Draft please do not cite or circulate Comments welcome: brendan.dekenessey@gmail.com Promises as Proposals in Joint Practical Deliberation Brendan de Kenessey Joint practical deliberation is the activity of deciding together what to do. 1 When two colleagues discuss where to meet for coffee, when two partners have a conversation about whether to have a second child, when friends on a road trip argue about whether to take the scenic route or the freeway, they are engaged in joint practical deliberation. They are trying to arrive at a single plan of action that expresses and governs both of their wills. This paper begins from two basic observations about this activity. First, we can make joint decisions regarding not just what we will do, but also what you will do or what I will do. For example: if we are deliberating together about how to write a coauthored paper, we might jointly decide that I will write the introduction, leaving it quite open what you will do. Second, if we do in fact deliberate together, we must have some speech acts by means of which we propose potential joint decisions, putting them forward for joint consideration. Combining these two ideas, we can infer that there are speech acts the force of which is to propose joint decisions regarding what you will do or what I will do. If we investigate what such speech acts would look like, they are surprisingly similar to speech acts we already have: requests, demands, offers, and promises. This suggests a hypothesis: these speech acts just are proposals within joint practical deliberation. Requests and demands are two ways of proposing joint decisions about what you will do; offers and promises are two ways of proposing joint decisions about what I will do. In this paper, I defend one part of this hypothesis: the claim that promises are proposals to make joint decisions about what the speaker will do. 2 1 I am very grateful to Dylan Bianchi, Stephen Darwall, Nilanjan Das, Sally Haslanger, Julia Markovits, Daniel Muñoz, Sofia Ortiz-Hinojosa, Bernhard Salow, Kieran Setiya, Bradford Skow, Judith Jarvis Thomson, Quinn White, and the MIT philosophy community for their helpful feedback on the ideas presented in this paper. 2 I am not the first to propose this view. Margaret Gilbert has proposed a joint decision account of promises: for one person to make a promise to another is for them jointly to commit themselves, by an appropriate, explicit process, to the decision that one of them ( the promisor ) is to perform one or more specified actions (Gilbert 2011a: 99). So, what does this paper have to add? First, Gilbert s defense of the joint decision account is limited to the final few
Call this the deliberative theory of promises. The deliberative theory is an analysis of promises; it addresses the question, what are promises? Its answer is: to make a promise just is to make a certain kind of proposal in joint practical deliberation. This analysis aims to explain a phenomenon recognized in ordinary language (promises) by identifying it with a posit within a theory (proposals in joint practical deliberation). Compare the claim that water is H2O: we accept this analysis because it enables us to explain water s various observable properties by appeal to the theory of molecular chemistry. Similarly, I will argue that we should accept the deliberative theory because doing so enables us to explain promises various properties by appeal to the theory of joint practical deliberation. My argument proceeds as follows. I begin by setting out an account of joint practical deliberation, motivated by the idea that joint practical deliberation has the same structure as individual practical deliberation ( 2). Combined with our analysis, this account yields substantive and independently motivated predictions about the properties of promises. In 3, I argue that the deliberative theory predicts and explains several features of promissory obligation: (i) the fact that promissory obligations are owed to the promisee in particular, (ii) the fact that promises require uptake from the promisee to be binding, (iii) the ways in which promises constrain the promisor s subsequent deliberation, and (iv) the promisee s power to release the promisor from her obligation. In 4, I argue that the deliberative theory provides the best explanation of why promises elicited by coercion and deception are not binding. I then show how the basic ethical premise of the deliberative theory that we have reason to abide by our joint decisions can be accommodated by various different ethical theories ( 5). I conclude by raising a general challenge to alternative theories of promising ( 6). If the arguments of this paper are sound, then the activity of joint practical deliberation necessarily brings with it a speech act that has all the characteristic features of promises: namely, a (certain pages of a largely critical paper; this paper offers a more sustained development and defense of this idea. Second, the theory of joint practical deliberation I present in 2 departs from Gilbert s in substantive ways, with important implications for the nature of promises. I compare my view to Gilbert s in more detail in footnote 25. 2
kind of) proposal to make a joint decision regarding the speaker s actions. Anyone who denies the deliberative theory is committed to holding that this speech act is not identical with promising. But this implies that promising is redundant: everything we do with promises, we could do instead with proposals in joint practical deliberation. So, the opponent of the deliberative theory is left in the unattractive position of maintaining that promises are not all that important: once we have joint practical deliberation, we have no need for promising. This undercuts the motivation for alternative theories: why posit a further, sui generis speech act of promising if joint practical deliberation includes a speech act with just the same properties? Better to accept that promises and proposals in joint practical deliberation are one and the same. Before we jump in to these arguments, it will be helpful to locate my proposal within the prior literature. Of the extant theories of promises, the deliberative theory is closest to conventionalism, the view that promises are a move within a conventional social practice. The deliberative theory also holds that promising is a move within a social practice: the practice of joint practical deliberation. This may be cause for worry, as conventionalism has been almost universally rejected over the past quarter-century. It will thus be worthwhile to begin by taking a careful look at conventionalism and the objection that has led so many to reject it ( 1). 