1. Shared Intentionality as a Theme of Modern Philosophy

One way of telling the story of modernity, as a social or political phenomenon, is that it involved a dawning recognition of the power of shared or collective intentional agency as a means of shaping who we are and how we see the world. As a tour through modern philosophy, this telling might include discussion of Rousseau's conception of the General Will (see Miller and Tuomela Reference Miller, Tuomela, Chant, Hindriks and Preyer2014; Tuomela Reference Toumela2013: 94–96), Kant's understanding of the categorical imperative as the demand to constitute a universal kingdom of ends (Reference Kant and Gregor1996: 83–85), Hegel's claim that ‘the experience of what Spirit is’ arises as the recognition of an ‘“I” that is “we” and [a] “we” that is “I”’ (Reference Hegel and Miller1977: 110), Peirce's view that the identities of individual persons have their ground in the universal community of inquiry (see Burgh, Field, and Freakley Reference Burgh, Field and Freakley2006: chapter 2), and Wilfrid Sellars's effort, which he understood as working out ideas in Kant and Peirce, to characterize the moral frame of mind as a species of shared intentionality (for example, Reference Sellars1951, 1Reference Sellars, Casteñeda and Nakhnikian963, [1968] Reference Sellars1992, Reference Sellars, Brand and Walton1976). Prominent contemporary examinations of the ground of social reality in shared intentionality include Margaret Gilbert (Reference Gilbert2014), John Searle (Reference Searle1995, Reference Searle2010), Raimo Tuomela (Reference Tuomela2007, Reference Toumela2013), and the essays in Frederick Schmitt (Reference Schmitt2003a), Michael Schmitz, Beatrice Kowbow, and Hans Bernhard Schmid (Reference Schmitz, Kobow and Schmid2013), and part six of Marija Jankovic and Kirk Ludwig (Reference Jankovic and Ludwig2018). Because of increased interest over shared intentionality in recent Anglophone philosophy, and in related areas of evolutionary anthropology, developmental psychology, and the social sciences, a story of this sort offers the prospect of relating major trends in the modern period to contemporary debates about human thought and agency in the sciences and the humanities, by using developments in one field to shed light on work in others.

In the contemporary philosophical literature on shared intentionality, one of the recurring questions concerns whether and in what sense shared intentions can be reduced to individual intentions (for representative discussions see Birch Reference Birch2019; Bratman Reference Bratman2014: chapters 1 and 6; Ludwig Reference Ludwig2016: chapters 9–12; Schweikard and Schmid Reference Schweikard, Schmid and Zalta2020: section 3; Tuomela Reference Toumela2013: chapter 2). The effort to provide some such reduction is often connected to a felt need to avoid reifying collective minds as bearers of mental states. More or less overt ontological reification occurs across much of this literature. The full passage from Hegel partially quoted above, for instance, reads: ‘What still lies ahead for consciousness is the experience of what Spirit is—this absolute substance which is the unity of the independent self-consciousnesses which, in their opposition, enjoy perfect freedom and independence: “I” that is “we” and “we” that is “I”’.Footnote ¹

To disarm the apparent reference to collective minds that talk of shared or collective mental states seems to involve, philosophers often try to reduce any apparent mindedness in a collective to the mindedness of its members, trading the tendency toward what Schmitt (Reference Schmitt and Schmitt2003b) calls supraindividualism for a tendency toward individualism. Meanwhile, some philosophers hold that even if there are no irreducibly collective minds there are irreducibly collective mental states (see the discussion of we-mode intentionality in Tuomela Reference Tuomela2007, Reference Toumela2013; Miller Reference Miller2014). The bearers of these mental states may be individuals, but as mental states they are essentially collective (views that tolerate more supraindividualist perspectives include Gilbert Reference Gilbert1989, Reference Gilbert2000; List and Pettit Reference List and Pettit2011; Pettit Reference Pettit and Schmitt2003; Schmid Reference Schmid2018; Schmitt Reference Schmitt and Schmitt2003b; Searle Reference Searle1995).

Despite the variety in their views, two commitments common across much of this literature contribute to worries about supraindividualism. First, philosophers tend to model theories of shared intentionality on theories of individual intentionality: starting with an antecedent understanding of the latter, the capacities thought operative there are repurposed to account for shared intentions (for example, Bratman Reference Bratman2014: 14, 128–30; Ludwig Reference Ludwig, Jankovic and Ludwig2018: 477). Sara Rachel Chant (Reference Chant, Jankovic and Ludwig2018: 14) calls this the ‘wash, rinse, and repeat’ approach: ‘Employing WRR [wash, rinse, and repeat] as a method for analyzing group-level concepts such as collective intention is to imply a kind of isomorphism between theories of individual action and theories of collective action. The isomorphism is between the respective sets of individual- and group-level concepts (e.g. “individual agent” and “collective agent”), and the explanatory relationships between them. Taking on WRR is tantamount to assuming that if there is an explanatory relationship between two individual-level concepts, there should be a corresponding pair of collective-level concepts bearing the same explanatory relationship to each other’.

Alongside this first framing commitment, debates about shared minds have been shaped by representational analyses provided for sentences like ‘we built the house together’ and ‘the homeowner's association is responsible for the change in policy’. These sentences are thought to be true just in case certain conditions hold, and they are semantically evaluated as representations of these truth conditions (for example, Bratman Reference Bratman2014: chapter 4; Ludwig Reference Ludwig2007, Reference Ludwig, Chant, Hindriks and Preyer2014; Ludwig Reference Ludwig2016: chapters 12 and 13; Tuomela and Miller Reference Tuomela and Miller1988). We gain a rather different view of shared intentionality if we trade these two commitments for, in the first instance, an analogical rather than isomorphic understanding of shared intentionality, and in the second instance an analysis of the sentences that give expression to shared intentions rather than those that purport to report on or represent their existence.

