance in the narrow sense as well as experiential belief).
We can rewrite 1*) and 2*) as
1**) For every a in G+, Bap
2**) BmG+(for every a in G+, Bap).
If we denote tacit acceptance by A and, for the sake of simplicity, assume acceptance to take place for the rights reason and also ignore the tendency-part, 3') and 4') of (BC' ) respectively become:
3*) For every a in G-, Aap;
4*) BmG(for every a in G-, Aap).

Now we can say that the conjunction 1')&2')&3')&4') (of the clauses of (BC' )) is true if and only if 1**)&2**)&3*)&4*) is true (given the above simplifications). We recall that tacit acceptance in the sense of the operator A is weaker than belief in the sense of B: belief in the latter sense always is acceptance in the sense of A but not necessarily conversely. Acceptance is an action leading to the state of acceptance-belief. Tacit acceptance is weaker than explicit acceptance: tacit acceptance only means (roughly) something like "not opposing overtly" (viz. refusing to obey the various rules brought about by the operative agents; disagreement in a weaker sense is, however, allowed).10
Our analysandum above is BGp, viz. the group believes that p; and this belief need not amount to more than acceptance in the wide sense, although it may occasionally be (or involve) acceptance in the narrow sense. Of course, 'B' here must be interpreted to covary with what 'B' in the analysans involves: a group belief has no independent content over and above what its basis - group members' beliefs and tacit acceptances - can give us. Let me note that one might in fact want to say that tacit acceptance does not really belong to the group belief (the notion itself) but has to do with the group's having the belief in a group involving an authority system for group-view formation.
As usual, let us analyze mutual belief in terms of the "Everybody believes" -operator Be, applying to the fullfledged and adequately informed group members: Bm = Bep&BeBep&..., with a finite number of operator-iterations (see Tuomela, 1991, Chapter 1 for discussion). We get this:
#) Suppose that G is a social group (with a functioning authority system) in which all the members are operative members with respect to belief formation and also suppose that the social and normative circumstances C are "empty", viz. do not impose constraints differing from normal condition-assumptions. Then, ignoring again the "oughts" involved in 1**), M) is true, viz. G believes that p in the aggregate sense if and only if the conjunction 1**)&2**) is true, viz. if and only if, according to (BC' ), the group G believes that p.
The proof of #) goes as follows: Let us define Bmp = Bep&BeBep. (For simplicity, I assume that only two iterations of the Be-operator are needed for mutual belief, but the generalization of the proof to the case of mutual belief defined analogously in terms of any finite number of iterations is obvious.) Now M) becomes Bep & BeBmp, which amounts to Bep & BeBep & BeBeBep. On the other hand, 1**) amounts to Bep, and 2**) is BmBep. The latter gives the conjuncts BeBep and BeBeBep. We have the same conjuncts in M) and 1**)&2**) and hence equivalence.
It is worth emphasizing that group beliefs in the sense of shared we-beliefs are at bottom mutual beliefs in the group. We can see from the above equivalence proof that even if the positional account and the aggregate account are equivalent under the conditions of #) and for any explicate of mutual belief stated by means of, say, k iterations of Be, we are at the same time dealing with group beliefs in the sense of mutual beliefs with k+1 iterations. (Thus, for instance, M) with k=2, gives a mutual belief in the content p with the value k=3, and so on.) This shows that, apart from the smallest possible difference in the number of iterations, group beliefs in the sense of mutual beliefs really are shared we-beliefs and vice versa. It follows that those social norms which are based on mutual beliefs are group beliefs at least in the sense of shared we-beliefs. If, in addition, a relevant exercise of an authority system is involved, creating "oughts" and upgrading mutual belief into mutual agreement in the rule-sense, we also get proper group beliefs here.
I have shown that the simple we-belief (or aggregative) approach and the positional account are in de facto agreement in the case of groups in which all the members are operative ones, given that the social and normative circumstances C are empty (or can otherwise be ignored). Of course, this is only one side of the coin: first, we have above ignored the normative dimension (the oughts); secondly, we have also ignored the right normative and social circumstances C; and, thirdly, even on the factual side there is only a partial overlap between the two approaches (see below). The group as a whole will be normatively involved (because of the commitment-creating exercise of its authority system) even when less than all the members are operative ones (with respect to belief); and in this situation the aggregative approach is inferior (if extended to cover those cases, which I am of course not suggesting in my final "combined" account). However, it is possible in that approach to say that those cases have to do with group belief only insofar as we-beliefs are concerned; the rest is concerned with operative members' acceptances of views for the group, and that has nothing to do with group belief (viz. the belief that p, although it may be connected to the group belief that the group has accepted that p). I wish to emphasize that the dispute is then only verbal (at least as far this issue is concerned): the positional approach has (proper) group beliefs in the sense of (BC' ) and group beliefs in the sense of shared we-beliefs (merely), while the aggregative approach correspondingly speaks of group decisions and acceptances (etc.) versus we-beliefs in these cases. I do not want to legislate about language-use here. The second problem and dispute concerns the role of the social and normative circumstances C. In the aggregative approach these play no role whatsoever in characterizing group belief, whereas in the positional account they are of course central. Actually this matter seems to boil down to the previous point, for surely there is no doubt that in groups (especially in formal ones such as organizations) views have to be formed under the right social and normative circumstances. It is only that the aggregation theorist would again say that this matter has nothing to do with group belief.