1. The rejection of conventionalism Conventionalism enjoyed a long reign as the orthodox theory of promises. Originally and influentially defended by David Hume (1739/1978: 516-525 [3.2.5]), conventionalism was then taken up in the mid-20 th century by such luminaries as John Rawls (1955; 1971: 344-350), John Searle (1969), and G. E. M. Anscombe (1978). All of these theorists held that promises are moves within a conventional social practice of making and keeping promises, and thus that our reason to keep our promises derives from our reason to obey the rules of that practice. But this idea has now been widely rejected, in large part due to the objections raised in T. M. Scanlon s nowclassic article Promises and Practices (1990; see also Shiffrin 2008 and Owens 2012: 129-141). 3
Whether or not they agree with Scanlon s positive proposal, most philosophers working on promises now share in his conviction that when promises give rise to clear obligations, these can be accounted for on the basis of general moral principles that do not refer to the existence of social practices (1990: 220). 3 The resulting post-conventionalist literature has produced several alternative theories of promising. 4 Though they each have their merits, I will not discuss these views here. My aim in this section is not to evaluate the extant theories of promising; it is to show that there is another, fruitful approach to promises that has been largely overlooked. To clear space for this idea, however, we first need to understand the reasons behind the rejection of conventionalism. Let s begin by stating the conventionalist view more clearly. The thesis of conventionalism is that promising is a move within a conventional social practice. Following Rawls, we can understand a practice as any form of activity specified by a system of rules which defines offices, roles, moves, penalties, defenses, and so on, and which gives the activity its structure (1955: 3). Examples of practices include games such as baseball and chess, institutional procedures such as committee meetings or court proceedings, rituals such as weddings and graduations, and more informal conventions such as queuing in a line. What these activities all have in common is that they have rules dictating what persons are to do depending on what role they occupy and what moves have been performed in the activity so far. 3 Some exceptions to this trend are Deigh (2002), Shockley (2007), and Taylor (2013). 4 The two most prominent non-conventionalist theories of promising are the expectation theory and the normative powers theory. On the expectation theory, the obligation to keep promises is an instance of the more general obligation to not mislead others, applied to the special case of expectations regarding our future conduct. Advocates of the expectation theory include Árdal (1968), Narveson (1971), MacCormick (1972), Thomson (1990: Ch. 12) and Scanlon (1990; 1998: Ch. 7). Both McMahon (1989) and Kolodny and Wallace (2003) propose hybrid theories that combine conventionalism with the expectation theory. For criticism, see Raz (1972; 1979), Kolodny and Wallace (2003), Gilbert (2004), Owens (2006; 2012: Ch. 9), Southwood and Friedrich (2009), Darwall (2011), and Heuer (2012a). The normative powers theory holds that our obligation to keep our promises is sui generis, not to be explained by appeal to more general principles governing either our care for others expectations or our participation in social practices. Its explanation of promissory obligations instead appeals to our interest in having the normative power to make binding promises. Versions of the normative powers theory have been defended by Raz (1972; 1977), Shiffrin (2008; 2011), Owens (2006; 2007; 2008; 2012), and Heuer (2012b). Though I do not have the space here to discuss either theory directly, I argue in 6 that both theories are vulnerable to the objection that, given the existence of joint practical deliberation, they render promising redundant. 4
An immediate clarification is in order. It is plausible to regard speaking a language as a social practice; so, if we assume (likely falsely) that promises can only be made using language, then there is a boring sense in which promises are moves within a practice. Friends and enemies of conventionalism alike should agree that this uninteresting claim is not its thesis (see Raz 1972: 100; Scanlon 1990: 213-216). One way to avoid this confusion is to distinguish between the claim that promising is necessarily performed by means of a move within a practice and the claim that promising is a move within a practice. Consider the difference between hitting a home run and making the crowd go wild. Hitting a home run is a move within the practice of baseball; making the crowd go wild is not itself a move within baseball, but instead something you can do by means of a move within baseball (cf. Austin 1962). The boring claim is that promising is typically performed by means of a move within a practice (uttering a sentence in a language); the conventionalist s claim is that promising is a move within a practice. But for those skeptical of this distinction, there is also an easier solution: we can say that conventionalism s thesis is that promising is necessarily a move within a practice other than the practice of speaking a language. 5 What is the practice within which promising is a move? The conventionalist s answer to this question has always been simple: the promising practice. The promising practice has two roles (promisor and promisee), two moves (promising and releasing), and one rule: if a promisor promises a promisee that she 6 will φ, then the promisor must φ unless (i) the promisee releases her from her promise beforehand or (ii) certain invalidating conditions (e.g. coercion or deception) obtain (Jones 1966: 288; Rawls 1971: 344-345; Kolodny and Wallace 2003: 120). Like the deliberative theory, conventionalism is an account of what promising is: promising is a move within the conventional promising practice. To get a complete theory, however, we need an account of why we ought to obey the rules of the promising practice. Early conventionalists focused on the detrimental effects of breaking these rules on one s own reputation (Hume 1739/1978: 522; see also Sheinman 2008) or on general trust in the practice 5 Thanks to Judith Jarvis Thomson for pressing me to clarify this point. 6 I make the promisor female and the promisee male throughout. Otherwise, I default to the female pronoun. 5
