WE-INTENTIONS REVISITED

RAIMO TUOMELA

WE-INTENTIONS REVISITED

ABSTRACT. This paper gives an up-to-date account of we-intentions and responds to some critics of the author’s earlier work on the topic in question. While the main lines of the new account are basically the same as before, the present account considerably adds to the earlier work. For one thing, it shows how we-intentions and joint intentions can arise in terms of the so-called Bulletin Board View of joint intention acquisition, which relies heavily on some under-lying mutually accepted conceptual and situational presuppositions but does not require agreement making or joint intention to form a joint intention. The model yields categorical, unconditional intentions to participate in the content of the we-intention and joint intention (viz. shared we-intention upon analysis).

The content of a we-intention can be, but need not be a joint action. Thus a participant alone cannot settle and control the content of the intention. Instead the participants jointly settle the content and control the satisfaction of the intention. These and some other features distinguish we-intentions from “action intentions”, viz. intentions that an agent can alone settle and satisfy. The paper discusses we-intentions (and other “aim-we-intentions”) from this perspective and it also defends the author’s earlier account against a charge of vicious circularity that has been directed against it.

I. INTRODUCTION

You and I may share the plan to carry a heavy table jointly upstairs and realize this plan. In this case we both can be said to have the

joint intentionjointly to carry the table upstairs: the content of the intention here involves our performing something together and the pronoun ‘we’ of course refers to us, viz. you and me together. In my earlier work I have often taken joint intentions to be expressible by means of locutions like “We will do X”, where the word ‘will’ is used conatively (rather than predictively, in the future tense) and X is a joint action type (cf. Tuomela, 1984, 1995, 2000a, b; Tuomela and Miller, 1988). However, as joint intentions can also have other contents, I will in this paper speak of the jointly intending agents’

jointly see to it that they jointly build a house, that one of them builds it, or that some outsiders are hired to do the job, and so on. I have chosen jstit as my umbrella term as it covers many kinds of activities – e.g. jointly performing actions in a direct or in an indirect sense, jointly bringing about states, jointly maintaining states, and so on (cf. Sandu and Tuomela, 1996; Belnap et al., 2001). Notice however, that in the context of joint intention the actions in question can be of various kinds, as just mentioned, and they can be non-equivalent (e.g. jstit does not entail direct performance nor is the converse true). I take jstit to be a necessarily intentional notion. This fits well with its appearance in the content of joint intention as one cannot non-intentionally satisfy an intention.

For some agents jointly to perform an action there must of course logically be an opportunity for them to do it. Thus they cannot open a window if it is already open. Let us call the state of the world where the window is open the result state of the action of opening the window. Seeing to it that the window is open expresses inten-tional control over the state of the window’s being open and requires success (viz., the agents have not seen to it that the window is open unless it is open).

The following cases of intentional activity fall under an agent’s seeing to it that the window is open: The agent opens the window (by his own direct actions), if it is closed; he keeps it open, if some other agent or something else tries to close it; the agent gets some other agent to open the window, if it is closed; he refrains from preventing another agent from opening the window (see Sandu and Tuomela, 1996; also cf. Belnap et al., 2001, for stit and jstit.)

The content of a joint intention has, as it were, two parts. In the case of single-agent intention I take the intention to have the form “A intends, by his actions, to perform X” or “A intends, by his actions, to see to it that X”. Analogously with this, we have – corresponding to the second alternative – in the case of joint agency (here the dyadic case) “A and B jointly intend by their actions to perform joint action X” or “A and B jointly intend to see to it jointly that X” (or, from their perspective, “We will perform X together”). Accordingly, I suggest that we take the joint intention now to be about jstiting something. This could be also called the first part of the content of the joint intention. What is to be jstited constitutes the second,

“variable” part of the content, and this part need not be performed as a joint action. The first part is the same in all joint intentions. It indicates that the joint intention is oriented towards joint action. We can add that jstiting involves that each participant of joint intention is in principle actionally involved: he has a share or part in the participants’ jstiting that X. I accordingly claim that the structure of the joint intention here can be expressed by the following in the dyadic case: Agents A and B jointly intend to see to it jointly that X. Here X can be the participants’ joint action, somebody else’s action, some other agents’ joint action, a state in the world (like that a house is painted or the window is open). Consider the special but central case in which X is a joint action to be performed by A and B. Then my formula becomes, using jstit for joint action allover: A and B jointly intend to see to it jointly that they jstit X, which due to the “collapsing” property of jstit amounts simply to “A and B jointly intend to see to it jointly that X”.

All intentions are necessarily related to one’s own actions. This applies also to joint intentions. In the single-agent case, an agent may intend to see to it that his car is fixed. This intention has as its satisfaction condition that the agent by his own actions sees to it that his car becomes fixed – e.g. he can get a mechanic to fix the car or fix it himself. Similarly, in the case of joint intention with the jstit content the participants have to see to it jointly by their actions that the intended state or event comes about. If the intention concerns the direct performance of an action (e.g. when an agent intends to open the window) the agent must himself bring about the satisfaction solely by his own action. This kind of intention I will call (direct) “action intention”. A minimal rationality condition for an action intention, at least a prior intention, is that the agent must at least lack the belief that it is impossible for him to perform the action. Assuming that at least a prior or future-directed intention involves commitment to the content of the intention, we can ground the previous claim by saying that if he would not so believe it would be pointless for him to commit himself to his task.

An action intention contrasts with an “aim-intention”. In the latter case it is not required that the agent believes that he with some likelihood can alone bring about or see to it that the action or its result event comes about. The kind of aim-intention that will

concern us in the present paper is we-intention. A we-intention is a participant’s “slice” of their joint intention, so to speak. Or the other way round, it can technically be said that a joint intention consists of the participants’ we-intentions about the existence of which the participants have mutual belief. Even if we assume that a joint inten-tion is (ontologically) composed of the agents we-inteninten-tions about which there is mutual knowledge (or belief), these we-intentions are different from ordinary action intentions not only in being aim-intentions but also in that they conceptually depend on the joint intention in question.

A we-intention is a special kind of aim-intention involving that the agent we-intends to bring about a state jointly with the other participants or we-intends to perform an action jointly with the others or, to use my general formulation, to see to it jointly with the others that a certain state or event comes about. Considering the joint action case where the agents jointly intend to perform a joint action together, the central condition of satisfaction of the we-intention is that the we-intending agent should intend to participate in the joint action in question. That is, he should intend by his own action, his part or share, to contribute to the joint action. Thus the agent’s having the we-intention to perform a joint action entails his partici-pation intention, which is an action intention in my terminology. (This matter will be discussed in detail in Section VI.)

