Shared Intentionality and Automatic Imitation: The case of La Ola

3.1. Intentionalism

As noted, usual action-theoretical responses to the problem of cooperation imply an account of shared propositional attitudes such as intentions (plus additional conditions that should be met, such as the common knowledge condition; cf. Bratman 2014; Paternotte 2017). This implies intentionalism. Generally, intentionalism can be understood as the metaphilosophical view that the language of intentionality typical for the folk-psychological conceptual apparatus—allowing one to talk about intentions, desires, beliefs and the like—builds the most convincing framework to anchor explanations of the nature of shared actions (cf. Makowski 2017, 37). Intentionalism is frequently an implicit view and it may take various forms, depending on explanatory needs and ambitions. For the purposes of this essay, we may distinguish two versions of it. A stronger causalist view posits causal relations between the explanantia (e.g., intentions, plans) and the explanandum (e.g., joint action). According to this view, intentional language must appear in the explanation of what is needed for the execution of a joint activity (i.e., in actu). So conceived intentionalism belongs to a realistic story about agents’ psychological processes. A weaker constitutivist view posits only constitutive relations between the explanantia (e.g., intentions or plans) and the explanandum (e.g., joint action) and it is rather a rational story about how agents understand what they do when they act together. In other words, constitutivism assumes that agents involved in joint action are capable of giving reasons for what they do (for more on this, see section 3.2.). Here, the role of the folk-psychological language is to provide the scaffolding for practical reasoning.

These two versions of intentionalism may go together. For example, influential accounts of collective intentionality based on the concept of shared intention are good instances of combining them: the idea that intentions prompt practical reasoning or that they can be cited in reason explanations is here supported by the assumption that intention formation plays its role in action causation (Bratman 2014; Pacherie 2012). ⁴ However frequent in the literature this kind of fusion is, I distinguish the two versions for heuristic reasons. This will allow us to focus especially on the cognitive underpinnings of joint action. Hence, the causalist version of intentionalism will be our default frame of reference.

Let us turn back to La Ola. According to the strong version of intentionalism, individuals who make up the wave must form shared cooperative intentions (plus additional relevant conditions). Otherwise, their micro-activities could not affect the macro-level joint action. Staying on this path may be tempting, but this view appears to be in tension with the initially accepted intuition that La Ola is automatic or even mechanical. It seems that the waving is possible even if most spectators do not even have to think about it—in the sense of planning or intending—to perform it. Even if we agree that most spectators, having the overall schema of La Ola in mind, understand what it is and are basically aware that it may happen in the stadium, most of them can simply enjoy watching the game, minimally pay attention to what their neighbors do, and do the same when needed: stand up and sit down. In this sense, it would be difficult to accept the view that La Ola is a fully intended or planned shared activity: it does not seem to be convincing to claim that all parties involved in the activity must deliberately form intentions to take part. The acceptance of this claim would have several unrealistic consequences. For example, it would lead to the view that all spectators have sufficient attentional and cognitive resources to form such intentions suitably to their place in the stadium and to the role they may play in such performances. Or that these resources are sufficiently refined to warrant effectiveness of such intentions for the smooth propagation of the wave.

Still, the picture of intentionality for La Ola appears to be too complex to flat-out assume that agents’ shared intentions build its essence. The intentionalistic explanation, if it makes sense in this case, must do justice to the fact that La Ola is not only largely automatic, but also that it requires some sort of emotional contagion. ⁵ In this respect, the reductive idea that individuals are “excitable media” appears to be an instructive metaphor and a signpost to modify the account. What this idea amounts to is yet to be demonstrated.

First off, it is plain to see that this reductive idea does not seem to be correct for any stage of the collective action exemplified by La Ola. Not all agents involved should be reduced to “excitable media.” Inasmuch as the very beginning of La Ola involves broadly conceived decisions, it allows for more standard intentionalistic accounts. That is, there are convincing reasons to stipulate that intentions and other propositional attitudes are formed by those who initiate the wave. All this suggest that an intentionalistic ontology of La Ola is complicated—its conditions appear to be different for different stages of the whole performance. Hence, following Vicsek and his colleagues, I distinguish two basic stages and groups involved in them: the perturbation group for the formation stage and the followers group for the propagation stage.