To illustrate yet another type of conflict, consider an example of a dyad G consisting of the members A and B, and make the following three assumptions
1) Bap;
2) Bb-p;
3) A and B jointly accept that p.
In the positional account these three statements can be jointly satisfied, even if it must be admitted that an extreme case is at stake (a more typical one would be where 3), contrary to 1) and 2), is regarded as relativized to some special, non-empty social and normative circumstances C). Now, from 3) we can get, if an authority system indeed is at work, in the case of the positional account
4) BGp,
while the aggregative account gives (from 1) and 2))
5) -BGp.
(Furthermore, we may note that both approaches entail
6) BG(that they jointly accept that p as the group stand).)
4) and 5) contradict each other and show a conflict between the two approaches (and that a division of labor between them is needed). This is not really a dispute about the operative versus nonoperative members nor about the circumstances C but, it seems, about the nature of beliefs in question. I suggest we deal with this case as follows: In 1) and 2) we are dealing with beliefs prior to the use of the group-authority system, while 3) concerns the situation after decision-making (or after the use of a group-belief formation mechanism). However, A and B may maintain their personal beliefs 1) and 2) after 3) has come about - according to the positional account. According to the aggregative account, however, 3) is simply irrelevant to group belief (here: the belief that p). It is worth noting that the positional account can also handle cases of conflict on the level of acceptance - the group's belief that not 3), for instance, amounts to the (narrow) acceptance by A and B that they do not jointly accept that p.
I would like to emphasize in connection with the above example that it is not a logical consequence of the positional theory of group beliefs that i) 4) is true, given 1), 2), and 3) nor that ii) 1), 2), and 3) can in any situation be jointly consistent. But i) and ii) are both compatible with the positional theory (while i) is not compatible with the aggregative account).
Furthermore, anomalous cases of group beliefs such as false consciousness can also be formulated and analyzed within this approach, although I shall not here discuss them. Let me finally point out that the above example of conflicting beliefs in a dyad reminds one of game-theoretic structures involving some conflict, such as the Battle of the Sexes. It will be left to future research to consider the formation of joint views systematically within game-theoretic structures.
3. As a partial summary of the consequences of the accepted positional account of proper group beliefs and of the we-belief account in the case of other group beliefs I finally wish to stress the following features of my combined account:
1) Groups (e.g. formal and organized groups) can have group beliefs such as the belief that something p in the wide sense of accepting that p.
2) A group's having a proper group belief that p is compatible in the extreme (or spurious) case with even all the members' personally not believing that p (and indeed with their believing that not-p). However, it is relevant to the genuineness and firmness of the group's belief that p to what extent its members personally believe, and, indeed, we-believe that p.
3) Any member of a group is prima facie obliged to accept (in the discussed wide sense) that p qua a member of the group, if the group has a proper group belief that p; but such a member is under no obligation personally to believe that p.
4) Gallup-poll investigations of proper group belief can in some cases be impossible on theoretical grounds. Of course, statistical investigations of relative frequencies of people sharing a we-belief are at least theoretically possible.
5) A group can have a proper group belief (and, for that matter, a goal) only if it incorporates either an explicit or an implicit group-authority system on which joint acceptance of views relies.
6) Groups can have beliefs in the sense of shared we-beliefs. This is the case when no group-authority system is at work to create group-commitments involving the group as a whole. Indeed, social collectives which are not social groups in the sense of group-commitment systems can obviously have shared we-beliefs (but not proper group beliefs). As said, a proper group belief is the more genuine the more widely the group members corresponding personal we-belief is shared.
The final upshot of the above discussion is that a viable account of all group beliefs is the one combining both the positional approach, applying to the normative case, and the simple we-belief approach, applying to the non-normative case.11 The positional approach accounts for proper group beliefs (group beliefs involving the group as a whole) aided (but not replaced in this context) by the we-belief approach. The we-belief approach again accounts for group beliefs that are not in the mentioned sense proper group beliefs, and do not involve an exercise of an authority system, and it should be accepted for such cases (which fall beyond the scope of the positional account (BC' )).*


* I wish to thank Kaarlo Miller and Philip Pettit for critical comments.