The first shift might not seem that significant. Although Chant characterizes wash, rinse, and repeat as an isomorphic understanding of shared intentions on the basis of an understanding of individual intentions, it is not uncommon to see philosophers frame their accounts in terms of analogical relationships between the two domains (for example, Bratman Reference Bratman2014: 10, 30–31; Tuomela Reference Tuomela2007: 140). For this reason, I think we should understand Chant as diagnosing a tendency or trend in recent philosophy rather than an explicit or universally exemplified commitment. The important point, for both her and myself, is that some of the troublesome ontological puzzles philosophers have faced are artifacts of the modeling methods prevalent in the contemporary literature. In what follows, I argue that we can solve, or dissolve, some of these puzzles by thinking more explicitly about the analogical character of our models.

Work in the philosophy of science over the last half-century has equipped us with a set of protocols for the critical application of analogical reasoning in theory construction (see Bartha Reference Bartha2010; Gentner Reference Gentner1983; the essays in Gentner, Holyoak, and Nakinov Reference Gentner, Holyoak and Kokinov2001; Hesse Reference Hesse1952, Reference Hesse1966). In its general form, we use an antecedent understanding of one domain of inquiry (the analogical base) for framing an understanding of a different domain (the analogical target). After identifying hallmark features of the analogical base and target that are situated in similar networks of operation (such as causal, geometrical, or conceptual), we give a commentary on the sorts of relations in the base that can be inferred to hold among the correlate features in the target. While there must be some structural similarity between the analogical base and target if the analogy is to get off the ground—consider the sun at the center of the solar system and the nucleus at the center of the atom—much of the explanatory value of analogical reasoning derives from the commentary one gives concerning which inferences can be drawn from base to target. The solar system model of the atom concerns geometrical relationships, and in the commentary we block the inference that nuclei irradiate electromagnetic energy on the basis of gravity-induced nuclear fusion. Just so, an analogical understanding of shared intentionality may include a commentary sufficient to defuse it of the charge that it reifies the group into some kind of supraindividual mind. Metaphysical concerns may be avoided from the beginning, by simply not importing the relevant explanatory resources from base to target.

Analogical reasoning has the additional benefit of helping reframe our understanding of a subject matter that may be otherwise difficult to comprehend. Using an initial model drawn from the base, we investigate the analogical target to see whether the hypothesized relations posited from our reading of the base are present there. Over the course of the investigation, we update the model according to what we discover about the target domain. As a consequence of this revisionary process, an analogical conception of the relationship between two domains may reorient how we understand the very things whose metaphysical status we dispute. In one of its forms, analogical reasoning about two different domains allows us to construct a more general category that subsumes them both (as with general theories of teleological systems derived from Darwinian accounts of biology; for discussion, see Stovall Reference Stovall2015). This capacity to change our understanding of a target of inquiry led Hegel to associate inference by analogy with the syllogistic figure Peirce would label ‘hypothesis’ and subsequently assimilate to abduction, and both philosophers thought this form of reasoning was at the core of scientific and philosophical inquiry—for Hegel, it involved the ability to reason with universal terms as middle terms in a syllogistic inference, and so it constituted our grasp of universality as the most general kind of thinking (see Redding Reference Redding2003; Stovall Reference Stovall2020). They also thought this form of reasoning was going to be instrumental for the advance of intellectual and social projects rooted in the modern period, and at the end of the essay I suggest that we heirs of modernity can still profit from this way of thinking.

Concerning the second shift in commitment I take on here, recent debates have been unduly influenced by the thought that the mind's core task is to represent the world. Much of what we do is not strictly representational but agency-involving and meant to change the world instead of describe it. Representational capacities are involved in action, of course, but that is not so say a claim like ‘we are going to the park this afternoon’ or ‘you should be more polite to your neighbors’ purports to represent a fact—it may instead be a rehearsal of an intention, or a call for another to do the same. This distinction has its correlate in the organism. Sensory and motor neural pathways are physiologically distinct structures, each carrying its own intentionality and direction of fit: cognitive operations of the former are successful or correct when the mind ‘fits’ or corresponds to the world, whereas the operations of the latter are successful when the world fits the mind. Cases of mind-world fitness may be conceived in exclusively representational terms with a tolerable degree of accuracy, but an adequate understanding of world-mind fitness requires the categories of practical rationality. And we can recognize this point without losing sight of the fact that we in some sense represent intentional facts when we say, for example, that we built the house together.

At any rate, the philosopher is not required to begin studying shared intentionality through a truth-conditional analysis of sentences that appear to report on the existence of shared intentions. Problems of the apparent existence of superminds are more pronounced if we orient ourselves along the representational and truth-conditional dimensions of linguistic activity. Instead, if we are interested in shared intentional states of mind we should pay attention to the sentences used to give voice to shared intentions. Consequently, rather than propose truth conditions for sentences like ‘we built the house together’ I will offer an analysis of the semantic content of sentences like ‘we shall build the house together’, where the modal term ‘shall’ connotes the expression of an intention rather than a claim about its existence. This is a shift from a truth-conditional mind-world analysis of shared intentions to a plan-conditional world-mind analysis. More specifically, I propose we understand shared intentional mental states by analogy with the mental states associated with deontic judgments giving voice to what people ought, may, and ought not do, and I will use a planning semantics to understand both the intentional and the deontic modalities. This analogy is motivated by the observation that practical rationality is operative as a hallmark feature of both kinds of cognition, and my claim is that these frames of mind alike involve planning on how one would behave were one any member of the relevant community.