3.2. Formation Stage: Shared Intentionality of the Perturbation Group

La Ola is something that must be initiated (“excited”). So, even if it emerges spontaneously, the waving effect seems to require shared intentions and some sort of initial, more or less tacit agreement within a group of individuals. Those individual agents constitute the perturbation group which collectively intends or plans to form the wave. This is the formation stage that precedes, so to speak, the more populous and more automatic phase of the whole performance.

To describe the nature of the collective intentions required at this stage, we could refer to classic (intentionalistic) views, for example, “we-intentions” (Searle 1990), joint intentions in the “we-mode” (Tuomela 2006), or “intermeshing intentions” (Bratman 1992, 2014), characterizing them with respect to the aforementioned causalism/constitutivism divide (section 3.1.). One might also propose another account based on the idea that explanations of shared intentionality require the formation of propositional attitudes such as joint collective intentions. Here I resist such temptations. Instead, I would like to again accept the causalist version of intentionalism as a broad framework and propose a suitable intentionalistic modification of the idea that humans are “particles.” This can be done by introducing the (reductive) idea that the initial concerted effort to perform La Ola in the formation stage is possible because of shared interdependent intentions of individual initiators plus their additional psychological conditions: similar thresholds of readiness and enthusiasm (“excitation”), agreement and common basic beliefs about what and how it will be performed. Let us call them shared triggering intentions. These intentions can be ascribed to two cases of strongly cohesive groups who initiate La Ola: to spectators who intend to perform La Ola entirely spontaneously and to the group who first deliberately preplanned the activity (for example, ultra-fans or youth sport teams). In both cases, the idea of shared interdependent intentions may resemble, to some extent, the Bratmanian interdependent (“intermeshing”) intentions (cf. Bratman 1992, 2014), but there are two significant differences.

First, we should notice that understanding La Ola as a shared preplanned activity implies an important condition. According to video-aided observations, La Ola spreads when spectators are neither too bored nor too excited (Farkas et al. 2003, 19), so they have to be in an appropriate mood and share an appropriately moderate degree of emotional excitement. Similar things should be said of the perturbation group: their psychology implies “excitability”—a readiness to implement the preshared plan (as in the case of a deliberate group of initiators) or—to impulsively co-intend the activity (as in the case of spontaneous initiators). In both cases, shared emotions add a significant necessary condition for the intentionality of La Ola. ⁶ Whether shared triggering intentions are born spontaneously or as a result of preplanning, the condition of sufficient emotional contagion within the perturbation group means that they cannot be mechanically translated into the language of deliberate plans (a typical strategy for the Bratmanian theory).

Second, notice also that if La Ola emerges spontaneously (not as a result of preplanning), then the idea of a shared intention to form up the wave could be understood as a full-fledged plan only ex post. That would imply a constitutivist view: most individuals who spontaneously initiate the wave could respond to the Anscombean question of “what are you doing?” (Railton 2009), but their answers—it seems—would rather be post-factum rationalizations of their behavior than reports of conscious intention formation before starting the activity that created the “perturbation” during the initial phase (cf. Makowski 2017, 177). This explanatory strategy appears to be apt for those cases of intentional action that are relatively quick and do not require much thinking, but—as it is silent about their cognitive underpinnings—its usefulness is significantly limited.

Thus, we must pay attention to whether triggering interdependent intentions that provide initial coordination are spontaneous or deliberate. In the first case, they need not anchor an initial plan, but they must be sufficiently mindful to link the mental schema of the wave with the visual and emotional information about the situation to guide and control micro-activities that build up the wave. ⁷ To use popular distinctions from philosophy of action: one might say that shared triggering intentions are rather present-directed (Bratman 1987) or proximal (Mele 1992) than future-directed or distal, and share their specific features.