NOTES

1) In Tuomela (1984) two notions of a social group are employed, and both rely on shared group-intentions (and hence group-commitments, intentions being commitments). When discussing the notion of a group that social action concepts presuppose (see esp. pp. 118, 144, 263), the requirement is presented that a social group must involve shared we-intentions (or group-intentions). Another, "normative" notion of a social group is defined in Chapter 8, see esp. p. 264. Also this notion relies essentially on the notion of we-intention and requires the group-members to have relevant we-intentions. (An expanded and improved account is to be found in Tuomela, 1991, Chapter 4.)
Gilbert (1989) takes groups to be "plural subjects", expressible by the pronoun 'we' under appropriate conditions. Her final analysis on p. 201 says this: "'We' refers to a set of people each of whom shares with oneself in some action, belief, attitude, or other such attribute". (Here 'shares' is to be understood in a strong sense, such that "those who share in an action, and so on, constitute the plural subject of the action".) Another way of characterizing the proper use of 'we' in this context that Gilbert gives as her final analysis on p. 199 is this: "A person X's full-blooded use of 'we' in 'Shall we do A?' with respect to [some other persons] Y, Z, and himself, is appropriate if and only if it expresses his recognition of the fact that he and the others are jointly ready to share in doing A in relevant circumstances." The analysans of this latter notion of 'we' appears to define a stronger notion of 'we' than the first one, for sharing in an attitude (e.g. wish or imagination) might not amount to readiness to do anything at all. Another, and bigger problem is the fact that surely one can ask the question 'Shall we do A?', for some A, in the case of any arbitrary collections of people. This cannot be right. Judging from Gilbert's discussion surrounding this analysis, she probably means to be analyzing the use of 'we' in the context of joint commitment to action, rather. I shall not attempt to explore the details of Gilbert's conception of a group-constituting joint commitment here. Suffice it to say that - even if Gilbert does not herself point it out - there is at least a surface resemblance between her account of groups and my account in terms of shared we-intentions, insofar as each of us invokes the concept of commitment.

2) Cohen (1989) treats acceptance and belief as completely separate notions. However, I will speak of acceptance belief (or belief in the acceptance sense) and experiential belief. This accords with the way standard English dictionaries characterize belief. Thus, for instance the Oxford Advanced Learner's Dictionary of Current English gives two main meanings for the word 'belief': 1) the feeling that something is real and true; trust; confidence; and 2) something accepted as true or real. I take it that 1) refers to the experiential aspect of belief while 2) is concerned with the acceptance aspect (propositions accepted as true).
3) We have to deal with the following components of belief. First, a belief can analytically be a (more or less) pure acceptance belief or a (more or less) pure experiential belief or some mixture of these. Secondly, we can speak of proper beliefs (standing or - also - occurrent) versus dispositions to acquire beliefs (experiential or acceptance beliefs).
Audi (1982) discusses the following principle:
For S to have a non-occurrent belief that p, at t, is for him to be such that, at t, a) he can entertain p, and b) should he entertain it he would spontaneously and sincerely assent to it.
This account undermines the distinction between i) dispositionally (in the sense of non-occurrently) believing that p, and ii) being disposed to come to believe that p "instantaneously" upon considering it. In some cases we must be prepared to live with ii) instead of i) (cf. Audi's 1982 discussion).

4) In her 1989 book Gilbert gives the following analysis of group belief:

A group G believes that p if and only if the members of G jointly accept that p.

This she explicates as follows:

The members of G jointly accept that p if and only if it is common knowledge in G that the members of G individually have intentionally and openly* expressed their willingness jointly to accept that p with the other members of G.
As the mentioned book came into my hands only after I had written this chapter I shall not here comment on the above analysis except for saying that the above criticisms with the partial exception of the first one (viz. 1)) still apply to this slightly modified analysis.

5) We may try to analyze a positional belief that p of a position-holder A in collective G as the belief that (it is true that) p has been accepted by A in the social-normative circumstances in question (assumed to specify the right circumstances for the task-performances related to that position). Then the positional belief would be related to the truth of its whole "content-radical", even if it would not be related to the truth of p. I shall not here explore this possibility further.

6) Our classification of the five kinds of conceptually possible "normative" group beliefs can be presented more conspicuously as follows. Both in the case of operative and nonoperative members we may speak of objective and subjective oughts (O = objective ought exists, S= subjective ought exists):

OS O-S -OS -O-S
Operative 1 2 3 not possible
Nonoperative 4 5 6 not possible

The conceptually possible cases are: 1&4, 1&5, 2&4, 2&5, 3&6. In other words, there are five possible kinds of proper group belief:
Operative Nonoperative
1) OS OS 2) OS O-S
3) O-S OS
4) O-S O-S
5) -OS -OS
Of these 1) and 2) are clearly possible objective cases, while 5) is a merely subjective case. Cases 3) and 4) are rather peculiar but still conceptually possible.