2. Background and OutlineFootnote ²

I begin by addressing two sets of worries one might have at the outset. First, it might be thought that the use of logical tools as a basis for investigating the nature of cognition was the resuscitation of a kind of metaphysical speculation that had long ago died out, to the benefit of our intellectual and scientific ecosystem. Here I emphasize two things and encourage the reader to keep them in mind in what follows. First, the model presented below is probative and subject to revision. I do not supply the final answers on what I discuss, and I try to use the formalism as a tool to help clarify what seem like otherwise plausible views about human cognition. Additionally, it should be remembered that the empirical study of human cognition is at least as relevant a source of model construction as the formal details that motivate my commitments here, and I strive to keep the discussion consonant with and informed by empirical work (more on this below). As a second general worry, one might think that in an essay of this length the treatments of these ideas will be too quick and superficial to carry the full weight of the claims placed on them. Here I concede the point while appealing to material that develops the view elsewhere.Footnote ³ And because this essay is more or less a philosophical investigation into the relationship between intentionality and deontic cognition, the social-scientific and evolutionary background is kept out of view entirely. In part to allay worries that my approach here is either too formalist or too superficial, then, it may do to say something about that background now.

In the last three decades, the empirical study of shared intentionality has developed into a burgeoning field of research across the behavioral, psychological, social, and evolutionary sciences. This work has begun to frame a picture of ourselves as paradigmatically norm-tracking, norm-enforcing, and collectively intentional animals (for overviews, see Schmidt and Rakoczy Reference Schmidt, Rakoczy, Newen, Bruin and Gallagher2018; Tomasello et al. Reference Tomasello, Carpenter, Call, Behne and Moll2005). While there is some evidence for coordinated activity that might be thought of as shared intentionality in nonhuman animals, from an early age human beings exhibit a range of shared and norm-enforcing behaviors unlike what is generally seen in even our closest primate relatives. Before their first birthday, human infants appear to understand human-like movements as goal directed, and around their first birthday they begin to share attention and emotions with other humans (Tomasello et al. Reference Tomasello, Carpenter, Call, Behne and Moll2005: 688–89); between twelve and eighteen months, human infants show a tendency to engage in both instrumental and playful cooperation (Schmidt and Rakoczy Reference Schmidt, Rakoczy, Newen, Bruin and Gallagher2018: 698); by the age of three, children distinguish a range of features of the normative landscape, including the distinction between conventions and moral norms (where the former depend on the agreement of the community; see Chudek and Henrich Reference Chudek and Henrich2011); by this age they also recognize that the entitlement to do something induces commitments of noninterference by others (Schmidt, Rakoczy, and Tomasello Reference Schmidt, Rakoczy and Tomasello2013), and they enforce norms on others (Rakoczy, Warneken, and Tomasello Reference Rakoczy, Warneken and Tomasello2008); between the ages of three and five, children gain the ability to cooperate better in a shared activity by swapping roles with other people, an ability that chimpanzees appear not to possess (Fletcher, Warneken, and Tomasello Reference Fletcher, Warneken and Tomasello2012); and by the age of five they spontaneously create and enforce norms when interacting with others (Göckeritz, Schmidt, and Tomasello Reference Göckeritz, Schmidt and Tomasello2014). So it appears human cognition is intimately connected to both deontic judgment and shared intentionality. From a physiological standpoint, there is good reason to treat these forms of cognition as involving multiperspective planning states. In the 1990s, the motor-representation complex of the central nervous system in macaque monkeys was shown to be operative both when the monkey performs various activities and when the monkey observes others perform them (for an overview, see Rizzolatti and Sinigaglia Reference Rizzolatti and Sinigaglia2010). It is as if the observer monkey was putting itself into the position of the observed and preparing to act from that point of view. Since then, we have discovered that this system is present across much of the animal kingdom. Work on the mirror system went through a period of intense popularization in the philosophy of mind and the philosophy of science during the first decade of the third millennium, and it has occasionally been put to use in philosophical accounts of shared intentionality as well (Butterfill Reference Butterfill and Misselhorn2015, Reference Butterfill, Jankovic and Ludwig2018; Tuomela Reference Tuomela2007: 69–74, 268n35, 297n26).

The formalism I defend is informed by the view that norm psychology and shared intentionality are characteristic features of human cognition, and that shared intentionality is partially realized, in the wetware of the brain, in patterns of multiperspective neural coordination across a community. I say ‘partially’ because mirroring in the motor-representation complex is at best a (naturally) necessary rather than a sufficient condition for some kinds of shared intentionality: it was perhaps a condition on its evolution in human beings, not something that must be present in every case of shared intentionality. For many animals exhibit motor-representational mirroring, but shared intentionality appears to be very rare in the animal kingdom—so mirroring is not sufficient for shared intentionality. And it is not necessary, as we can settle a business decision by conference call or decide as a nation to go to war with very little in common across our motor-representational complexes. Instead, it seems as though the coordinated neural mirroring that characterizes much of the animal kingdom interacts, in human ontogeny, with a suite of more cognitively complicated capacities peculiar to human beings, resulting in our capacity to adopt the shared affective, cognitive, and practical stances that characterize so much of human life (see Butterfill Reference Butterfill, Jankovic and Ludwig2018). I suspect that shared intentionality in human beings is best conceived as the product of a process of evolutionary and educational development that includes not only neural mirroring but also the human proclivity to share attention, emotion, and a range of other affective and motivational states.