If shared triggering intentions are deliberate (the second case), they can be conceived as anchoring an initial plan, so that they inherit its content: as already noted, within the perturbation group there must be agents who literally decide together that during the game they plan to watch they will intend to begin the wave. This is the case of strategic/planned behavior that might or might not appear, depending on specific circumstances. A more precise account of those two cognitive contexts would require some empirical psychological work to estimate the degree of presence of deliberation and propositional attitudes. Moreover, although the initiating groups are relatively small, probability distributions would also be needed to make such estimations: the number of spectators who, either deliberately or spontaneously, rapidly employ their triggering intentions and dissolve them in their co-dependent micro-movements through motor representations must exceed a minimal threshold to have an effective impact on other spectators.

Let us take stock: the intentionalistic ontology in the perturbation group during the formation stage of La Ola can be built with the use of classic accounts of shared intentions plus additional correlated conditions to initiate the wave. Let us enumerate them: (1) proper mood (shared emotions), (2) shared beliefs about initiation and the whole performance, (3) shared triggering intentions to initiate and perform La Ola, which require (4) communication channels (verbal and nonverbal) that facilitate the whole formation stage. It is also worth noting that a triggering intention can simply be communicated by initiating standing-up micro-activities which constitute shared perceptions for other spectators in the group and are almost simultaneously repeated by those from the perturbation group who intuitively understand it as the initiation of the wave (communication through movement). Naturally, the ontology of those micro-activities can be investigated further. Especially those concepts that allow one to pinpoint cognition-movement links appear to be particularly important. Hence, (5) nonpropositional motor representations that represent kinematic features of bodily movements during the sequences of standing up and sitting down should be included into the intentionalistic framework. The concept of motor representation, originally related to the idea of so-called motor intentions—informationally encapsulated and unconscious motoric controllers of movement (Jeannerod 2006; Pacherie 2008)—is not free from theoretical intricacies (Brozzo 2017; Butterfill and Sinigaglia 2014; Mylopoulos and Pacherie 2017). Still, understanding it as a nonpropositional (or even nonconceptual) representation of motoric goals that at least partially determine the content of intentions (Brozzo 2017) prompts to include this concept in those intentionalistic accounts of shared action that are open to ideas from the area of motor cognition. ⁸

To wrap up: if we put the above conditions on one list, the schema of the intentionalistic ontology in the formation stage of La Ola would be as follows:

Perturbation group:

This schema can be interpreted as a basic answer to the question about the intentionalistic mechanisms responsible for the “excitation” of the wave. Two things that now call for explanation are what intentional resources are needed for thousands of spectators to further transmit the collective movement that is excited by the perturbation group, that is, perform the waving effect around the whole stadium, and how such a domino-effect is possible.

3.3. Propagation Stage (I): Automaticity in the Followers Group

We are moving now to the propagation stage, in which the waving effect is diffused onto other spectators, the vast majority of a stadium. As noted, the key stipulation here is that the intentionalistic properties of those who belong to the perturbation group need not be instantiated in the psychological profile of most spectators performing La Ola. Cognitive factors responsible for this situation may be diverse. The simplest intuition, hinted above (3.1.), leads to bounded resources (the idea rooted in the work of Herber Simon 1976, 1969). According to this intuition, cognitive and attentional resources of bounded agents cannot always warrant formation of full-blooded intentions to take part if the group of those agents is large and their cooperative movements are to be as fast as it is required by the performance of a typical La Ola. On the other hand, due to the high pace of those movements, it might also be that the conditions for formation of propositional attitudes acted as inhibitors of the whole activity if they belonged to the psychological profile of each and every spectator. As in the case of many automatic performances of individual actions, ⁹ here, on the micro-level, too much thinking (attentional control, deliberative processes and self-consciousness) could block a desired macro-level outcome. Whatever the psycho-cognitive mechanisms responsible for this empirically complex issue are, it is reasonable to assume that movements in the propagation stage of La Ola must be highly automatized to provide sufficient speed and fluency of the whole activity. And this requires the strong presence of a special sort of organization.

In the literature, that kind of organization is called “emergent coordination” (Knoblich et al. 2011). It differs from planned coordination because it is not only spontaneous but also highly automatic, that is, it requires neither collective decision-making and deliberation nor joint perceptions and full-fledged shared intentions. ¹⁰ How, then, is such coordination achieved? According to Knoblich et al. (2011), it has four sources: (1) motor coordination (“entrainment”), (2) shared affordances, (3) perception-action matching, and (4) action simulation. Since (as I assume) all of them can be detected in La Ola in the propagation stage, let us briefly discuss them.