An obvious rationality constraint on we-intention is that an agent cannot we-intend unless he believes not only that he can perform his part of their joint action X, but also that he together with his fellow participants can perform X jointly (can jstit X with them) at least with some nonzero probability. The jointly intending agents must believe that the “jstit opportunities” for an intentional jstiting of X are (or will be) there at least with some probability. Yet another property of a we-intention is that in each participant’s view it must be mutually believed by the participants that the presuppositions for the (intentional) jstiting of X hold or will hold with some probability. The formation of a joint intention (and hence we-intention, a personal “slice” of the joint intention) requires that the participants jointly make up their minds to jstit something, thus exercising joint control over the possible courses of action and settling for a partic-ular content of jstiting. The formation of a joint intention (or plan)

is based on their various personal and, especially, joint desires and mutual and other beliefs. In this sense a joint intention can be said to “summarize” or reflect the motivation underlying joint action. Of course, this final motivation underlying the joint inten-tion need not be anything like an aggregainten-tion of private motivainten-tions but may instead be a compromise based on discussion, negotiation, or bargaining. In contrast, joint desires and wants do not simi-larly involve making up one’s mind and do not lead to intentional jstiting, to attempts to function rationally in a coordinated way, so as to fulfil the already formed plan; and desires and wants do not require success beliefs as their normal (or “normal-rational”) accompaniments. Joint intentions (and hence shared we-intentions) entail collective (or, here equivalently, joint) commitments to action, and this joint commitment also includes that the participants are socially committed to each other to perform their parts of the jstiting.

The notions of stiting and jstiting clearly cover more than direct action performance, and this is one reason for employing these notions. Another reason is that they serve to make the structure of joint intention more perspicuous, as seen. Thus they help us to see that joint intentions typically involve not one but two kinds of activity. There is the activity in which the participants see to it that the (second part of the) content comes about and there is the very activity that makes the content obtain. Thus the participants might hire some other agents to build a house. Here the first kind of activity is hiring the agents and monitoring their work; and building the house (performed by the hired agents) is the second kind of activity. Having made these points, I will, however, mostly in this paper speak as if the content of joint intention were a joint action (directly) performed by the jointly intending agents. This is because speaking in terms of jstit becomes rather clumsy.

Section II, The Bulletin Board View of Intention Formation, of this paper discusses the conceptual and “structural” aspects of joint intention formation. While it in part draws on published material, the ideas of this section are extended and deepened later in Section III, We-Intentions and Joint Intentions Analyzed. This section gives a summary account of my theory of joint intention and adds some new aspects to the account. Section IV discusses in more detail the

epistemic and normative bonds between jointly intending partici-pants. Taken together II–IV give an up-to-date theory of joint intention and we-intention. Another central task of this paper is to answer some criticisms directed against my account by Seumas Miller and John Searle. These criticisms relate to the issues of what one really can intend and whether my original (viz. 1984 and 1988) account of we-intention is viciously circular. These prob-lems are assessed, respectively, in Section V, Collective Ends and We-Intentions, and Section VI, On the Alleged Circularity of the Concept of We-Intention. The concluding section VII summarizes the main achievements of the paper.

II. JOINT INTENTION FORMATION

In joint-intention formation each participating agent accepts to participate in the participants’ seeing to it jointly that some state or event X obtains. Concentrating on the central case in which X is a joint action, the agents jointly intend, as a group, to see to it that they perform X jointly. Here each participant accordingly is assumed to intend with a we-perspective together with the others. What does this involve? The agents jointly intend as a group to further the content of the joint intention that they have accepted as the group’s intended goal (broadly understood). I have elsewhere used the term “we-mode”, contrasting with the private or “I-mode”, to describe the present kind of thinking as a group member or thinking with the we-perspective and have offered the following analysis of a we-mode intention (Tuomela, 2002b, p. 30):

(WM) Agents A1,. . ., Am forming a group, g, share the

inten-tion to satisfy a content p (e.g. in our example p = X is jointly performed by the participants) in thewe-modeif and only if p is collectively accepted by them qua group members as the content of their collective intention and they are collectively committed to satisfying p for g. Here functioning as a group member entails for our example that the participants function so as to satisfy their shared intention to perform X together (and in more general cases function to further the group’s constitutive or main interests, goals, beliefs, and stand-ards). Collective commitment in the joint intention case need not

be stronger than what joint intention conceptually entails: When our agents jointly intend X (e.g. a joint action) they must collec-tively bind themselves to X and what its satisfaction requires. This I call the instrumental sense of collective commitment. This sense is intention-relative and, strictly speaking, non-normative. (In this sense my non-normative account sides with Bratman and Miller against Margaret Gilbert; see Bratman (1999, p. 125ff.), Miller (1995, p. 64), and Gilbert (1990, p. 6f.).)

In all, the we-mode in the case of joint intention amounts to saying that the participants must have collectively accepted “We together will do X” (or one of its variants) for their group, and they must have collectively committed themselves to doing X. Here “We together will do X” applies to each participant, and in the case of a single participant it expresses his we-intention. An agent’s we-intention then is his “slice” or part of the agents’ joint inten-tion, and conversely a joint intention can, upon analysis, be said to consist of the participants’ mutually known we-intentions. The collective acceptance of an intention as the group’s intention entails the satisfaction of the so-called Collectivity Condition. Applied to satisfaction, the Collectivity Condition says, roughly, that neces-sarily, if the joint intention (goal) is (semantically) satisfied for one of the participants, then it is satisfied for all participants. (For a more detailed recent discussion of the we-mode versus the I-mode or individual mode, see Tuomela (2002b, c); note that I-intentions, viz. personal intentions, can be either in the we-mode or in the I-mode.) In the joint intention to perform a joint action X, it is precisely the content of the intention that is shared,viz., the content of doing X jointly is shared. Each agent tokens this content, and because a necessarily act-relational intention is involved this amounts to his intention to perform his part or share of X. The basic argument for assuming that each participant must intend to perform his part of the joint action is that the joint intention can only be satisfied if each participant performs his part – for only then will the intentional joint action satisfying the joint intention come about. The part perfor-mance must be intentional, of course, and thus based on the agent’s intention.

In the general case, each agent can be taken to accept “We together will jointly see to it that X” (or its equivalent) and “I

will participate in, or contribute to, our jointly seeing to it that X”, while in the case of directly performable joint action the (“vari-able”) content of a we-intention can accordingly be taken to be “to perform X together”, entailing a participation intention for each participant. A we-intention is not by itself an “action intention” but an “aim-intention” involving that the agent intentionally aims at X and is “aim-committed” to X, while his action commitment is to performing his part of X. The agent’s intention to perform his

partof the joint action accordingly is a proper action intention, thus something the agent believes he can, at least with some probability, satisfy by his own action (given, of course, that the others perform their parts).

In this section I will consider the presuppositions of joint inten-tion and the central conceptual elements involved in joint inteninten-tion formation. I will focus on plan-based joint intentions which express

jointly intending as a groupand which arepublicin a group context. A central subclass of such joint intentions is formed by joint inten-tions based on the participants’ (explicit or implicit) agreement to act jointly. The making of an agreement in the full sense (viz., accepting a jointly obligating plan) is a joint action which is neces-sarily intentional. The point about emphasizing this kind of case is obviously that it is conceptually central and also common in actual social life (see Tuomela (1995, Chapter 2; 2002a), on which I will draw below).