While I leave this discussion behind in what follows, the central hypothesis suggested by the analogical reasoning below is meant to contribute to the empirical study of shared intentionality and norm psychology in human beings. It does so by proposing one sense in which the form of cognition associated with the latter is more sophisticated than that of the former. Toward that end, my terminology and framework is Michael Tomasello's (Reference Tomasello2014). Tomasello argues for a two-stage process of evolution from early hominids to modern humans: a stage of joint intentionality where dyads and small groups could share intentionality over spatio-temporally proximate kinetic behavior, and a stage of collective intentionality where an expanded capacity to share intentionality allowed for the sense of oneself as the member of an abstract community. At this second stage, Tomasello contends, we have the (in principle) universal sense of community that conditions our understanding of the true and the moral as that which constrains the thought and action of everyone. Tomasello notes that there may be additional stages between these two transition points from brute to civilization (Reference Tomasello2014: 151), and I can be read as positioning an intermediate stage in Tomasello's framework, between the development of collective intentionality and the capacity to think about what is true and moral. I do so by using the analogical reasoning below as a basis for reexamining some of the presuppositions of recent philosophical work on shared intentionality, and I defend the following hypothesis: the intentional state of mind is a more physiologically primitive form of practical rationality than that which is characteristic of the deontic state of mind, so that intentional mental states are prior in the order of being to deontic mental states. More specifically, I show that the need to distinguish the strong and weak deontic modal forces (such as ought and may) requires appealing to the notion of a single-minded choice attitude, which involves bearing attitudes both toward what one chooses and toward choices that are incompatible with what one chooses. An animal able to think single-mindedly could then adopt an indifferent frame of mind as well, according to the notion of indifference defined below. And with the choice attitudes of single-mindedness and indifference it is possible to model the planning mental states associated with the strong and weak deontic modalities. The intentional modalities, by contrast, do not distinguish strong and weak modal forces, and so we can model the sentences giving voice to individual and shared intentions with planning states that make no appeal to the capacity to bear attitudes both toward what one does and toward choices incompatible with what one does. The conclusion I draw—adding detail to Tomasello's philosophical anthropology, and presented as a hypothesis for further examination—is that the transition from the joint intentionality of small groups to the truth- and morality-mongering intentionality of modern humans required the development of a more sophisticated kind of practical cognition (the single-minded thinking that comes with deontic reasoning) than is necessary for the exercise of shared intentionality, whether joint or collective. This fits remarkably well with a claim made in the closing paragraph of the review article by Schmidt and Rakoczy (Reference Schmidt, Rakoczy, Newen, Bruin and Gallagher2018: 698, emphasis added): ‘So, one picture that is worth being explored more systematically in future research is that while humans and other species, notably primates, share basic forms of individual intentionality (and the corresponding natural norms of correctness and success), uniquely human forms of norm psychology and uniquely human forms of shared intentionality develop in close tandem in early ontogeny, the former building on and growing out of the latter.’

At the end of the essay, I offer a second hypothesis suggested by this model: we can understand individual intentions on the basis of our understanding of shared intentions rather than vice versa, so that while shared intentional mental states might be prior in the order of being to deontic mental states, shared intentional mental states are prior in the order of understanding to individually intentional ones. In a sense to be explicated in below, original intentionality is the shared intentionality of a community. That this is a theme resonant with certain ideas in modern philosophy will remain mostly in the background.

3. Understanding the Intentional Modalities by Analogy with the Deontic Modalities

Though not equivalent in all contexts, I will indifferently speak of sentences under an interpretation, of utterances, assertions, judgments, or claims, of mental states associated with utterances, and of propositional commitments.

It is widely accepted that shared intentions, as a species of intention, are or involve plans of action (see, for example, Bratman Reference Bratman2014; for a discussion of Bratman's view consonant with the one developed here, see Butterfill Reference Butterfill and Misselhorn2015). Though it has not received much attention in the literature on shared intentionality, an expressivist planning semantics has been developed for the deontic modalities. Allan Gibbard (Reference Gibbard2003: chapters 3 and 4), Seth Yalcin (Reference Yalcin2012, Reference Yalcin, Fogal, Harris and Moss2018), and I (Stovall Reference Stovall, Koreň, Schmid, Stovall and Townsend2021b, Reference Stovallforthcoming a, Reference Stovallforthcoming b) offer the most precise formulations of this world-mind semantics, but it has its roots in a body of research Wilfrid Sellars began developing in the 1950s (such as Reference Sellars1951, Reference Sellars, Casteñeda and Nakhnikian1963, Reference Sellars1966a, Reference Sellars and Lehrer1966b, Reference Sellars1967, [1968] Reference Sellars1992: chapter 7, Reference Sellars, Brand and Walton1976). The assertion of sentences falling under these modalities are interpreted as giving expression to plans of action concerning what one would do were one any member of a relevant group. In this regard deontic modality marks a capacity to put ourselves into different points of view within some community: the claim that cooks ought to clean their knives when dirty expresses commitment to the plan to clean one's knife in any situation in which one is a cook whose knife is dirty. Just so, features of the world that constrain or are otherwise conditions for the satisfaction of these plans can be accounted for in terms of possible worlds. The claim that my knife is dirty and I ought to wash it at once represents the world as being a particular way, and expresses commitment to a plan to do something.