(Pickard-Cambridge 1932: 153-167). But these accounts wrongly predict that one may break a promise so long as one knows that one won t get caught (Ross 1930: 39; Scanlon 1990: 221). Thus most conventionalists now endorse Rawls account: breaking a promise is wrong because to do so is to accept the benefits of the promising practice without doing one s part by obeying its rules (Rawls 1971: 342-350; for criticism, see Nozick 1974: 93-95; Robins 1984: 127-132; and Shiffrin 2008: 482). Rawls claimed that such free-riding is wrong independently of its consequences, thus avoiding the objection faced by earlier conventionalists. We can now ask why conventionalism has been so widely rejected. Though several objections have been raised, the most influential has been the objection that conventionalism makes the existence and nature of promising too contingent. 7 Conventionalism entails that our ability to make promises depends upon the prior existence of a social practice of promising. This seems to imply that the existence of promises is contingent in the following sense: socially and morally competent agents could fail to have a shared practice of promising, and so could lack the ability to make promises to each other. Yet many have thought that moral agents necessarily have the power to make binding promises to one another (Locke 1689/1980: 13 [Ch. 2, sec. 14]; Raz 1977: 214-215; Scanlon 1990: 201; Thomson 1990: 303-304; Shiffrin 2008; Pink 2009: 400-401). Scanlon makes this point vivid with a state of nature case. Suppose that Scanlon encounters a stranger on a desert island: they are on opposite sides of the river and each has lost their weapon on the other s side. Scanlon manages to convince the stranger that he will throw the stranger s boomerang across the river if the stranger first throws Scanlon s spear back to him. The stranger throws Scanlon the spear and Scanlon walks away, leaving the boomerang where it fell. Scanlon judges that what I have done in this example is no less wrong than it would have been if I had promised the stranger that I would return his boomerang if he threw back my spear it is the same kind of wrong (1990: 201). It seems the conventionalist cannot account for this: 7 A second important objection is that conventionalism cannot capture the fact that breaking a promise wrongs the promisee in particular (Scanlon 1990: 221; Darwall 2011: 263-264; Owens 2012: 135). On Rawls account, for example, breaking a promise seems to wrong everyone who participates in the promising practice equally, not just the promisee. Though I take this objection seriously, I propose we table it until later ( 3). 6
by hypothesis, the two are members of different societies that have made no contact, and so presumably do not have any shared practice of promising. Buttressing this intuition with argument, Seana Shiffrin (2008) claims that the existence of promising is a necessary condition for interpersonal relationships that meet minimal conditions of moral decency. Since the power to promise is an essential tool to forestall and neutralize morally problematic inequalities between people, a world without promising would be a world without relationships of equal respect (485). If, as is plausible to assume, moral agents necessarily have the capacity to engage in equal and respectful interpersonal relationships, then moral agents must necessarily have the power to make binding promises. This gives us reason to reject the conventionalist s idea that promises are inventions that we could have failed to invent and still gotten by morally, although perhaps less well and less efficiently (498). These objections focus on the conventionalist s claim that the existence of promises is a contingent matter. But conventionalism also seems to entail that the nature of promises is a matter of contingent convention. Just as we could have ended up with a practice similar to baseball, but with four strikes instead of three, conventionalism implies that we could have ended up with a practice similar to our actual practice of promising, but with different rules. Perhaps we could have counted deceived promises as binding, allowed promises made in solitude to bind without any need for a promisee, or given promisors a limited number of free passes they could use to release themselves from prior promises. But this seems to render the nature of promising too flexible and contingent on historical happenstance. It seems to be far more necessary that (say) promises require promisees than it is that three strikes makes a strikeout. The existence and nature of promises thus seem to be less contingent than the conventionalist claims. But the conventionalist is only vulnerable to this objection because she has bundled together two claims that can and should be separated: the claim that promising is a move within a practice, and the claim that this practice is optional, in the sense that social interaction between moral agents would be possible without it. I suspect that conventionalists and 7
their critics have bundled these claims together because they assume that all social practices must be optional. But this assumption is simply false: there can be non-optional social practices. Consider the case of assertion. Assertion has all the signature features of a move in a social practice. It is governed by a rule that dictates which assertions are permissible: the muchdiscussed norm of assertion (for a review, see Weiner 2007). Like the rules of other practices, the norm of assertion does not determine when we ought to make an assertion all things considered after all, sometimes we ought to lie instead, it says how we ought to behave qua assertors. (Just as the rules of baseball say how we ought to behave qua baseball players). In addition to this basic norm, conversation is governed by a complex set of further rules determining what assertions are relevant, cooperative, or even make sense (e.g., Grice s maxims of conversation, see his 1989). And, as in all practices, the permissible moves in a conversation depend upon the moves that have already been made (Lewis 1979a). No wonder philosophers of assertion, just like philosophers of promises, have felt irresistibly pulled to the analogy with baseball. 