What does this kind of plan-based joint intention presuppose? Firstly, it must obviously be required of the participants that they understand – at least in some rudimentary sense – that a joint action in some sense, however weak, is being proposed. The joint action must be taken to include a “slot” for each participant’s intention. In general, all the relevant generic action concepts need to be possessed to a relevant extent by the participants – a kind of “hermeneutic circle” is at play. Thus the notion of joint action opportunity needs to be available. Secondly, there is much other background knowl-edge, most of it culture-dependent, that is presupposed. Thirdly, and this is most significant, there is situation-specific information that must be presupposed. If the performance of a joint action, X, in a situation, S, is at stake, the concept of X must be possessed by the potential participants, and they must also understand what S

involves concerning the performance of X. It is also required that each participant believes that the participants mutually believe that the joint action opportunities for X hold in S (cf. the third clause in my analysis (WI) of we-intentions below in Section III). Some direct or indirect communication (or signaling) between the participants is needed for the reason that the participants are autonomous agents who, nevertheless, must make up their minds depending on what the others are thinking and doing. More concretely, communication is required for them rationally to arrive at unconditional intentions (we-intentions as well as intentions to perform one’s part of the joint action). The indirect communication may be previously “codi-fied” and may relate to certain specified types of situations (cf. “in situation S we always form a joint plan of a certain kind and act together”).

We can view the joint intention formation in intuitive terms from the group’s angle and say in functional terms that we want to have unified group action as a result of a group’s intention being properly satisfied. This involves that the group members’ actions must be suitably bound together and coordinated with each other. Part of this will have to take place at the level of the group’s plan of action, assuming that we are dealing with intentional action. The bond here is due to the group’s intention to act and its ensuing commitment to action, which makes the members collectively or jointly committed to the action. In cases of jointly intending as a group and thus being collectively committed as group members, the participants are also socially committed to each other to participate.

Viewing the matter from the “jointness” level, viz. on the level of the group members, jointly intending as a group amounts to jointly intending to realize a joint plan. Intuitively, the participants must be suitably bound together for proper collective action purporting to realize the shared plan to come about (and accordingly for their actingas a group). As seen, this activity in general requires public exchange of information between them if it is to lead to mutually known (and not only mutually believed) unconditional participation intentions.

Another philosophical reason for publicity in a group context is that such central social notions as the speech acts of agree-ment making, promising, commanding, and informing – all relevant

to joint-intention formation – are in their core sense not only language-dependent but public.

The view to be developed below takes all this into account. It presupposes that the participants understand in a colloquial sense what acting together and a plan (or an agreement) to act together are. I will below analyze the conceptually central elements in the formation of a full-blown joint intention to act together and do it in terms of a metaphor, the “Bulletin Board” metaphor. The resulting Bulletin Board View (BBV) bases joint intentions on a publicly shared plan of joint action and thereby emphasizes the epistemic publicity (the public availability of relevant information) of full-blown joint intention, as will be seen. While publicity is central in this view, in principle one can also formulate a similar view without the publicity requirement (see below).

Suppose that one of us comes up with the idea of cleaning a park. This is the proposed joint action content. That person may publicly communicate this to other group members. We may conceptualize and illustrate the present situation in terms of the followingBulletin Board Viewof joint intention formation. The initiating member’s or organizer’s proposal (or, more generally, plan for joint action) can be thought to be written on a public bulletin board: “Members of group g will clean the park next Saturday. Those who will partici-pate, please sign up here.” Here ‘will’ in the latter sentence is taken to express intention and not only prediction. Supposing that the ensuing communicative signaling of acceptance to participate (under the presupposition that sufficiently many others participate) results in a wide uptake and “whole-hearted” acceptance (signing) of this proposal, then – given (communication-based) mutual knowl-edge about this – there will be an adequate plan involving a joint intention to clean the park. The participants’ whole-hearted accept-ance is assumed to entail that the participants form the intention to participate in joint action. As seen, this is a two-faced intention, so to speak. There is, firstly, each signed-up participant’s we-intention to clean the park, and, secondly, his intention to carry out his part or share of the cleaning (qua his share of it). Furthermore, the partici-pants – because of having expressed their personal participation intentions – have jointly exercised control over what to do together

and made up their minds to clean the park. Thus their joint intention to clean the park has come into existence.

In a slightly stronger case the participants not only accept a shared plan of joint action but in effect make an explicit or implicit agreement (made up of interdependent promises) to perform it (cf. the discussion in Gilbert, 1993; Bach, 1995). Here it holds, on conceptual grounds, that making an agreement in this sense gives each participant a reason for action, viz., his promise. Furthermore, promising also gives a reason for the other participants to norma-tively expect that the other participants indeed will participate. Thus, we can say that here a participant has the right to expect that the others will perform their parts and is also obliged to respect their analogous rights. In this sense they are normatively socially committed.

The central thing about BBV is of course its general conceptual content, although I have in part used concrete and partly meta-phorical language in stating the view. From a conceptual and theoretical point of view, the present model of joint intention forma-tion involves the following elements. First, the content of the joint intention must be brought to the participants’ attention. I call this the topic problem. There may be an initiator who proposes the topic, or the participants may arrive at the topic by means of their negotiation or joint decision making. They might thus consider their preferences for the different action alternatives and arrive at a joint decision by a suitable decision rule, e.g. the majority vote. In this kind of case the participants of course, so to speak, sign the bulletin board proposal only after the topic has been decided upon. In other cases, the topic may be suggested to them by their shared history or background knowledge in conjunction with some relevant contextual informa-tion. For instance, the participants may share the standing want to keep the park clean and when they learn that a garbage collector will arrive the next morning they may gather that the park cleaning is the thing to do tonight. In BBV this element can be indicated by the appearance on the bulletin board of the description of the topic (here park cleaning).

The set of potential participants will be the members of a group, g, and this must be knowable to the potential participants. The actual participants – or at least a suitable subset of them sufficient to get

the joint action initiated and under way – will have to be publicly named or indicated. The central element again is the public avail-ability of the information about the intention to participate; this aspect is also relevant concerning newcomers and persons who have to change plans for some reason. The participants will pick up that information and this will lead them to believe that those signed up will participate. What is more, they will also be able to acquire mutual knowledge (or minimally mutual true belief) about this, for they will come to know that the others know that those persons will participate; and this can be iterated if needed.