In order to account for the difference between the strong deontic modalities (‘ought’ and ‘forbidden’) and the weak (‘may’ and ‘may not’), we must distinguish two ways of choosing: single-mindedly, whereby one rejects all choices incompatible with the choice one makes, and indifferently, whereby the agent is capable of choosing single-mindedly and there is at least one other choice incompatible with the choice made which could have been made without changing any of the agent's single-minded choices (the capability to choose single-mindedly is a condition ensuring that nonhuman animals do not trivially count as choosing indifferently in the relevant sense). In choosing indifferently, one has rejected rejecting the choices one could have taken without changing one's single-minded choices. A version of this distinction appears to have been independently developed at least three times over the last half century, and it has not received the attention it deserves. Sellars marks it in at least two places (Reference Sellars and Lehrer1966b: 113–15; Reference Sellars, Brand and Walton1976: 57), though he does not discuss it in detail. Gibbard (Reference Gibbard2003: chapter 3) then proposes it, and James Dreier (Reference Dreier and Shafer-Landau2006, Reference Dreier2009) advances a similar idea. Jeremy Koons (Reference Koons2019: 135) briefly discusses Sellars's account of indifference, and Alex Silk (Reference Silk2015: 53n8) notes that many of Gibbard's critics, by ignoring his discussion of this distinction, have not done justice to his view. Sellars, Gibbard, and Dreier each speak of preference rather than single-mindedness as the contrast with indifference, but as any choice might be thought to exhibit one's preferences—and as the relevant notion is the doubled-stance one adopts both toward what one chooses and toward choices incompatible with what one chooses—I prefer to speak in terms of single-mindedness (furthermore, the definition of indifference in terms of single-mindedness is my own; and see Stovall [forthcoming b: chapter 3], for an examination of the physiological ground of single-mindedness in the exercise of what Strawson Reference Strawson1962 calls the ‘reactive attitudes’).

The distinction between these two ways of choosing, and its role in a planning semantics for deontic judgments, can be appreciated through an example. Suppose I am a cook who thinks I ought to wash my knife, that I may wash it either by hand or in the dishwasher, and that I ought not wash it with turpentine. I am thereby in a state of mind of planning on single-mindedly choosing to wash my knife (rejecting not washing it), single-mindedly choosing not to wash it with turpentine (rejecting washing it with turpentine), and indifferently choosing between washing it by hand or with the dishwasher (rejecting rejecting each of these choices). Even where I have a general policy of washing knives by hand, if I regard both that action and use of the dishwasher as permitted, then I am planning on indifferently choosing to do the former and not the latter, in the sense introduced above: for I could just as well have chosen the latter without changing any of my single-minded choices. On the other hand, my choice to wash the knife by some means or other, as well as my choice not to use turpentine, are choices I undertake single-mindedly: for I could not make some choice incompatible with either of these choices while remaining of the same mind. To choose indifferently among permitted actions, each of which satisfies some single-minded choice, is to discriminate practical species or determinations of a practical genus or determinable: for example, washing a knife by hand and with the dishwasher are practical species of the practical genus of knife washing. And cognitive grasp of that relationship is reflected in the practical capacity to distinguish the corresponding single-minded and indifferent choices.

A semantics making use of these planning states offers a rigorous means for explicating the idea that descriptive sentences play a world-representing role while prescriptive sentences play an action-guiding or expressive role. Consider a language ${\cal L}$ composed of descriptive sentences, prescriptive sentences, and constructions from these and the Boolean operators (in the interest of readability I generally ignore use/mention distinctions below). Let a possible world w be defined as a maximally determinate state of affairs. That is, for every atomic descriptive sentence d _atom ∈ ${\cal L}$ and every w, either d _atom is true at w or ~d _atom is true at w (I will also speak of worlds as collections of facts or true propositions). Let a deontic hyperplan h_D be defined as a maximally consistent plan of action such that, for every circumstance C, every agent α able to make a choice at C, and every action A that α is able to choose to perform at C, either (exclusively):

1. 1. α single-mindedly chooses to A in C on h_D; or

2. 2. α single-mindedly chooses not to A in C on h_D; or

3. 3. α indifferently chooses to A in C on h_D; or

4. 4. α indifferently chooses not to A in C on h_D

Let a deontic hyperstate 〈w, h_D〉 be an ordered pair of a possible world and a deontic hyperplan (I will occasionally speak simply in terms of worlds and plans). With standard definitions for the Boolean operators, any sentence of ${\cal L}$ can be uniformly interpreted in terms of these hyperstates (see Stovall [forthcoming a] for details).

This hyperstate semantics can be straightforwardly extended to account for individual and shared intentions. We do so by supplementing deontic hyperplans with intentional hyperplans (I explain the need to keep two sets of books on planning below). An intentional hyperplan h_I ∈ H_I is a maximally consistent plan of action such that, for every circumstance C, every agent α able to make a choice at C, and every action A that α is able to choose to perform at C, either (exclusively):

1. 1. α chooses to A in C on h_I; or

2. 2. α chooses not to A in C on h_I

Because the intentional modalities do not distinguish a strong and a weak modal force, the distinction between single-mindedly and indifferently choosing to do something does not need to be marked in the plans that model intentional states. In part 5, I argue that this feature of the semantics, together with the joint fact that (1) the ability to distinguish single-minded and indifferent choices underlies the ability to mark off genus/species relations in practical reasoning, and (2) single-minded and indifferent choice attitudes each require that an agent entertain attitudes toward classes of action that are incompatible with what one chooses, suggests that the capacity for shared intentionality is a more primitive form of the kind of practical rationality exhibited in deontic cognition.