8 Yet no one is tempted to say that the assertion practice is socially optional or that its rules are arbitrary. The ability to share information in the ways assertion makes possible seems to be a necessary precondition for cooperative social interaction. And the rules governing assertion are widely held to be derived from general principles of rationality, cooperation and/or cognition (Korta & Perry 2015). Thus though assertion is a move in a practice, this practice is not conventional in the Lewisian sense of the term: it is not one of multiple different possible solutions to a coordination problem (see Lewis 1969/2002: 70). The practice of assertion, with just the rules it has, is the unique solution to the problem of sharing information. Taking assertion as our model, we can thus make room for a nonconventionalist practice theory of promising. Such a theory would claim, with the conventionalist, that promising is a move in a 8 Compare Rawls (1955) with Lewis (1979a). See also Williamson (2000): someone who knowingly asserts a falsehood has thereby broken a rule of assertion, much as if he had broken a rule of a game; he has cheated (238). Others who endorse the assertion-game analogy include Wittgenstein (1953/2009), Searle (1969), Brandom (1983), Dummett (1993), and Rescorla (2009); for criticism of this analogy, see Cappelen (2011) and Maitra (2011). 8
practice, but depart from the conventionalist in claiming that this practice is not socially optional and that its rules could not easily have been different. The deliberative theory of promises occupies this uncharted position in logical space. Like conventionalism, the deliberative theory is a practice theory: it claims that promises are a move within a practice. But unlike conventionalism, the deliberative theory does not take the power to promise to be activated only by the happenstance that a social convention of promising has developed, the major and minor details of which are also socially contingent and subject to change (Shiffrin 2008: 483). For the practice within which it locates promises joint practical deliberation is plausibly as essential to social life as is promising itself. Like the practice of making assertions, joint practical deliberation serves a function essential to social interaction: it enables us to plan and act together on terms that are justifiable to all. And as we shall see, this function dictates the rules of joint practical deliberation, which in turn explain the features of promises. So it is not a contingent fact that moral agents engage in joint practical deliberation, nor that this practice has the particular rules it does. Thus it is no more contingent that moral agents have the power to promise, or that this power takes the particular shape it does. 9 2. Joint practical deliberation The thesis of this paper is that promises are proposals in joint practical deliberation. More precisely: to promise to φ just is to propose (in a certain way) to decide together with your 9 Anyone who asserts the necessity of promising has to contend with Korn and Decktor Korn s (1983) argument that there exists at least one society with no practice of promising: the Tonga Islands of the South Pacific. In Tonga, they claim, statements of the form I will φ are not taken to commit the speaker to φing, but instead merely express solidarity for the addressee s desire that φ be done. No alternative form of speech plays the role of promising. Though the case of Tonga does seem to show that promising is not culturally universal, it does not demonstrate the same for the practice of joint practical deliberation. For it may be that Tongans engage in joint deliberation, but refrain from proposing joint decisions in the particular way involved in promising. Though Tongans do not promise, they do appear to make demands and requests of one another; they also marry and exchange gifts (Evans 1996). Given my hypothesis that demands, requests, and agreements are also moves in joint practical deliberation (see 2.4), this indicates that the Tongans do engage in this practice. Why, then, don t they promise? Perhaps because Tongans regard the future as indefinite and uncertain (Korn and Decktor Korn 1983: 449). They thus avoid long-term planning, restricting the potential scope of joint decisionmaking to the here and now. While I might demand or request that you do something right now, there s no point in promising to do something right now, since I can just do it. Promises require a longer-term view; so, it makes sense that Tongans might avoid promises while still making joint decisions in other ways. 9
addressee(s) that you will φ. My argument for this thesis is an inference to the best explanation: if we accept the hypothesis that promises are proposals in joint practical deliberation, we are in a position to explain many of promises puzzling features. To give this argument, however, we first need a theory of joint practical deliberation. That is the task of this section. 2.1. The parallel structures of joint and individual practical deliberation Our investigation begins from the premise that joint practical deliberation is an activity of the very same kind as individual practical deliberation, the activity individuals engage in when they deliberate alone about what to do. Like making lasagna, writing a philosophy paper, or watching a movie, practical deliberation is the kind of activity that can be performed both together and alone. This idea suggests a strategy for investigating joint practical deliberation: look at the structure of individual practical deliberation, and then ask how two or more agents could engage in a single activity with the same structure. 10,11 What is the structure of individual practical deliberation? Consider a concrete case: I am deciding where to get lunch. I am faced with a set of options, the potential actions I am choosing between (to go to Restaurant A, to go to Restaurant B, to go to Restaurant C). My aim is to make a decision that narrows those options down to one: to go to one of the available restaurants (to go to Restaurant B). However, my decision will (typically) not be arbitrary it will be based on reasons. For example, the fact that Restaurant A sources its food from local farmers might give me some reason to go there, while the fact that I m craving a burrito right now might give me more reason to go to Restaurant B, which makes a mean burrito. 