In social motor psychology, entrainment is a basic intersubjective process providing synchronization (Schmidt et al. 2011), for example, as in the case of an audience spontaneously clapping in unison. Similar synchronization seems to occur among spectators who participate in the “waving.” Furthermore, the “waving” must be somehow environmentally structured and enabled. La Ola is structured by stadium seats, which build and specifically induce activity repertoires (standing up and sitting down) for spectators. Ecological psychologists would say that they are spectators’ affordances. ¹¹ Naturally, they are not sufficient to perform the wave, but they significantly condition the performance. Perception-action matching is the process that induces certain activity on the basis of visual information and other perceptual data—for example, if someone knows how to dance, seeing someone dancing may automatically induce doing the same (Cross et al. 2006). Individuals-spectators who jointly perform La Ola in the propagation stage are, in a sense, stimulated by the behavior of their neighbors and this must be minimally guaranteed on the perceptual level. Indeed, to achieve the proper “waving” effect, these agents must also somehow forecast the timing of the micro-activity they will perform in response to micro-activities of their neighbors. This requires the process of minimal action simulation (Knoblich and Sebanz 2008).

Psychological and neurocognitive literature on the above four processes related to intersubjective coordination has been growing (see Passos et al. 2016), but among the many cases studied (dance, sports, etc.), the Mexican Wave is still missing. Research streams related to the above sources of coordination certainly open promising paths for explorations in our context, but the accounts provided by them appear as too general to claim that they sufficiently describe the intentionalistic mechanisms of the waving effect. Before more detailed empirical studies emerge, an action-theoretical account that uses cognitive-psychological research will help us understand these mechanisms. Where empirical data and methods are scarce (or none at all), philosophically elaborated intuitions—supported by relevant theories—are highly desirable. There are many instances where the philosophy of action, informed by social and cognitive psychology, is able to convincingly explain intentionalistic mechanisms of actions. ¹² There are good reasons to continue this path in the context of such actions as La Ola.

I stipulate that entrainment, perception-action matching and action simulation in La Ola require the phenomenon called automatic imitation (Heyes 2011). It is neither an entirely blind and mechanical “doubling” of a given activity (like mimicking mannerisms in the Chameleon effect; Chartrand and Bargh 1999) nor a complex imitation enabling new skills (like intended exercising—cf. Makowski 2017). Automatic imitation is based on associations linking sensory representations of topographical features of a given activity with motor representations of the same or “corresponding” topographical features of that activity. Automaticity here means—similarly to the case of research on automaticity in psychology (Bargh 1994) and in the philosophy of action (Di Nucci 2013; Makowski 2017)—relative independence from the agent’s intentions: motor mimicry occurs without intention formation, without direct self-conscious attentional control of the stimuli and independently from already active behavioral response sets. Although the phenomenon in question has been observed in relatively simple scenarios (such as finger tapping or hand closing), it can reasonably be claimed that it conditions the performance of La Ola at its micro-level psycho-cognitive mechanism. Spectators automatically imitate activities performed by their neighbors; when these imitative behaviors are properly entrained and simulated on the micro-level, the relatively stable macro-level waving effect appears. This is why the propagation stage is performed by spectators who can be called the followers group—its members simply follow activities initiated in the formation group by imitating them.

Nevertheless, one may reasonably doubt that automatic copying is sufficient to begin the domino-effect and to keep it going through the whole stadium. Coordinated movement of a stadium crowd may require additional resources: (1) during the beginning of the domino-effect that emerges after the formation stage and (2) for surviving the whole activity during later phases of the propagation stage. Let me begin with the first.

According to Vicsek and his colleagues, the domino-effect depends on external stimulus which is, recall, “a distance-and direction-weighted concentration of nearby active people exceeding a threshold value” (Farkas et al. 2002, 131). The number of active spectators certainly plays an important role here, but it does not explain the psychological micro-level mechanism needed to further transmit the domino-effect from the formation stage. Mere automatic imitation may not be enough. More specific theories of imitation and emulation can help us understand the issue.