A deontic-intentional hyperstate 〈w, h_D, h_I〉 ∈ S_DI consists of an ordered triple of a world, a deontic hyperplan, and an intentional hyperplan. Let s_DI range over elements of S_DI. Where the semantic interpretation of a sentence φ is denoted as ⟦φ⟧, we use deontic-intentional hyperstates to evaluate semantically (mental states associated with) the utterances of descriptive, prescriptive, and intentional sentences as follows. Descriptive sentences d are semantically evaluated as the worlds at which d is true:

$$\lobrk {{\it d} } \robrk = _{{\rm def}.}\{ { \langle {\rm w}, \;{\rm h}_{\rm D}, \;{\rm h}_{\rm I} \rangle \colon\; {\it d}\,{\rm is}\,{\rm true}\,{\rm at}\,{\rm w}} \} $$

Read this as ‘the semantic value of the sentence d (under an interpretation) is the set of deontic-intentional hyperstates such that d is true at the world of those hyperstates’. The mental states associated with utterances of the three kinds of prescriptive sentences are evaluated as follows:

$$\eqalign{& {\lobrk {{\rm people}\,{\rm are}\,{\rm obliged}\,{\rm to}\,A\,{\rm in}\,C} \robrk } = _{{\rm def}.\!\!\!} \,\{ \langle {\rm w}, \;{\rm h}_{{\rm D}}, \ {\rm h}_{\rm I} \rangle \colon \; {\rm for}\,{\rm every}\,\alpha , \;\,\cr&\alpha \,{\rm single}\hyphen{\rm mindedly}\,{\rm chooses}\,{\rm to}\,{\it A}\,{\rm in}\,{\it C}\,{\rm on}\,{\rm h}_{\rm D}\} $$

$$\eqalign{& {\lobrk {{\rm people}\,{\rm are}\,{\rm forbidden}\,{\rm to}\,A\,{\rm in}\,C} \robrk } = _{{\rm def}.\!\!\!}\, \{ \langle {\rm w}, \;{\rm h}_{{\rm D} },\ {\rm h}_{\rm I} \rangle \colon\; {\rm for}\,{\rm every}\,\alpha , \;\cr &\alpha \,{\rm single}\hyphen{\rm mindedly}\,{\rm chooses}\,{\rm not}\,{\rm to}\,{\it A}\,{\rm in}\,{\it C}\,{\rm on} \;{\rm h}_{\rm D}\} $$

$$\eqalign{&\lobrk { {{\rm people}\,\,{\rm are}\,\,{\rm permitted}\,{\rm to}\,A\,{\rm in}\,C} } \robrk = _{{\rm def}.\!\!\!}\, \{\langle {\rm w}, \;{\rm h}_{{\rm D} },\ {\rm h}_{\rm I} \rangle \colon\; {\rm for}\,{\rm every}\,\alpha , \;\cr &\alpha \,{\rm either}\,{\rm single}\hyphen{\rm mindedly}\,{\rm choose}s\,{\rm to}\,A\,{\rm in}\,C\,{\rm or}\,{\rm indifferently}\cr &\,{\rm chooses}\,{\rm whether}\,{\rm or}\,{\rm not}\,{\rm to}\,{\it A}\,{\rm in}\,{\it C}\,{\rm on} \;{\rm h}_{\rm D}\} $$

Because negation is interpreted as a complement operator on sets of hyperstates, the usual equivalences among what is obliged, what is permitted, and what is forbidden hold (see Stovall [forthcoming a] and Stovall [forthcoming b: appendix 3]).

To represent the contents of intentional sentences, I use ‘ι’ as a metalinguistic variable taking either the first person singular or plural pronoun as its value. Formally, the semantic value of ‘ι shall A in C’ is the set of deontic-intentional hyperstates where everyone who is one of the group determined by the pronoun chooses to A in C on the corresponding intentional hyperplans:

$$\lobrk {\iota } \,{\rm shall}\,A\,{\rm in}\,C\robrk = _{{\rm def}.\!\!\!}\, \{ { \langle{\rm w}, \;{\rm h}_{\rm D}, \;{\rm h}_{\rm I}\rangle \colon \;{\iota }\,{\rm choose}\,{\rm to}\,A\,{\rm in}\,C\,{\rm on} \;{\rm h}_{\rm I}} \} $$

Here a concern might be raised that this does not deliver the right result as an analysis of shared intentionality, for there are cases where two people each want to do A in C, and want the other to do so, and yet do not share a plan. For instance, you and I might each have a plan to paint a house with the other, yet one of us plans to deceive the other in some of the details. Here it seems we do not share the intention to paint the house together, or share it in the right way.

To avoid misclassifying such cases as instances where an intention is shared, in the literature on shared intentionality it is common to impose a requirement that the plans be executed in the right way. This may include Bratman's requirement that the plans ‘interlock’, Ludwig's requirement that each agent's plan include a specification of the plans of the other agents in the group, or Tuomela's requirement that we-mode shared intentions are acted on in part because of, or for the reason that, the intention is shared. In the interest of being as noncommittal about these debates as I can, my official view is that, in cases of shared intentionality, the adverb together is used in the specification of the action content of the plan (for example, for a shared action A the action should be understood as doing A together), and I assume that this stands as a placeholder for the conditions someone like Bratman, Ludwig, or Tuomela might impose. For present purposes, I also assume that whatever natural or scientific conditions satisfy togetherness in shared intentionality can serve here as well (see Butterfill Reference Butterfill and Misselhorn2015 and Reference Butterfill, Jankovic and Ludwig2018 on the use of entrainment, motor-representation mirroring, task co-representation, emergent coordination, collective goal-states and joint affordances as a basis for constructing a minimal theory of shared intentionality). As long as these conditions do not presuppose a capacity for planning single-mindedly, that assumption will not undercut the inferences about ontological priority made below.