10 It is this strategy that distinguishes my account from other work on joint practical deliberation (e.g., Westlund 2009; Laden 2012; Bratman 2014: Ch. 7): none of these philosophers pursue the idea that joint decision-making is isomorphic in structure to individual decision-making. 11 Let me clarify at the outset the relation between joint practical deliberation and joint action more generally. I take joint deliberation to be a joint activity, to be listed alongside tennis and dancing as one of the joint activities in which we engage. Thus I do not aspire to analyze joint activity in terms of joint deliberation: I cannot say, for example, that what it is for an activity to be joint is for it to be based in a joint decision, for joint decisions require joint deliberation, which is itself a joint activity (cf. Arpaly and Schroeder 2015: 104). My account of joint practical deliberation thus takes the concept of joint action as given, and is intended to be neutral between the leading theories of this phenomenon (e.g., Tuomela 2006; Gilbert 2011b; Bratman 2014). Thanks to Kieran Setiya here. 10
Putting these elements together, we can characterize individual practical deliberation as the process of making a decision about which of one s options to take on the basis of one s reasons. If joint practical deliberation has the same structure, then it must be a process of making a joint decision about which of our shared options to take on the basis of our shared reasons. Let us treat the italicized phrases in this schema as functional terms: each refers to whatever it is that plays the same role in joint practical deliberation its analogue plays in individual practical deliberation. Our question now is what could play the roles picked out by these terms. Start with shared options. In the individual case, an agent s options are the potential actions she might perform. This suggests an idea: in parallel with the popular possible worlds framework for modeling theoretical deliberation (e.g., Stalnaker 1984), we can model practical deliberation using a set of possible actions (cf. Gibbard 2003: Ch. 3). 12 In this model, the content of a decision (to go to Restaurant B) is a set of possible actions (the set of all actions in which I go to B), and the effect of this decision is to exclude from my future deliberation any possible actions outside of that set (e.g., going to C instead). My options are the result of partitioning my possible actions into exclusive subsets, each of which is the potential content of a decision. A virtue of this model is that it can easily be extended to the case of shared options. Instead of considering possible actions an individual agent can undertake, we can consider the possible combinations of actions multiple agents could compatibly undertake. This will include not only paradigm cases of joint action, but also any possible combinations of our individual actions, e.g. the combination in which I eat a sandwich in New York while you conduct a symphony in Los Angeles. We will model the shared options of two or more agents as sets of possible combinations of their actions. 13 Joint decisions take one of these sets as their content (e.g., the set 12 Different theorists will read the possible in possible actions differently: see Hedden (2012) for a review. 13 Let me emphasize how weak this model s claim is. To claim that our shared options are composed of combinations of our individual actions is just to say that whenever we do something, that involves you doing something and me doing something. This does not imply that everything we can do is a sum of something I could do without you and something you could do without me: you and I can carry a heavy couch together, though neither of us would be able to do our part in this action without the other. Nor is it to say that when we break down what we do into what you do and what I do, there will be any natural way of describing those individual components. For 11
in which we take a trip to Vegas together), and exclude from joint consideration combinations of actions that fall outside that set (e.g. my staying at home while you go to Vegas). We can now see how it is possible to make a joint decision about what I will do. You and I can decide upon a shared option that substantively constrains only my actions. Consider the set of all possible combinations of our actions in which I read your draft. Jointly deciding on this set would significantly constrain my actions, excluding from our future consideration any possible combinations of actions in which I don t read your draft. However, it would leave your actions largely unconstrained: any action you could perform that is compatible with my reading your draft is part of one of the combinations of actions in this set. Thus it seems that nothing of substance is missing if we describe our joint decision simply as a decision that I will read your draft, omitting the clause and you will not prevent me from doing so. 14 This is a crucial point for the deliberative theory, since promises concern only the speaker s actions. Our account of shared options implies that the content of a joint decision will be a set of possible combinations of the participating agents actions. However, there is more to a representation s being a decision than the mere fact that it represents actions. I can suppose that I will φ, predict that I will φ, or tell a story in which I φ. What distinguishes my deciding to φ from these other representations of my φing is its distinctive connection with my behavior. If I decide to φ, then I will be disposed to take the means I believe to be necessary to φing and avoid performing actions I believe to be incompatible with φing (Bratman 1987: 15-17). If I do not do so, then I am failing to conform to the norms of individual practical deliberation (see Bratman 2009). Similarly, for a representation of a shared option to count as a joint decision, it must have implications for what the deliberating agents will do. If we jointly decide to φ, we should be disposed to take the means we believe to be necessary to φing and to avoid performing actions we example, there is likely no natural way to describe what I do when we dance a waltz other than to say that I am doing my part in our dancing a waltz. Thanks to Kieran Setiya for discussion on this issue. 14 Some evidence that this implicit clause is present is that we would normally be affronted if someone accepted our promise and then prevented us from carrying it out. See Darwall (2011: 268-269) and Gilbert (2011a: 99). 12