7) The "surprise argument" for the presence of mutual belief (which argument also Gilbert, 1989, uses in a related kind of context) applies not only to standard conscious mutual belief but also to "subconscious" or "prereflective" mutual belief. As for the latter, consider our example i) from the beginning of the chapter. To argue for the presence of mutual belief in this case suppose that my discussant at a cocktail party comes very close to me, say at the distance of twenty centimeters. I become surprised and feel uncomfortable. Why does mu fellow discussant behave so? Is he from a different culture (even if I originally thought him to be a European)? Upon reflection I find myself having the belief that he should not stand so close and that he should believe so, too, and also believe that I believe so. In this way I come to see that among the members of my group there is the mutual belief (however prereflective and dispositional it may be) the discussants should keep at least half a meter apart from each other. (In fact, it suffices here that there only be a disposition to acquire the mutual belief in question in baffling situations like the above.)

8) A person can have the positional belief that p (viz. he might for some reason or other accept that p, where acceptance need not be narrow acceptance) while he may fail to believe that p personally - he may not have any belief of the matter. Furthermore, in some cases he might personally believe that not-p while having the positional belief that p. This is possible because a positional belief is not a proper truth-related belief but an acceptance belief, which falls short of taking into account all things considered relevant by the agent. Thus we are not forced to assume in this kind of cases that the agent has a schizophrenic mind (or anything of the kind), that he must pretend or fake, when acting in his position, that p is true or acceptable, or, as a final alternative, that he cannot be aware of his conflicting beliefs or of their conflict.
Note that in some cases a positional belief can have a content different from any corresponding or related personal belief. For instance, the secretary of the Flat Earth Society might have only the positional belief that his collective has adopted the view that the earth is flat while her relevant personal belief would be that the earth is not flat. There is no conflict here.

9) Some passages of Sections I and II of this paper have been drawn from Tuomela (1990b).

10) Let me point out that we can similarly symbolize and formalize condition-involving group beliefs, the condition being that the non-operative members believe that the operative members jointly believe that p:
A group believes that p relative to the nonoperative members' beliefs about the operative members' belief in the circumstances C if and only if in C, 1'), 2'), 4'), and 3'') the nonoperative members of G tend tacitly to accept - or at least ought so to accept - p, as members of G, if they believe the operative members jointly believe (when performing their rule-based tasks in their positions) that p.
In our symbolic terms we now get the following analysis of condition-involving group beliefs (ignoring the oughts involved as well as the circumstances C):
A group G believes that p relative to the nonoperative members' beliefs about the operative members' belief that p if and only if
1**) For every a in G+, Bap.
2**) BmG+Bep.
3**) For every a in G-, Aap, given that BaBmG+p;
4**) BmG(for every a in G-, Aap, given that BaBmG+p).
Here Be symbolizes "Everybody believes ...".

11) As a summary the following final comparison with Gilbert's 1987 and 1989 analysis can be made. Gilbert does not use such notions as my positional belief, wide acceptance, or shared we-belief in her account, nor does seem to use "functional counterparts" of these notions. Her notion of joint acceptance is different from mine. As to the consequences of my combined account presented above, 2) and 3) clearly are features of Gilbert's account (and stressed by her), and so is the first part of 4). But 1) and 6) are not consequences of her approach, and neither does she seem to be committed to 5) (or anything resembling it within her account). Thus my final combined theory of group belief is very much different from Gilbert's account, even if some of the consequences of these accounts coincide, as pointed out.



REFERENCES

Audi, R., 1982, 'Believing and Affirming', Mind XCI , 115-120
Cohen, L. J., 1989, 'Belief and Acceptance', Mind XCVIII , 367-389
Gilbert, M., 1987, 'Modelling Collective Belief', Synthese 73 , 185-204
Gilbert, M., 1989, On Social Facts , Routledge, London and New York
Miller, K., 1990, 'Shared We-beliefs as Group Beliefs', manuscript
Tuomela, R., 1989, 'Actions by Collectives', Philosophical Perspectives 3 , 471-496
Tuomela, R., 1990a, 'Intentional Single and Joint Action', forthcoming in Philosophical Studies
Tuomela, R., 1990b, 'Can collectives have beliefs'? In Haaparanta, L., Kusch, M., and Niiniluoto, I. (eds.), Language, Knowledge, and Intentionality: Perspectives on the Philosophy of Jaakko Hintikka , Acta Philosophica Fennica 48
Tuomela, R. 1991, The Importance of Us: A Philosophical Study of Basic Social Notions , book manuscript
Tuomela, R. and Miller, K., 1988, 'We-Intentions', Philosophical Studies 53 , 367-389