We are now in a position to see how this plan-theoretic analysis of the deontic modalities supports an analogical understanding of shared intentionality. For this is a view of moral reasoning as a capacity to put ourselves into the perspectives of other people and to consider how we would behave were we in their position. That capacity may take effort to cultivate, as with the ability to enjoy the arts of a community we did not grow up in. But insofar as we possess this capacity, we represent our community as one whose members we identify with in the practical sense that we think about their actions as, after a fashion, actions that we ourselves take part in. The moral stance is the universalization of this transpersonal practical point of view, permuted through the category of single-mindedness. To put the point in Kantian terms, the universal kingdom of ends is one whose members settle on what morality requires of them independently of their person and place in space and time, and they pursue that kingdom single-mindedly. And on this semantics, the transpersonal point of view is characteristic of shared intentionality as well.

4. A Commentary on the Model

As with any analogy, it is important to note where the similarities between the analogical base and the analogical target break down. In this section, I provide some remarks on the single-minded/indifferent frame of mind and the merely intentional, I discuss some of the ways they differ, and I show how these differences are represented in the semantics.

As I am using the notions, it is possible to choose without choosing either single-mindedly or indifferently. For the single-minded choice attitude is one that involves adopting an attitude of rejection toward all choices incompatible with what one chooses, and this requirement is not in force when simply choosing to do something (similar remarks hold for indifference). This is an idealization, of course, as we are at best approximate or partial hyperplanners. Much of what we do is habitual and unreflective, and when we do reflect on our actions we are often motivated by all sorts of things that we either are not aware of, or are actively trying to ignore. Still, our planning mental states can be modeled as approximations to what perhaps only the angels are capable of. Because single-mindedness and indifference are needed to model the strong and the weak deontic modalities, and because the intentional modalities lack a strong and a weak modal force, one can satisfy a merely intentional plan without choosing either single-mindedly or indifferently. At the same time, one way of making a choice is to choose single-mindedly (again, similarly for indifference). It follows that one way of realizing one's intentions is to do what one thinks one ought. That is, every single-minded choice to A is a choice to A, and so such a choice satisfies the intentional plan to A. But not every choice to A is single-minded, and so the mere fact that one chooses to A does not imply that one satisfies one's deontic plans—we may say that the agent did the right thing, but not for the right reason (or in the right frame of mind). This leaves open that the action may nevertheless be praiseworthy, as doing the right thing is often difficult enough on its own. But there is special praise that comes with doing the right thing under the right attitude (think of the difference between the executive forced by a court of law to donate to charity, and one who does so in the interest of the charity).

The fact that choosing simpliciter can be satisfied without an agent adopting any consideration toward incompatible choices, whereas the single-minded and indifferent choice attitudes cannot, is one reason for distinguishing intentional hyperplans from deontic hyperplans in the hyperstates that model the contents of theoretical and practical cognition, and so specifying where the analogy between these forms of thought breaks down. Another reason is provided by the following consideration. As exercises of practical rationality, in general (and where there are no defeaters) one who utters ‘I ought to A in C’, just as one who utters ‘I shall A in C’, will be disposed to A in C. But the two states of mind differ on the strength of the connection to the performance of the action in question, and we sublunary beings are characterized by the possibility of akrasia. For this reason, ‘I ought to A in C and I shall not A in C’, though in some sense involving a kind of error in cognition, is not incoherent. This is a regrettable, but understandable, state of mind. The mental state associated with an assertion of this sentence is determined by the intersection of the two sets of deontic-intentional hyperstates determined by the two conjuncts, and owing to the fact that deontic and intentional sentences are evaluated relative to different sets of plans, there is a nontrivial semantic value for this mental state. If deontic and intentional cognition were alike evaluated according to the same sets of plans, however, this conjunction would be incoherent as there would be no set of hyperstates that model it. Nevertheless, there is a straightforward explanation for the practical irrationality of one who is in such a state: given the plans that constitute the contents of the separate conjuncts, there is no choice one can make that satisfies everything one has planned. Akrasia is therefore conceptually coherent, though practically irrational.

Note that this semantics makes no appeal to superminds in accounting for the contents of shared intentions. The only novel entities postulated are the plans that individual thinkers may have in mind. Concerning the subjects who have these plans in mind, we suppose that they have the capacity, individually, to put themselves into the perspectives of other people and plan from these different points of view. That we are able to collectively have a plan in mind is a function of the fact that (1) plans are individuated relative to different agents and (2) we are able to consider how we would behave were we in various situations, including situations in which we are other people. Thus, numerically identical planning contents can be shared across minds. But each point of evaluation at a plan is always the point of a single individual, so there is nothing in the semantics that commits us to anything but the mental states of individual people.

5. A Hypothesis about Shared Intentionality and the Deontic Frame of Mind

At the beginning of the essay, I noted that Hegel associated analogical inference with the same syllogistic figure that Peirce labeled ‘hypothesis’ and later assimilated to abduction. Both philosophers held that reasoning in this mode allows for changing the way we understand some target of inquiry, and they thought that a study of the operation of this reflective spontaneity (to think the idea in Kantian terms) was essential for future intellectual and social development. Just so, this analogical conception of the relationship between shared intentionality and deontic reasoning allows us to rethink our understanding of the relationship between individual and shared intentionality. In part 2, I suggested that some of the empirical work on shared intentionality and norm psychology supports my proposal; here I want to indicate how the empirical work might in turn be informed by this view. Despite not forcing our hand, the analysis does raise some intriguing hypotheses.