believe to be incompatible with our φing. If we fail to do so, we are failing to conform to the norms of joint practical deliberation. Thus we can state the first rule of joint practical deliberation: act in accordance with your standing joint decisions. 2.2. The standard of correctness for joint decisions Individual agents do not just decide what to do arbitrarily: they decide on the basis of reasons. To count as deliberating at all, an agent must evaluate potential decisions on the basis of her reasons and try to make a decision those reasons support. In this way, we hold our decisions up to a standard of correctness: a standard that separates those decisions that are justified from those that are not. Similarly, to count as deliberating together, we must evaluate potential joint decisions on the basis of our shared reasons, and try to make a decision that those reasons support. In doing so, we are holding our joint decisions up to a standard of correctness. Call joint decisions that meet this standard of correctness warranted, and those that fail to meet it unwarranted. What is the standard of correctness for joint decisions? It may seem the answer is simple. In the individual case, it seems uncontroversial to say that a decision to φ is justified just in case the agent has sufficient reason to φ: that is, there is no incompatible option ψ such that the agent has more reason to ψ than to φ (cf. Schroeder 2010). So, in parallel, we might think that a joint decision that we φ is warranted just in case we have sufficient shared reason to φ: that is, there is no incompatible option ψ such that our shared reasons give us more reason to ψ than to φ. Recall here that shared reasons is a placeholder term: it refers to whatever body of reasons it is on the basis of which we ought to evaluate joint decisions. This makes the proposal even more plausible: if our shared reasons are just the reasons we ought to use to evaluate joint decisions, then isn t it obvious that a joint decision is warranted when it is sufficiently supported by the shared reasons? The problem with this simple proposal is that it opens up the possibility of conflict between joint and individual practical deliberation. Our shared reasons might not always be identical with each of our individual all-things-considered reasons, simply because my reasons might be 13
different from yours. So, there is no guarantee that our shared reasons will always support the same actions that each of our individual reasons do. If we assess joint decisions just on the basis of how well they are supported by our shared reasons, then, we may reach a warranted joint decision that conflicts with what one of us has individual, all-things-considered reason to do. If such conflicts are possible, they cast doubt on the rationality of joint practical deliberation. An agent faced with such a conflict would be forced to choose between acting on the warranted joint decision and respecting her all-things-considered reasons for action. If she conforms to the joint decision, she will be acting irrationally, since by hypothesis her all-thingsconsidered reasons do not support doing so. But if she acts on her all-things-considered reasons, it s unclear why she should have bothered to participate in joint deliberation, since she is now ignoring the joint decision and acting on her own judgment. So, the agent will either be irrational for participating in joint deliberation at all, or for acting on its results. If participation in joint practical deliberation can be individually rational, then, we must have some way of avoiding such conflicts. We can do so by building a conflict-blocking constraint into the standard of correctness for joint decisions. I propose the following: SUFFICIENT REASON: A joint decision J is warranted only if (a) each participant has sufficient all-things-considered reason to act in accordance with J (given that the other participants do so as well), and (b) each participant s sufficient reason to act in accordance with J is based upon the shared reasons of the process of joint practical deliberation within which J was made. The first clause of SUFFICIENT REASON is meant to block the possibility of joint-individual conflict: for a joint decision to be warranted, any individual participant in joint deliberation must have sufficient reason, all things considered, to act in accordance with that decision. This captures the fact that but I shouldn t φ! is always, if true, a decisive objection to a joint decision that requires one to φ. If each participant has sufficient reason to act in accordance with the joint 14
decision, as SUFFICIENT REASON requires, then this joint decision will not require anyone to do something they individually ought not do (given their reasons at the time the decision is made). 15 The second clause captures the idea that joint decisions must be based on a body of reasons that is shared by the deliberative participants. When we are deliberating together about what to do, it seems that there is a shared set of considerations that we are taking to be relevant to our decision. For our joint decision to be warranted, it must be supported by these shared reasons. This was the simple idea with which we began. SUFFICIENT REASON augments this idea by requiring not only that a warranted joint decision be supported by the shared reasons considered alone, but also that those shared reasons provide the considerations on the basis of which each agent is individually justified in acting upon the joint decision. SUFFICIENT REASON has an important upshot for the content of shared reasons: all of the shared reasons must be reasons for each of the deliberative participants individually. To see this, suppose that the fact that p is not a reason for S to φ. By SUFFICIENT REASON, the joint decision that S will φ is warranted only if the shared reasons in support of it are the basis of S s sufficient reason to φ. Since the fact that p does not give S reason to φ, it cannot be among the reasons that give S sufficient reason to φ, and thus cannot contribute to the warrant of the joint decision that S will φ. Therefore, the fact that p can be a shared reason in support of the joint decision that S will φ only if it is a reason in support of φing for S individually. 