The publicity requirement in BBV is a kind of public communi-cation requirement in the sense discussed earlier. It is a particular contingent feature of BBV that the information gathering and delivery is centralized so that, e.g., pairwise communication is not needed. However, this is a practical feature that is not conceptually essential and can easily be changed. But publicity in group context is still philosophically central in that it creates a quasi-objective realm,viz., a realm which is objective for the participants, and which is more prone to lead to actual objective knowledge than weaker views (as the participation intentions are “objectively out there” as stated on the bulletin board). There is thus a kind of group-relative objectivity both ontically and epistemically involved here. Further-more, as compared with less public methods, in the case of large groups, new participants, and participants that have changed their intentions, etc., knowledge can then better be gathered and checked. In our metaphor, there may be information written on the board and there is also information in a special box beneath the board called “Presuppositions and Background Knowledge”. Typically only situational information is written on the board, and the rest,

viz., general background assumptions and maybe some obvious kind of situational information is available in the presuppositions box. Somewhere there should also be information about whether the participants only are forming a shared plan for joint action based on publicly expressed intentions or are also making a full-blown agreement (in terms of interdependent promises) to act jointly and in this “thick” normative sense accept a joint plan (see Section IV).

The present approach has several virtues. Firstly, it gives categor-icaljoint intentions (without the problems concerned with

decondi-tionalization or with “change of view” – see Tuomela, 2002a). BBV is not concerned with proper conditional intentions at all (although extendable to deal with them as well). The belief that sufficiently many participate can be regarded as a presupposition rather than a (contingent) condition, and this is of course a presupposition binding all the participants (a collectivity condition is at play here). A participant thus categorically commits himself when signing up, although he may retract his commitment if he comes to believe that a relevant presupposition is not satisfied. Secondly, there is no need for a prior joint intention to form a joint intention, as mere personal intentions are enough for entering one’s signature on the board. Thirdly, the view can treat the participants either symmetrically or asymmetrically, depending on the demands of the situation (e.g., Bach’s (1995) offer-acceptance model concerns pairwise communi-cation and is asymmetric). Fourthly, BBV is capable of yielding epistemically strong (if not the strongest) joint intentions in the sense that all the information that so to speak goes into the joint intention is publicly available and publicly checkable.

The BBV covers all public joint intentions to act jointly or, for that matter, to jointly see to it that a state or event obtains. Thus it covers all publicly indicated and accepted joint intentions, be the acceptance thick or thin, and all “group-public” cases subsumable under the label “jointly intending as a group”. In such cases it will typically also be correct in general to attribute the joint intention to the group in question and to say that the group intends to perform the action in question – see Tuomela (1995, Chapter 5) for some qualifications.

However, in contrast to the public BBV spoken above, one can also formulate and deal with a purely intersubjective and non-public BBV in which everything is based only on beliefs and mutual beliefs – viz., beliefs about who are potential and actual participants and beliefs about the topic of the joint intention and about the participants’ participation intentions, about their shared background knowledge and situational information. Thus, if the participants accept a content (intention content or belief content) which they purport to be for the group, if they are collectively (and socially) committed to the content, and if there is mutual belief (but perhaps not mutual knowledge about the participants

accept-ances), then the group intersubjectively intends as a group – e.g. the account in Tuomela and Miller (1988), allows for this possi-bility. (In the case of the merely intersubjective kind of BBV the metaphorical bulletin board will exist only in the minds of the participants – or believed participants, and no communication is required.) Between the epistemically full-blown BBV and the purely subjective BBV there are various intermediate views or models, depending on what is assumed about objectivity and about the reasons for the participants beliefs (see the discussion in Tuomela, 2000, 2002b).

III. WE-INTENTIONS AND JOINT INTENTIONS ANALYTICALLY ELUCIDATED

In accordance with the grounds presented in the previous section, I will next discuss we-intentions and joint intentions in a more analytical and precise fashion. These analyses will be needed not only for the sake of clarity but also for the sake of our discus-sions later in this paper. I will start with my earlier analysis of we-intention to act jointly. Such a we-intention is expressible by “We together will do X jointly”. The analysis of core we-intention is assumed to apply also to the mutual belief-based case which is not is not (fully) public (cf. my treatments in Tuomela, 1984, 1995, 2000a, b; Tuomela and Miller, 1988). This analysis, formulated for the case of X being a joint action, can summarily be stated as follows for a collective assumed to consist of some agents A1, . . ., Ai,. . .,

Am:

(WI) A member Ai of a collective gwe-intendsto do X if and

only if

(i) Aiintends to do his part of X (as his part of X);

(ii) Ai has a belief to the effect that the joint action

opportunities for an intentional performance of X will obtain (or at least probably will obtain), especially that a right number of the full-fledged and adequately informed members of g, as required for the performance of X, will (or at least probably will) do their parts of X, which will under normal conditions result in an intentional joint performance of X by the participants;

(iii) Ai believes that there is (or will be) a mutual belief

among the participating members of g (or at least among those participants who do their parts of X intentionally as their parts of X there is or will be a mutual belief) to the effect that the joint action opportunities for an intentional performance of X will obtain (or at least probably will obtain);

(iv) (i) in part because of (ii) and (iii).

I have assumed that the participants actually exist, but I allow that a participant might in principle be mistaken in his beliefs (ii) and (iii).1Thus a single agent can in principle have a we-intention, although of course this is an exceptional case. In such a case a we-intention is not a “slice” of a joint we-intention but at best of a believed joint intention. Below I will mostly use formulations which presup-pose that the beliefs (ii) and (iii) indeed are true and, what is more, that all the agents in question really have the we-intention and that we in this sense are dealing with “genuine” we-intentions. (Notice that intentional performance of X, dealt with by (ii) and (iii), can in some cases come about without all the participants having the we-intention.)

I will not here argue for the present analysis except that its clause (i) will be discussed later. The presupposed beliefs, expressing the minimal rationality of the we-intender, and condition (iv) accord-ingly will not be commented on here (see the mentioned references, especially Tuomela and Miller, 1988, for justification).

It is presupposed by my analysis that a minimally rational we-intender should in the standard case of direct joint action be disposed to reason in accordance with the following two schemas (W1) and (W2) of practical inference (or in terms of their variants):

(W1) (i) We will do X. Therefore:

(ii) I will do my share of X. (W2) (i) We will do X.

(ii) X cannot be performed by us unless we perform action Z (for instance, in the case of an action type X, teach agent A, who is one of us, to do something related to his performance of actions required of him for X).

(iii) We will do Z. Therefore:

(iv) Unless I perform Y we cannot perform Z. Therefore:

(v) I will do Y (as my contribution to Z).

The first of these schemas in an obvious way connects we-intending to the we-intender’s own action – to his performance of his part or share of the joint action X. The second of the schemas applies to all “normally rational” we-intenders, too, but of course only when the contingent clauses (ii) and (iv) apply, and it is to be exhibited by the we-intenders’ dispositions to reason in appropriate circum-stances. This schema expresses a part of what is involved in saying that a we-intention involves a joint commitment to contribute to the realization of the content of the we-intention. This joint commit-ment also involves social commitcommit-ment, viz. that the participants are committed to one another to participation in the joint performance of X. Accordingly, (W2) clearly makes we-intentions cooperative to a considerable extent and shows that in they require interaction between the individuals – or at least disposition to interact.2

Supposing that joint intentions can be expressed by “We together will do X” or its variants, in order to cover also “standing” intentions in addition to “action-prompting” intentions, we must also take into account dispositions to we-intend, as argued in Tuomela (1991). The following, elucidation of the notion of joint intention in the direct joint action case can now be given:

(JI) Agents A1, . . . , Ai, . . ., Am have the joint intention to

perform a joint action Xif and only if

(a) these agents have the we-intention (or are disposed to form the we-intention) to perform X; and

(b) there is a mutual belief among them to the effect that (a).