One of them is the theory of “string parsing” (Byrne 1999, 2003). Although it has been proposed in a different context (animals and infants learning new skills), it may offer a direction to further explore the imitative mechanisms in La Ola. The basic idea of string parsing is that perception of complex behavior that is subject to imitation must first be analyzed into simpler streams of elements. These streams are then interpreted to provide suitable representation of movement. In other words, agents need to perceive the whole action as built of “strings” or sequences of elements which are then “parsed.” Parsing allows for the generation of a statistical sketch of the organized behavioral structure which aids the comprehension of the movement as based on causal relations (Byrne 1999). Importantly, this sketch allows one to mirror the initial coordination of organized behavior, that is, the conditions of its beginning, its outcome, and the statistical regularities which link them and build the whole action. This, in turn, allows for copying or imitating the action without the need to comprehend the intentions of other agents. The whole process is automatic (similarly to Heyes’ idea of automatic imitation). Surely, in the context of the Mexican wave, people do not learn anything new for their repertoire of skills. Still, all these preliminary micro-activities that are repeated several times by the formation group to start the action can be understood as structures of linear strings. These strings are parsed by spectators who make up the statistical sketch of the whole action before it is mirrored and copied in their performance. “String parsing” may help us understand how the formation stage gains micro-level stability and fluency to the degree that the waving propagates. Imitation of singular micro-activities of the waving—in the crucial moment of the beginning of the domino effect—can be understood as a result of parsing. Ultimately, whether subsequent copying activities require parsing or they may be performed more automatically is an empirical issue which cannot be determined from the armchair. What appears to be crucial for the intentionalistic profile of the domino-effect in the propagation stage, is that it does not seem to require intention formation. ¹³

3.4. Propagation Stage (II): Shared Intentionality of the Followers Group

We build on the idea that most spectators performing La Ola in the stadium automatically copy what their neighbors do. Micro-level copying shapes macro-level outcome. But if we agree that La Ola is a case of collective intentionality, then another question arises: does the micro-level copying provide a sufficient dose of coordination to keep the whole action going, after a successful initiation of the domino-effect?

Classic accounts of collective intentionality which are based on strong folk-psychological premises would, again, refer to propositional attitudes, but they seem to be redundant in emergent coordination based on automatic copying. Nonetheless, although we introduced automatic copying as a “social glue” for La Ola, its rudimentary collective intentionality requires—as I stipulate—something more to provide sufficient coordination in the propagation stage. There must be something throughout the whole action that provides a relevant (to use the Bratmanian term) “mesh” between all the micro-activities. The sole automatic imitation, aided by proper mood of the crowd, may not be enough to provide it among thousands of spectators. Stability of such a large performance may require something else. This is again an empirical issue, but for the reason already mentioned—the need for philosophical insights where empirically testable knowledge is scarce—more theorizing is required.

It is rather clear that the triggering intention to perform the wave, spontaneously produced in the formation stage, must be effectively transmitted during the whole process among a sufficient number of agents. I stipulate that this is enough for the occasional emergence of mental states which are nonetheless propositional. The reason for this stipulation is as follows: longer strands of imitative micro-activities of standing up and sitting down (an imitation of an imitation of an imitation . . . etc.) could lose their automatic power and effectiveness to transmit the triggering intention. Thus, occasional appropriate dispersion of mindful intentions to contribute to the performance of the whole action—call them shared sustaining intentions—appears to be necessary to provide a minimal dose of collective intentionality in the propagation stage. In other words, some spectators do not merely copy what their neighbors do: they also fully understand what is happening, believe that their micro-activities will contribute to the overall outcome and accordingly co-intend to play their part. Sustaining intentions are also proximate: although they can have propositional form, they need not precede the execution of the activity. They guide it, control it, and integrate perceptual information about action, its idea (schema) and its situational context (see Pacherie 2008). Verbal communication channels are not necessary to transmit sustaining intentions: they appear stochastically, dispersed throughout the stadium on the basis of shared visual information and motivational impact that is made by neighboring micro-activities. Nor are they conditional with respect to what others may or may not intend. Their condition is the sufficient presence of copying behavior around those who have them. ¹⁴ What the specific phenomenology of sustaining intentions looks like is a matter for further studies (see Pacherie 2012).