16 15 To be clear: SUFFICIENT REASON states that the warrant of a joint decision depends on the reasons of the individual participants, not the other way around. It thus does not imply that the fact that we have made a warranted joint decision that I will φ gives me any further reason to φ. For a joint decision to be warranted, it must be true that I already had sufficient reason to act in accordance with it; but this does not mean that the decision itself gives me reason to do so. Whether joint decisions do provide additional reasons in this way is a further question (see 5). 16 Some might worry that this requirement is too strong. Can t an admissions committee decide to admit a student because she is a legacy, even though not all the members of the committee agree that this consideration is relevant (Bratman 2014: 134)? Can t a democratic legislature make joint decisions in spite of the plurality of incompatible priorities and values represented within it (List 2006)? Despite the apparent tension, I think these kinds of cases are compatible with the SUFFICIENT REASON constraint. The key is to see the procedure by means of which we compromise between our disparate views as itself based upon the unanimously shared reason we have to endorse decisions made by this procedure. For example, the admission committee s joint decision to give legacy considerations the weight they do is itself justified by the committee members unanimously shared reason to settle on a standard procedure for evaluating applicants. I hope to explore this idea further in future work. 15
I do not claim that this is the only condition that must be met for a body of reasons to count as shared. There are likely to be further constraints: perhaps the shared reasons must be common knowledge, or perhaps they must be connected to the participants joint projects in the right way. But for our purposes, we can leave it open what these further constraints might be. It is important to be clear about the normative significance of SUFFICIENT REASON. This standard is a rule of the practice of joint practical deliberation: qua participant in joint deliberation, one is required to propose and accept only joint decisions that meet this requirement. Insofar as we have reason to perform joint deliberation well, we have reason to conform to this rule. However, as with other practices, we may have all-things-considered reason to violate the rules of joint practical deliberation, if practice-external considerations are at stake. I might be justified in proposing a plot to assassinate the president, if doing so will convince my captors that I have converted to their cause, though the joint decision I am proposing is clearly unwarranted. In this way, the standard of correctness for joint decisions is similar to the norm of assertion: qua assertors, we are required to obey this norm, but we may have practice-external reason to violate it in particular cases (e.g., by lying). 2.3. Two ways of making joint decisions One major question remains: how do we go about making a joint decision? Our joint decision-making procedure must meet two constraints. First, the process of making joint decisions must be guided and regulated by the standard of correctness for joint decisions. In order to count as engaged in individual practical deliberation, an agent must be aiming to make a decision that is warranted by her reasons. Similarly, for us to count as engaged in joint practical deliberation, we must be aiming to make a joint decision that meets the requirements of SUFFICIENT REASON. Thus any account of joint practical deliberation must explain how this process enables us to approach and (sometimes) attain this goal. The second constraint is simply that the process of 16
making a joint decision must be a joint activity. This means at minimum that it must involve all the participants playing some part. There are two distinct methods of joint deliberation that meet these two criteria, which I call propose-and-challenge and propose-and-ratify. Both methods proceed in two steps. First, a speaker proposes a potential joint decision that she takes to be warranted. Second, the addressees evaluate the proposed joint decision, judging whether it is warranted, and then signal their acceptance or rejection of the proposal on that basis. A proposed joint decision comes into force just in case all participants signal their acceptance of it. Where the two methods differ is in how the addressees go about signaling their acceptance or rejection of the proposal. In propose-and-challenge, an addressee is expected to challenge a proposed joint decision if she thinks it is unwarranted. When a challenge has been raised, it is open for counter-challenge: perhaps the speaker can show that the challenge does not in fact show the proposed decision to be unwarranted. Such counter-challenges can themselves be challenged, and so on. A proposed joint decision counts as accepted just in case there have been no successful challenges to it (meaning that either no challenges have been raised to the proposal, or all challenges to it have been successfully counter-challenged). The absence of successful challenges signals that each individual takes the shared reasons to give her sufficient reason to accept the proposed joint decision, indicating (defeasibly) that it meets the standard of correctness. Thus propose-andchallenge meets our first constraint. It meets the second as well: each participant plays a role in the propose-and-challenge process by being prepared to raise any challenges that come to her mind. Even when a joint decision is proposed by one participant and then accepted with no further words, it is made jointly by all participants, since the other participants silent acceptance signals that the proposal has passed a collective process of evaluation. 17 17 In the right context, however, even silence could count as a challenge (e.g., if combined with a pointed stare). What matters is not the difference between being silent and uttering words per se, but the difference between performing a communicative act that counts as raising a challenge, and performing no such act. 17