In the case of joint intention the conatively used “We will do X” is true of each participant Ai.

I would like to emphasize that my analysis of joint intentions and we-intentions is conceptually non-reductive, although it is ontically individualistic or, better, interrelational (cf. Tuomela, 1995, Chapter 9, also cf. Section VI below). These notions presuppose at least a pre-analytic notion of joint intention – viz. one involved in the

participants’ minds when engaged in joint intention (and joint plan) formation such as expressed by simple exchanges of the kind: (i) “Let’s go swimming”; (ii) “OK” (recall BBV and cf. Section VI below). Thus the full concept of a person A’s we-intention to do X entails that he we-intends to do X – and hence intends to do his part of X – in accordance with and because of the agents’ endorsed “plan”, thus preanalytic joint intention, to do X together. A central argument for this kind of partial reflexivity is based on the view that an intention, firstly, cannot be fulfilled non-intentionally. If A does his part of X accidentally – e.g. does something that unbeknownst to him turns out to be describable as his part of X – that does not qualify as fulfilling his we-intention to do X: A intended to perform his part intentionally and not unintentionally. Indeed, secondly, not only must A act intentionally in the right way, he must act on the basis of the agents’ preanalytic joint intention to do X. Otherwise the participants would not properly satisfy their joint intention in terms of their intentional joint action (cf. the analogy with the single-agent case). Another argument for the reflexive nature of joint intention comes from the requirement that the participants must accept the joint intention (goal) as their joint intention.

The upshot of our present analysis is this. In the direct joint action case,joint intentionsare intentions that several agents among them-selves have, and they are expressible by an intention-expression of the form “We together will do X” endorsed by these agents. We can say, using the terminology that Mathiesen (2002) suggests, that the

intentional subject of a we-intention is “we” while theontological

subject of a we-intention is a single person (or more generally, an agent, if we consider the possibility of groups as agents).2

IV. WHAT CONTEXTUALLY DIFFERENT KINDS OF JOINT INTENTION ARE THERE?

As seen, joint intentions can involve bonds of different strength between the participants. They can be bound by explicit or implicit agreements, by public acceptances of joint plans involving joint intentions or even only by mutual beliefs about joint plans. In this section I will discuss these matters at some length under the assump-tion that the analyses of joint intenassump-tion and we-intenassump-tion given in

Section III still hold true in the different contexts and cases to be covered below. Thus, in a sense there is only one kind of joint inten-tion, but it can so to speak appear in different contextual guises. These have to do with the normative and epistemic connections that hold between the participants – either due to their voluntary choice or due to environmental factors.

In the full-blown case of joint intention we speak of “thick”, normative acceptance of a plan (or intention) of joint action. My analysis thus distinguishes between thick, normative acceptance of a plan (or intention) to perform X and a thin, non-normative acceptance of a plan. For instance, if the participants have made an agreement (consisting of mutual, interdependent promises) to perform X together, they accept the intention in the thick, normative sense. This can be called the full or fullest case of joint intention.

As claimed, in the case of accepting an intention to participate in a joint action a participant can be taken to accept “We together will perform X jointly”. Taken in the thick sense he here makes the promise to perform X together with the others. If only the thin sense is involved, he merely expresses his intention to take part in the joint action X. In the case of thin acceptance a participant merely forms the intention to participate, without promising to do it or accepting publicly to do it (cf. below).

Generally speaking, our present distinction relies on the distinc-tion between the “promise” family and the “intendistinc-tion” family of concepts. To the promise family belong agreements and acceptances of plans (when acceptance is understood in the thick, normative sense of “I accept to do this together with the others”) and to the intention family belong plain intentions, be they language-based “acceptance” intentions or “mere” intentions of the “lower” kind applicable also to children and possibly chimpanzees and the like. In the collective case the promise family involves a joint obligation to the agreed-upon content while this is not the case for intentions and intention-based notions. The possibility of thin, non-normative acceptance of a plan is indicated by the legitimacy of statements such as “I accept to participate in this plan but cannot promise to do it”.

The Bulletin Board View explicates intention formation both in the case of thick and thin acceptance. In the case of thick,

norma-tively accepted (e.g. agreement-based) joint intention BBV serves to explicate the central idea in the practice of forming a joint plan of action in the intersubjective public space. If agreement making is in question, there will also be a publicly existing social (or, if you like, “quasi-moral”) obligation to participate in joint action. This entailment of an obligation can be regarded as a conceptual truth about the notion of agreement. An agreement consists of mutual interdependent promises, and a promise involves putting oneself under an obligation. Each signed participant is taken to endorse the obligation in question and to be committed to the agreed-upon joint action and to his share of it. That he is thus committed will show up in the practical reasoning he may engage in. Such reasoning will have a content that may be described by him, roughly, by locutions like “On the basis of our agreement I am socially obligated to do my part of the joint action and hence I think I ought to do it”. The partic-ipant cannot be released from this obligation merely by changing his mind (as he can in the mere mutual belief-based account) because of the interdependence involved in agreement (cf. the discussion e.g. in Gilbert, 1993; R. Tuomela and M. Tuomela, 2003).

To summarize the case of a thick, agreement based joint inten-tion to joint acinten-tion, there are the following conceptual elements in an accepted, effective agreement, assumed to be mutually believed by the participants: (1) An intersubjective obligation to fulfill the content, say X, of the agreement, and (2) a joint commitment to X by the participants in virtue of their accepting (1). The joint commitment (2) entails that each participant is (i) committed to the participants’ collectively performing X, thus sharing the partici-pants’ collective responsibility to perform X, and (ii) committed to performing his part of X (this is expressible roughly as thoughts of the form “I will perform my part of X because of (1)”), and that (iii) each participant is suitably persistent but also flexible in his performance of his part of X. Relative to the joint intention to perform X and the joint commitment to X, (iv) in virtue of (ii) each participant is also committed to intendingto perform his part (and not only committed to performing it), as this intention is concep-tually part and parcel of the joint intention in question. The joint intention to perform X here basically consists of the participants’ we-intentions (and non-private personal commitments) to perform X

(cf. singing a duet), and therefore a participant is committed to both intending to perform his part of X and to performing it. Further-more, there is also (v) a social commitment by the participants to each other concerning their part-performances – based on the endorsement of the intersubjective obligations concerning the part-performances entailed by (1). This social commitment involves that each participant is committed to responding to the others’ normative expectations related to his performance and is thus responsible to the others for performing of his part. Correspondingly, he is also entitled to expect that the others perform their parts.