For instance, it falls out of my account that while moral reasoning and shared intentionality are of a common kind, the former involves a more fine-grained capacity for exercising practical rationality than is needed for the latter. To reason in a deontic frame of mind one must be able to choose single-mindedly, and the single-minded choice to do A differs from merely choosing to do A insofar as it requires that one also bear a negative attitude toward actions that are incompatible with so choosing. Simply choosing to do A, however, amounts to stumping for A, and that choice can be made without regard to incompatible alternatives.Footnote ⁴ This conclusion is reinforced by a feature of practical rationality noted above. The ability to discriminate the indifferent choice from the single-minded choice is the ability to mark off the practical species or determinations that instantiate a practical genus or determinable. That is to say our theoretically rational capacity to make the deontic judgments we do is conditioned by our practically rational capacity to mark off genus/species or determinable/determination relations in agency. For an agent can only comprehend the requirement to do the right thing insofar as she is aware of what counts as ‘the right thing’, and in most of the contexts we face there are many ways of acting so as to satisfy this requirement. The world being the multifaceted thing it is, then, one can only plan from a deontic point of view insofar as one has the capacity to distinguish practical genera and their species, and this is conditioned by our ability to adopt single-minded and indifferent practical stances.

Finally, it is worth pointing out that a language of intentionality can be used to specify positive and negative reinforcement behaviors sufficient to habituate the members of a community into norm obedience of the sort expressed with the deontic modalities. And this can be done without presupposing that the individuals whose behaviors are shaped by this process have the capacity to discriminate single-mindedness from indifference—indeed, even the ones instituting this behavior might not be capable of explicit deontic reasoning. Where ‘stickbeating’ and ‘flocktending’ refer respectively to negative and positive behavioral reinforcement, and are understood on supposition that the hominids who were undergoing this conditioning shared a form of life (an affective/cognitive/practical stance toward a common environment), we have the following:Footnote ⁵

• • acting on the intention to stickbeat doing A in C and flocktend not doing A in C is one way of training people to conform to the norm that A is forbidden in C

• • acting on the intention to stickbeat not doing A in C and flocktend doing A in C is one way of training people to conform to the norm that A is obliged in C

• • acting on the intention to neither stickbeat doing A in C nor flocktend not doing A in C is one way of training people to conform to the norm that A is permitted in C

It may be that the ability to recognize that something is obliged, permitted, or forbidden requires the ability to distinguish single-minded from indifferent choices. That is to say, self-consciousness may require the resources of deontic cognition. Nevertheless, the ability to conform to the norms governing what is obliged, permitted, and forbidden can be instituted within a community simply on the basis of behavior specifiable in terms of intentions unmarked by the distinction between single-mindedness and indifference. This again suggests that the merely intentional state of mind, whether individual or shared, is a more primitive form of cognition than that which attends deontic mental states. To return to Tomasello's philosophical anthropology, the hypothesis I am proposing—as a further specification of his account of the development of truth- and morality-tracking cognition from joint and then collective intentionality—is that a capacity for single-mindedness and indifference had to emerge between the evolution of collective intentionality and the appearance of discursive theoretical and practical rationality. And I am suggesting that, nevertheless, shared intentional habit-shaping acculturating practices (stickbeating and flocktending) could have driven the development of a selection pressure in favor of that capacity among human beings, who could thereby more effectively control the development of the norms that bind them together. This hypothesis has the added virtue of explaining what it is to be capable of exercising discursive cognition of the sort exemplified in the human use of natural language: one must be capable of the self-government that comes with choosing single-mindedly by binding oneself to a course of action in rejecting everything incompatible with it (I develop this line of thought more extensively in Stovall [forthcoming b]).

6. Understanding Individual Intentionality by Analogy with Shared Intentionality

A second hypothesis raised by this framework concerns the status of shared intentions vis-à-vis individual intentions. It is generally held that individually intentional mental states are more basic than shared intentional mental states, with philosophers scaling up and supplementing the former in the interest of explaining the latter. By shifting to an analysis of the states of mind expressed with the language of the intentional modalities, and understanding these on the basis of a planning semantics for the deontic modalities, we arrive at a different view on the relationship between shared and individual intentions. For from the standpoint arrived at here it is more parsimonious to understand individual intentionality by analogy with shared intentionality rather than vice versa. Starting with the notion of a shared intentional state as a collection of hyperstates that are general with regard to the agents whose choices are planned, the mental state constituting an individual intention can be understood as the localization of that general set of plans to a single agent. Rather than seeing the shared intentional state of mind as an expansion of the individually intentional state of mind, we can see the individually intentional state of mind as a contraction of the shared intentional one. On this analysis, the universal practical point of view is the conceptual foundation for the practical mental states that express senses of self that are localized to particular communities (such as cooks), with the individual point of view as the limit of this circumscription of the primordial practical stance.

This hypothesis can be understood as one way of spelling out the idea that the shared intention is the locus of original intentionality within a community of rational beings. Regarding metaphysics or ontology, the only minds in play are those of the individual members of different communities. This is to say that the individual intention may be prior to the shared intention in the order of being (just as the shared intention, I have argued, is prior in being to deontic cognition). But in the order of understanding, as reflected in a hyperstate semantics for statements that give expression to intentions, it proves simpler to suppose that shared intentionality is basic and to define the mental state of an individual intention as a special case of the former (just as a planning semantics originally developed for the deontic modalities was positioned above as prior in the order of understanding to a planning semantics for intentionality). This line of thinking provides some support for the supposition that modern philosophy's interest in Spirit or the General Will was a conceptual advance over pre-modern notions of individual and shared agency. And, if the first hypothesis is correct, then this analogical investigation into the metaphysics of practical rationality helped discover something new about the nature of Spirit: viz., that in the order of being the shared intention is prior to deontic cognition.Footnote ⁶