In propose-and-ratify, the addressees are expected to ratify a proposed joint decision by explicitly accepting it if they think it is warranted. If one of the addressees fails to accept the proposal, then it counts as rejected. Even more clearly, propose-and-ratify is a collective activity of evaluation, since a joint decision is only made after every participant has explicitly affirmed it. One way of characterizing the difference between these methods is in terms of their default results. In propose-and-challenge, the default result is to accept the speaker s proposal: if the addressees raise no challenges, the proposed joint decision counts as accepted. In proposeand-ratify, the default result is to reject the speaker s proposal: if the addressees do not explicitly signal their acceptance, the proposed joint decision counts as rejected. Just as we can make a new joint decision by means of propose-and-challenge or proposeand-ratify, we can also retract an already standing joint decision by means of these methods. When a speaker proposes to retract a certain standing decision, the addressees will assess whether the proposal is warranted, and the decision is retracted just in case all participants signal their acceptance. When a joint decision is retracted by this method, the shared options are expanded to include some of the possible combinations of actions that had previously been ruled out. 18 This concludes our theory of joint practical deliberation. Though there is much more to be said on the topic, I have intentionally included only those details that are necessary to generate the deliberative theory s central predictions regarding promises. The claims I have made all follow from three basic ideas: (i) joint practical deliberation has the same structure as individual practical deliberation; (ii) it must be possible for individual agents to participate in joint practical deliberation rationally; and (iii) joint practical deliberation must be a joint activity in which all participants play an active role. Thus we can now cash in on the promissory note of 1: the rules of joint practical deliberation are not arbitrary; they are the unique result of the constraints imposed by the practice s function. 18 The question of which actions are added back in to the shared options is an instance of the problem of permission (Lewis 1979b; Yablo 2009). I won t try to address this notoriously difficult question here. 18
2.4. From joint practical deliberation to ordinary speech acts The opening move in any process of joint practical deliberation is a proposal that puts forward a potential joint decision for consideration. So, if we are in fact able to deliberate together, we must have some way of proposing joint decisions. How do we do this? A simple hypothesis is that we propose joint decisions just by uttering sentences with their contents: I propose the joint decision that we will φ whenever I say, we will φ. But this can t be right: in saying, e.g., we will go to Vegas together someday, I could just be making a prediction, rather than proposing that we decide to go to Vegas. So, merely uttering a sentence with the content of a potential joint decision is not enough to propose a joint decision with that content. To capture the distinction between predictions and proposed joint decisions, then, we must posit at least one distinctive kind of speech act that has the force of proposing a joint decision. I say at least one, because there may be further distinctions to be drawn between different speech acts that aim to propose joint decisions. I submit that our ordinary categories distinguish between such proposals on the basis of their content and their intended method of evaluation. First, we distinguish between proposals to make joint decisions concerning only the speaker s actions ( I will φ ), those concerning only the addressee s actions ( you will φ ), and those concerning multiple participants actions ( we will φ ). Second, we distinguish between proposals meant to be evaluated by propose-and-challenge and those meant to be evaluated by proposeand-ratify. I submit that our ordinary concepts map on to these distinctions as follows: Propose-andchallenge Propose-andratify I will φ Promises Offers You will φ Demands Requests We will φ Agreements 19
Promises are proposals to make joint decisions of the form I will φ by means of the propose-and-challenge method. They are distinguished from demands, requests, and agreements by their content: promises concern the speaker s actions, while demands and requests concern the addressee s actions, and agreements concern the actions of the speaker and the addressee. 19 Promises are distinguished from offers by their intended method of evaluation: promises are meant to be evaluated by means of the propose-and-challenge method, while offers are meant to be evaluated by means of the propose-and-ratify method. This captures the fact that offers need to be explicitly accepted in order to bind the speaker to perform, while promises do not. If I offer to drive you home, and you do not explicitly accept my offer, then I am under no obligation to drive you home. If I promise to drive you home, on the other hand, then you need not say anything for my obligation to come into effect: so long as you do not object, I am bound (cf. Robins 1984: 101; Darwall 2011: 273). The table presented above outstrips the ambitions of this paper: I cannot defend here the claim that all five of these phenomena are best understood in terms of joint practical deliberation. Here I will focus on the proposed analysis of promises, which we can now state in full: The Deliberative Theory of Promises: For S to promise A that she will φ just is for S to propose to A that they make a joint decision, by means of the proposeand-challenge method, to the effect that she will φ. Now that we have a complete statement of our hypothesis, we can begin to evaluate it. What predictions does the deliberative theory make about the nature of promises? Do these predictions capture and explain our intuitive judgments? We now turn to these questions. 19 Agreements have two odd features. First, unlike our other terms, agreement is a success term. While it is perfectly possible to have a rejected promise, the idea of a rejected agreement seems incoherent. If the parties didn t agree to it, it wasn t an agreement. So while our other terms refer to proposed joint decisions, the term agreement only refers to accepted joint decisions. One upshot of this is that agreements are not themselves speech acts; proposals to make agreements are speech acts. Second, we don t have words that distinguish between the propose-and-challenge and propose-and-ratify methods of making agreements. But the distinction still applies: consider the difference between Let s go on a walk! and Would you like to go on a walk? The former proposal is meant to be evaluated by the propose-and-challenge method, while the latter is meant to be evaluated by propose-and-ratify. 20