We should also notice that shared sustaining intentions should be considered significantly different from shared triggering intentions. Although the goal of action coded by the former is more or less the same as the goal of action coded by the latter and both can be interpreted as present-directed or proximate, these two should not be conceived of as different tokens of the same type of intention. Why? The propositional content of sustaining intentions seems to be more related to contributing to the overall performative success of La Ola (playing one’s part), while the content of triggering intentions is more related to the successfully initiating the activity. There is a semantic difference between them. The difference is also cognitive: in comparison to triggering intentions, sustaining intentions appear informationally more encapsulated: their content cannot be directly inherited from the action plan, so their cognitive permeability is limited.

The claim that stable propagation of the wave is conditioned by dispersed shared sustaining intentions is important not only because it reveals another condition of cooperation in La Ola, but also because it suggests that full automaticity of imitative behavior on a large social scale featuring multi-agent cooperative actions seems to be impossible. This stipulation, remaining to an extent in accordance with classic approaches to collective intentionality, gains some support from the psychological theory of imitation. Heyes (2011) points to three ways in which intentions interfere with automatic imitation:

. . . input modulation, where broadly attentional processes influence the extent of action stimulus processing; output modulation, where imitative motor representations are inhibited or allowed to gain control of overt behavior; and intervention, where context-dependent intentions have a direct effect on the process that converts activation of an action stimulus representation into activation of a topographically similar response representation. (Heyes 2011, 497, my italics)

This “tripartite scheme”—analogous to other psychological models of automatic behavior (see, e.g., Wood and Rünger 2016)—refers to both individual and social contexts, so it can also be considered in the case of La Ola. Heyes found evidence for the first two types of intention-imitation interference. It seems that in the present context, output modulation is especially important: shared sustaining intentions to collectively perform the wave do not stifle automatic mimicry, but they allow for imitative motor representations to gain control over a micro-activity that in other circumstances would be the result of copying. ¹⁵ One might also say that output modulation from intentions catalyzes imitation.

Let us take stock. The intentionalistic ontology of the propagation stage of La Ola is significantly different from the ontology of its formation stage. Analogously to the previous stage, we can enumerate the conditions of shared intentionality. First, the propagation stage requires (1) proper shared emotions (here they are more related to the mood that keeps La Ola alive than to excitability). Furthermore, (2) automatic copying provides for smooth propagation of the activity; that automatic copying is occasionally catalyzed by (3) shared performance beliefs about the generic sense of the performance and playing a part in it, and by (4) shared sustaining intentions to contribute to the whole performance. Motor representations (5) again play their roles in performing micro-movements. But this is not the full picture.

It is clear that La Ola can persist even if some spectators simply happen to miss copying what their neighbors do. So, there must be sufficient statistical spread (dispersion) of automatic imitation throughout the stadium that is additionally catalyzed by sufficient dispersion of sustaining intentions. It is plain that the former must be drastically higher than the latter due to the overall automaticity of the wave, but still the dispersion of sustaining intentions (and the beliefs that accompany them) must also be noticeable to warrant minimally shared cooperative intentionality. A similar claim should be made in regard to the level of shared emotions (enthusiasm). In all these cases, there are thresholds below which La Ola dissipates. So, if we consider those basic statistical indicators, the ontology of the propagation stage can be summarized as follows:

Followers group:

Dispersion indicators related to thresholds of the presence of beliefs, intentions, emotions, and copying reflect the fact that the propagation of La Ola can be more or less automatized. What these thresholds look like it depends on probabilistic measures adjusted to the specific empirical contexts of each and every instance of La Ola. In this respect, the simulations made by Vicsek and his colleagues (Farkas et al. 2003) can certainly be instructive, although they do not touch upon the question of intentionality. Thus, the intentionalistic ontology of groups performing La Ola opens a direction in which adjustments to such simulations can go to warrant their adequacy.