Here are some examples of joint intentions involving only thin, non-normative acceptance falling within the standard BBV. Firstly, there is the park cleaning example discussed earlier. It can have tokens satisfying BBV, even if only some kind of behavioral indica-tion of togetherness is involved (e.g., the participants show by their behavior which area they leave to the other). Next, there could be a large number of people sharing the joint intention to push a broken bus up the hill without a socially grounded obligation to do it. The joint intention could be based on only the participants’ mutual knowledge about the others’ pushing action being expressive of the intention to participate in joint pushing (rather than of their explicit agreement). Here the bus pushers jointly intend as a group and act as a group. A third example is provided by a group of people going out for a drink after a talk. There might be some leaders or, better, “operative” members who agree about joint action, but many others merely follow suit. In this asymmetric case only (and at most) the operative members would be obligated to perform the joint task.3

In these kinds of thin cases there are joint or collective commit-ments generated by joint intentions. These commitcommit-ments are inten-tion-relative and instrumental concerning the satisfaction of the joint intention. There are no (or need not be) intrinsic obligations for them either to keep their intentions (we-intentions and intentions to perform their parts) or to participate in the joint action and hence to perform their part actions. These joint we-mode commitments are appropriately persistent and are not properly consummated before the agents have jointly achieved what they we-intend or have achieved relevant consensus about the unachievability of the intended goal (or if some other mutually recognized “revocability”

condition comes to apply). Note, furthermore, that the agents can of course collectivelychange their mind, and thus joint commitments are changeable.

To end this section, I will summarize my classifications of the different cases that BBV allows. While the characterization of joint intention in Section III explicates what a joint intention is, the classification below concerns various contexts of jointly intending. The cases below are claimed to exhaustjointly intending as a group to perform a joint action:

(1) Joint intention in a thick, normative context. There are two cases: (a) Agreement-based joint intention, a strong case of BBV involving agreement making in conditions of mutual knowledge or “group-knowability”. (b) Expressed acceptance of plan in conditions of mutual knowledge. Here the acceptance isnormative, thick acceptance. (This case requires the analysans of the “bridge principle” of Tuomela (1995, Chapter 3) to hold true.4)

In (a) the participants make an agreement to act jointly; for instance, the participants make an agreement to paint a house together tomorrow. In contrast, (b) includes the somewhat wider kind of case where the participants can sign up for joint action. (Thus, if sufficiently many persons sign up e.g. to go for a bus tour next Sunday, then joint intention is created and, normally, action ensues.) In general terms, case (1) is (strongly) norma-tively group-binding on the basis of a joint obligation and

collective commitment.

(2) Joint intention in a weakly normative context. In these cases there are normative participation expectations based on an agent’s leading the others to expect normatively that he will participate in the joint action in question. When all the agents do the same, there will be a base for the participants’ mutual belief about collective commitment to participate and thus about the others’ participation intentions. We may speak of this case as being based on mutual weak promises to participate. Example: By expressing – by his words or by his actual action – that he will help to clean a park an agent leads the others normatively to expect that he will participate in the joint action. When this kind of weak promise is mutual (and understood to be mutual among

the participants) a weakly normative joint intention to partici-pate is at stake. Each of the participants is committed to this joint task under conditions of mutual belief, and thus we have collective commitment here. (Recall, however, that there can be a joint intention to clean the park, etc., also in a non-normative sense.)

(3) Joint intention in a non-normative context. In this case there is a publicly shared plan-based joint intention which falls short of being even a weak promise in the sense of case (2). In this case there, because of the publicity requirement, will be an expression of intention, which nevertheless is based on non-normative, thin acceptance (cf. I intend to participate but I do not promise that I will). In all we can say that this type of case of BBV is based on mutual-knowledge -based explicitly expressed intention to share a plan (involving a non-normative acceptance of the plan). The participants will be collectively (or jointly) committed to the plan partly because of the mutual knowledge that they share a plan, and thus intention.

(4) Non-standard cases of joint intention not satisfying the publicity requirement.This is the weakest case of joint intention. While the publicity requirement is not satisfied and while there thus are no public participation expressions, there must still be amutual beliefconcerning participation and this creates intersubjectivity sufficient for joint intention and collective commitment. This case thus falls outside the standard BBV (but not the mentioned merely intersubjective version of it).

For instance, a participant may be personally collectively committed, I say we-committed, to the joint action in question on the grounds that he will not achieve his goal that the joint action comes about without the others performing their parts properly; thus he and the others commit themselves to the action at least in an instrumental sense based on mutual belief. (Think here e.g. of the Roman military where the soldiers in a military unit were punished or rewarded collectively.) This case is not normatively group-binding in a sense allowing for justified criticism of violation (for such justified criti-cism would have to be based on public facts, e.g. public intention expressions).

In all, cases (1)–(4) entail group-binding intentions, and thus group-binding group action can and normally will ensue. Cases (1) and (2) are normative cases while in (3) only mere joint intention formation (e.g. joint decision) is involved. In (4) joint intention formation cannot even be based on a joint decision but must be based on some kind shared implicit understanding of the situation and the other participants’ relevant mental attitudes.5

V. COLLECTIVE ENDS AND WE-INTENTIONS

It is often claimed that one can intend only one’s own actions. Recently, Baier (1997), Velleman (1997), Stoutland (2000), and S. Miller (2001) have presented versions of that view. I will below discuss this thesis, which at least seemingly is a criticism against my notion of we-intention as an aim-intention. As Seumas Miller has explicitly criticized my account concerning this point in his 2001 book, I will discuss primarily his formulations. The considera-tions in this section seem not to be sensitive to the distinction between the use of different action concepts – such as performing, bringing about or jstit. I will therefore mainly speak of performance or bringing about, as they fit best the joint action case involved here. Furthermore, I will use the words ‘collectively’ and ‘jointly’ interchangeably.

I take Seumas Miller to claim that in the case of joint social action the participants’ can only have as theircollective endthat the action in question (or, alternatively, its logically inbuilt result event or state) comes about but that no participant of joint action can intend that the action or its result comes about (see Miller, 2001, p. 64).6 Let me state this as:

(1) When intentionally performing a joint action no partici-pantcan(on conceptual grounds) intend the action, or its result event, to be brought about by the participants. In contrast, I claim this:

(2) When intentionally performing a joint action every partic-ipantmustintend, viz. we-intend, the action, or its result event, to be brought about by the participants.

(1) and (2) contradict each other. (1) is based on Miller’s stipula-tion, but when the stipulation is removed there is no incompatibility. We-intentions, viewed as aim-intentions, are perfectly acceptable intentions. Intentions are conations and so are collective ends in Miller’s sense (by his own admission). The main idea involved in conation is intentional striving towards a goal (“content”) to which the agent has bound himself. If an intention is rational, the agent must believe that there is some chance that the intended content will be realized – not necessarily due to the person’s own action but e.g. due to his group’s action. However, Miller claims that an end is a conation but not an intention. This I find a conceptually impossible combination. ‘Conation’ is an old-fashioned term for intention (and striving) and that is at bottom all there is to the matter, fine distinctions apart.