4.1. Conditions of Cooperation and Automaticity

Mass actions or crowd behaviors are prominent phenomena in the social sciences. Although there are many interesting theories here (e.g., contagion theory, convergence theory, emergent norm theory, or value-added theory), the mere question of cooperation in this domain can be solved with the aid of the Bayesian game theory. The conditions of the so-called minimal cooperation instantiated by the aforementioned flash mobs or demonstrations studied recently by Paternotte (2012) make a good case here (see section 1). Unlike in the classic game theory, minimal cooperation can be justified on the basis of Bayesian Nash equilibria related to propensities to act—not to routinized decisions and strategies that maximize expected utility. Agents involved in demonstrations do not need to know others’ decisions, but only the probability that they form them. Stadium behavior exemplified by La Ola reveals both some similarity and a significant difference from minimal cooperation explored through the prism of game theory.

Certainly, La Ola also minimizes the conditions of cooperation: it requires neither an awareness of cooperation from all of the participating agents nor their having a shared or preshared plan for whole action. In fact, the constraints of cooperation for La Ola seem to be even more minimal than in the case of flash mobs or demonstrations. Here is the most salient difference: in terms of stability, frequency or dispersion, the key condition for cooperation is provided by the automatic aspect of human movement and motoricity (automatic imitation), so the stability of tendencies to form decisions to cooperate in La Ola—if it appears at all—is statistically insignificant; it cannot serve as a social glue. In fact, Bayesian Nash equilibria cannot be used to explain the cooperative role that shared sustaining intentions play either. A sustaining intention (which can be understood, roughly, as a decision) is not based on the probability of spectators intending to perform La Ola, but solely on the probability of their automatic copying. So, even if Bayesian Nash equilibria are interpreted in the weakest sense possible, as a formalization of a spectator’s identification with a group performing La Ola (cf. Paternotte 2012, 59, footnote 30), it does not change the overall impression that the case discussed requires a significant departure from the game-theoretic framework. To properly understand such phenomena like La Ola, the idea of probabilistically interpreted decisions and propensities to act should give way to the automatic aspects of behavior. All this means that studies of cooperation in mass behavior should fully consider motor psychology to an extent that has not been envisioned by extant theorizing. Rationalistic approaches may not always be apt or sufficient.

The role of emergent cooperation has already been recognized in the field of social and cognitive psychology, but the same does not appear to be the case in the social sciences. Such issues as the micro-constitutive and micro-causal role of automatic processes (automatic imitation) for macro-level collective actions can be systematically introduced to all those branches of multilevel or micro-foundational research in which collective movement and motoric aspects of agency are important. ¹⁶

The novum that I wanted to highlight for the social-scientific approaches to cooperation in mass behavior is related to the theme of automaticity that has also long been ignored on the ground of the philosophy of action (Di Nucci 2013). Although, as already noted, the situation somewhat changed recently (especially in the area of research on skills; Brownstein 2014; Brownstein and Madva 2012; Douskos 2017); this theme still has not gained sufficient attention from philosophical theories of shared actions. Some preliminary work is available (Paternotte 2014), but the topic is nonetheless new. In this light, the arguments proposed here aim to show that automaticity should be fully introduced and considered in theories of collective intentionality as well. Automatic individual actions occupy a separate place on the philosophical map of practical intentionality. We should therefore expect future studies in the philosophy of action to recognize the role of automatic performances at the level of collective actions just as they do for individual actions.