According to Miller, collective end is not only a conative notion but even the intentionality of action seems to be analyzable in terms of ends (cf. Miller, 2001, p. 112). This strongly suggests to me that ends form a kind of aim-intentions (rather than that a new account of the intentionality of action is being proposed). I accordingly claim that basically collective end in Miller’s account has, and must have for it to work, the same function as it has in my account, except that in my account collective intentions (in the sense of Section III) explicitly deal with the strong notion of jointly intendingas a group. Let us now apply the above ideas to the case of a group of partici-pants who collectively or jointly intend to perform an action, X, jointly or to jointly bring about a certain end, E. We can say that jointly bringing about E amounts to performing a joint action, X. (In the case of E being a collective end, it is not in general required that E be brought about by the agents’ jointaction, but here I will assume this for the sake of exposition.) The following now holds:

(3) The participants can (‘can’ in at least a conceptual and metaphysical sense) rationally jointly intend to achieve E if and only if they mutually believe that E can with some probability be brought about by it.

Rationality here only means that there is a subjective success condition involved: the intention is realizable with some probability. In order not to lose the collective conation that (3) involves we must here require that each individual has the conation in

ques-tion, because otherwise we cannot arrive at intentionally performed joint action when the joint intention is realized. Furthermore, I will require that intentions are reflexive in the sense that successful satisfaction of intention must take place as intended (see Searle (1983, Chapter 8) and Tuomela (1995, Chapter 3) for arguments and discussion).

Moving to the individual level corresponding to the collective level, we arrive at we-intentions:

(4) When some participants jointly intend E each individual participant must we-intend E and consequently “action intend” to participate in the bringing about of E, viz. to perform his share or part of the participants’ joint action to bring about E.

Seumas Miller adopts the view that one can only intend what one can bring about by means of one’s own actions. Thus, according to him, one can only intend one’s bodily actions and their direct consequences. But this is a stipulation and a not a very happy one, although it contains the acceptable core idea that intentions are act-relational (viz. always related to one’s actions, see below). I have claimed that one can strive for (“conate”) and thus intend that a joint action, X, or a state, be realized by the participants’ collective action. One can claim this while accepting that intentions also concern one’s own actions. Thus:

(5) Individual agent A intends to perform an action X together with the other participants only if A intends by means of his actions to bring about some result such that when all the other participants similarly perform their part actions, a performance of X is intentionally generated at least with some probability.

An intention of the present kind I have called an “aim-intention” in contrast to an “action intention” (viz. basically the only kind of intention that Miller accepts in his theory). An aim-intention can be satisfied without the aim-intending agent alone satisfying the intention, whereas in the case of an action intention the agent must believe, if rational at all, that he can (with some probability) satisfy the intention by means of his own actions. An aim-intention can in typical cases be rendered as “agent A intends, by his actions,

to bring about state or event E”, where E can in principle (viz. on conceptual grounds) be any kind of (contradictory or non-tautological) state (including another agent’s mental state of his having an intention). While, trivially, nobody can (directly) intend another person’s actions, one can still – depending on the circum-stances and one’s success beliefs – intend to bring about, by his actions, e.g. that another person comes to intend or to perform a certain action.7 An aim-intention can be called an end (or goal) in the sense that its content E is an end that the agent has. In the context of a joint action the end is collective. A collective end according to my view is, roughly, one that the agents have collectively accepted as their collective end, assumed to satisfy the Collectivity Condition (recall Section II).

It is essential that there is an individual conation (intention) in the present context concerning also E (or X) and not only one’s part action. An intention involves commitment, and the crux is that in the case of joint intention – required of intentionally performed joint action – there must be collective and, consequently, also individual-level commitment concerning end E (recall the discus-sion in Sections II and IV). For instance, you and I can intend and be committed to some joint activity, e.g. you and I can jointly intend to build a bridge together. This joint intention consists of our we-intentions to build a bridge. Each we-intention is an aim-intention but it entails a participation intention in the case of each partic-ipant: I intend to participate in our building the bridge, and similarly for you, provided the right presupposition beliefs are present. As emphasized in Section III, the we-intending participants are accord-ingly entitled to infer according to schemas (W1) and (W2). While the participation intentions are personal intentions, they are not private intentions but we-mode intentions, thus intentions qua group members.

To summarize, Seumas Miller has not succeeded in showing in his 2001 book that intentional joint action can be based on collective ends which are not intentions. I agree of course that collective ends need not be action intentions. However, aim-intentions must be present. The collective end that Seumas Miller speaks about (but does not sufficiently clarify in the book) is an aim-intention, more specifically an intention satisfying the intention expression “We

together will bring about E” accepted basically by all participants in this context. Miller says that his notion of an end is a conative notion and bases the intentionality of action on it. This indicates that it is an aim-intention, for conation involves striving for an end to which one has bound oneself. The conflict here is partly if not completely terminological, as collective ends are conative states (intentions – at least as I understand conation). Furthermore, collective ends are in any case on a par with intentions in that one can ask how one can have ends that one cannot by one’s own actions reach. The answer is that ends and intentions at bottom concern one’s own acting, although it cannot be required that one alone make the end state realized.

To end this section, I will discuss the problem at hand in some-what more general terms. Bratman’s account of shared intention (reprinted as Chapter 6 in his 1999 collection) has been criticized e.g. by Velleman (1997), and Stoutland (2002) on the ground that his notion of a person’s intending that we (his group) perform a joint action, J, or briefly “I intend that we J” is not an intention notion at all. This discussion of course is closely related to the issues considered above. The following principles have been suggested for single-agent intentions (see Bratman, 1999, pp. 148–149):

(1) Own action condition (OA): One can intend only to do something herself.

(2) Control condition (C): One cannot intend what one does not take oneself to control.

(3) Settle condition (S): One can only intend what one believes her so intending settles.

The settle condition entails that “for me to intend that we J I must

. . .see my intention as settling whether we J” (p. 149). There is also

an additional principle that Stoutland (2001) suggests:

(4) Responsibility condition (R): To intend to do something is to commit oneself to do X so as to thereby commit oneself to take full responsibility for having done X (if and when one does X).

I take conditions (1)–(4) to be on the right track and (almost) acceptable for action intentions. (I say “almost”, because at least the control condition seems somewhat too strongly formulated –

less than full control will suffice – and full responsibility in (4) also needs relaxation.) Leaving a more detailed discussion of (1)–(4) for another occasion, we can notice that these conditions clearly do not – not at least without modification – apply to aim-intentions (such as we-intentions).

Assuming for simplicity’s sake that the presupposition beliefs (ii) and (iii) in the analysis (WI) of we-intention in Section II are true, our discussion in that section (and Section II) then warrants the following claims in the case of we-intentions, remembering that we-intentions are personal “slices” of joint intentions:

(1J) Our action condition (WA): A person Ai, can we-intend

to perform something X (if and) only if he is one among some persons A1, . . ., Am who jointly intend to perform

X together such that his action intention is to participate in these agents’ joint performance of X. (If we were dealing with the “irrational” case with false beliefs (i) and (ii) in (WI), the analysans here would have to be relativized to Ai’s belief.)

(2J) Joint control condition (JC): Some persons A1, . . . , Am

cannot jointly intend to perform an action X jointly and cannot we-intend to perform X in this case unless they mutually believe that they can control X to a substan-tial degree (at least these agents should mutually believe that their performing X together is not impossible in the circumstances in question).

(3J) Joint settle condition (JS): If some persons A1, . . ., Am

we-intend to perform X under conditions of mutual belief, they must also mutually believe that their so intending (psychologically, not perhaps in an overt action sense) settles that they will perform X together.

(4J) Joint responsibility condition (JR): For a person Aito

we-intend to do something X is in part to commit himself to X (in a context where some agents A1, . . . , Am, of

which Ai is one, jointly intend to perform X), so as to

thereby commit himself to take partial responsibility for their having performed X together and to take responsi-bility for his having participated in the performance of X (if and when he actually does participate). (Cf. (1J) for

the qualification concerning the case of false beliefs (i) and (ii).)

A joint intention to perform X together requires that the partici-pants believe that they will perform X intentionally, indeed as jointly intended (as seen above, in contrast to what Bratman, 1999, says on p. 147). This may seem viciously circular, but I will below in Section VI argue that the conceptual situation is not that bad. As to the functionality of joint intentions and their component we-intentions, they lead to the right kind of reasoning (recall the schemas (W1) and (W2)) and to the right kind of action intentions and consequent actions (recall my above discussion and emphasis on the conceptual and factual dependence on X of a we-intention).

VI. ON THE ALLEGED CIRCULARITY OF THE CONCEPT OF WE-INTENTION

In this section I will consider the charge, made against my account by Searle (1990) and S. Miller (2001), that the notion of we-inten-tion is circular in an unacceptable sense.8 Let me start by quoting Searle:

We are tempted to construe ‘doing his part’ to mean doing his part toward achiev-ing the collective goal. But if we adopt that move, then we have included the notion of a collective intention in the notion of ‘doing his part.’ We are thus faced with a dilemma: if we include the notion of collective intention in the notion of ‘doing his part,’ the analysis fails because of circularity; we would now be defining we-intentions in terms of we-intentions. If we don’t so construe ‘doing his part’, then the analysis fails because of inadequacy. (Cf. also S. Miller, 2001, pp. 71–73)

I will argue below that there is in fact no vicious circularity of any kind in the account presented in Section III above. To begin my response, consider an agent A and assume that he we-intends to perform X (or, alternatively, to bring about or see to it that X) together with some other agents. It seems that this must be taken to presuppose that A believes that these participants jointly intend to perform X; and as A’s we-intention at bottom is in a constitutive sense just a slice of the joint intention (recall Section III), we have circularity. (In normal cases the belief is a true one, of course.) But things are not so simple. As will be seen, in the

“preanalytic”, common-sense case a rather meager notion of we-intention is shown to be psychologically and functionally sufficient, and my response will rely on this. In particular, it is not required that joint intentions in this preanalytic account be taken to consist of shared we-intentions.

According to my analysis of “strong” we-intention (WI) (in Tuomela, 1984; Tuomela and Miller, 1988, etc., and section III above) A we-intends to perform a joint action X (or to perform X together with the others) if and only if (i) he intends to perform his part of X as his part of X in part because he (ii) believes (presupposes) that the “joint action opportunities” for X obtain (e.g. that there are sufficiently many participants for an intentional performance of X to come about, if there is that kind of numer-ical requirement, and they actually perform their parts of X) and also (iii) believes (presupposes) that there is mutual belief about the obtaining of the joint action opportunities. The central requirement (i) involves that he intends to perform his part of X (an action inten-tion) and intends to perform X together with the others (or put more technically, intends that X will be performed by them).

To recapitulate in somewhat more explicit terms, we have: (1) A we-intends to perform X if and only if

(a) A intends to perform his part of X as his part of X, and (b) the aforementioned presupposed beliefs (ii) and (iii) are in place and function as partial reasons for (i).

Next, I suggest that we can give a preanalytic, common-sense account of (1)(a) by:

(2) A intends to perform his part of X as his part of X if and only if

(a) A intends to perform his part of X, and

(b) A intends to perform X with the others in part because (of his belief that) the others intend to perform their parts of X and intend to perform X with the others.

(2) and especially its clause (b) can be viewed as preanalytic in the sense that ordinary people can well be assumed to actually think in these terms. (2) thus views the matter from the ordinary agent’s rather than from the theoretician’s point of view. In conditions of mutual belief (cf. (1)(b)), the present account gives a functionally

adequate solution to the problem of joint intention concerning what the agents need to have in their minds when jointly intending to perform X together. We can say more colloquially that in the present situation A intends to participate and takes the others to be in a similar position as he is with respect to X. The analysans in (2) does not (directly) refer to a we-intention or joint intention and thus it does not make (1) nor (WI) viciously circular.

However, while the preanalytic account does not, to quote Searle again, “construe ‘doing his part’ to mean doing his part toward achieving the collective goal”, it might still be argued, from a theoretician’s point of view, that underlying (2) there is implicit reference to joint intention. This is because in (2) A’s intention to perform his part of X as his part of X can be satisfied only if the participants intentionally jointly perform X and because such inten-tional joint performance may be argued upon analysis to depend on the joint aim or intention to perform X. I will grant that the preanalytic account presupposes intentional joint performance of X, but this does not yet entail circularity. Even if it thus is granted that the preanalytic account presupposes intentional performance of X it does not take a stand on what precisely the latter notion involves, and still it works perfectly well in actual practice.

However, when a theory of joint intention adds an analysis of intentional joint action referring to joint aim or intention, some amount of circularity comes about. While this need not be an unavoidable move, this view applies to the Tuomela and Miller 1988 paper where (on p. 377) it is required that the agent A should intend to perform his part of X and perform it with the (general) purpose of the action X coming about. In Tuomela (1995, p. 140 ff.), a similar requirement is made that the agent aim at, and be committed to, the realization of X. Analogously, Mathiesen (2002) defends the account by Tuomela and Miller (1988) against Searle’s criticism by imposing what can be regarded as an equivalent requirement that A intend to perform his part of X in order for the participants to successfully intentionally perform X.9

In accordance with what was just said, I can go along with my critics to the extent that doing one’s part in a sense presupposes collective (or joint) intention but only in an implicit and unanalyzed sense of aiming at the joint action. I deny that this sense creates