4.2. The Problem of Coding Goal Representations

As stated, La Ola is a highly automatized coordinated activity. But its ontology is puzzling when we discuss the conditions of shared intentionality it implies. Shared intentionality requires shared mental states of a certain sort that must meet the criteria of intentional action. Standardly, these criteria—variously conceptualized—amount to goal-directedness or, goal representations. In the cases of what I called shared triggering intentions and shared sustaining intentions the matter is rather obvious: both represent the collective goal to perform the whole action. But automatic copying, which is—according to the approach presented in this essay—statistically the most important micro-level coordinator in La Ola, appears to be much more difficult to assess in this respect. The problem can be reduced to the following empirical question: are representations of the shared goal of La Ola coded in micro-level automatic copying? A constitutivist interpretation of the common-sense intuitions based on examples of automatic individual actions would prompt a positive answer: spectators, even if they perform automatically, “have reason” to copy the activity of their neighbors, so their performance is intentional. At least, this is what we would want to say ex post, on the basis of standard intentionalistic (Anscombean) intuitions. However tempting it would be to rely on this strategy, the estimation of this issue according to the causalist version of intentionalism (in actu), that is, when automatic copying is happening as a sequence of representations, requires empirical studies. So, in this area, there are reasons to resist common-sense intuitions.

In fact, experimental psychology has begun to explore this problem and the answer it provides is rather negative. According to Cole et al. (2018), simple movement such as reaching out with one’s arm can be automatically copied without coding goal representations (more precisely, automatic copying requires coding of kinematic aspects of a given activity which is primary to goal coding). Although these authors do not rely on the concept of automatic imitation strictly in the same sense as Heyes and the empirical example they consider is very simple, I think that their study should be taken seriously in the present context. For there may be cases of the micro-level copying (in La Ola or in a different case of the collective action where it plays an important role) where these findings remain in force. Indeed, practically blind repetition, without representing the goal of the whole action, is a possible scenario for performances of La Ola in very large stadiums.

To understand why this is so, we should imagine the following example: you are watching a football game with your friend in a large arena full of people, but you two are not really absorbed by the game; instead, you are enthusiastically discussing something that will happen after the match. Suddenly, the wave approaches to your seats. You and your friend just repeat what your neighbors do, but you do so almost mechanically, without even stopping the conversation you were in. Still, you are minimally aware of what you do: the thing you perform does not merely happen to you. This peculiar instance of copying is certainly different from the typical micro-activities your neighbors performed, so it should be considered as a borderline case. Still, if we analyze it on the level of mere execution (i.e., in actu), it makes sense to take it as something that is performed as coding its kinematic aspect—without coding the goal representation (to be sure, the goal can be represented by you and your friend right afterwards, but that is a different question). ¹⁷ At any rate, the whole problem of goal representation in the case of La Ola is a matter of experimental psychology and its scientific solution should consider various micro-level scenarios, especially those ones which instantiate obtrusive behavioral automaticity.

4.3. Hybridity and Modularity

The overall question of intentionality in “stadium agency” leads to the conclusion that its philosophical account should be straightforwardly hybrid: full-blooded intentional states are needed as much as automatic aspects of human performance are. So, if we take La Ola, a collective coordinated motion, as an important example of shared actions, then its account ought to be hybrid in the sense that to explain the nature of its intentionality it must combine the traditional (philosophical) idea of intentional action with the idea of automatic action the study of which is to a large extent an empirical matter. This outcome contributes to those recent developments in the philosophy of shared action which stress minimal cognitive abilities needed for joint action (e.g., Butterfill and Apperly 2013; Michael 2011; Tollefsen 2005), without overly reducing the role of traditional philosophical theory of shared intentionality (in the causalist version).

It is worth emphasizing that the highlighted hybridity is a property of a theory of shared intentionality—not the phenomena it aims to explain: the cognitive-psychological picture of spectators who act together is diverse, but there is no reason to assume that the difference in theoretical approaches (an account of full-fledged intentions on the one hand, and an account of automatic processes on the other) is an upshot of the hybridity of cognitive-psychological properties of agents. I believe that it rather reflects the imperfection and variety of options when it comes to explanations of shared intentionality. On the one hand, classic intentionalistic approaches do not seem to be convincing when it comes to fine-grained explanations of automatic aspects of cooperation. On the other hand, although cognitive-psychological approaches appear to be much more suitable in this area, they still are not mature enough to fully replace more traditional intention-based theorizing. This is why certain salient examples of shared intentionality appear to be (so to speak) explanatory crossbreeds.

Notice also that the hybridity, detected as a specific feature of explanations of collective intentionality instantiated in La Ola, is additionally boosted by the fact that the case on the basis of which an account is proposed is necessarily modular. This property of groups of collective agents appears frequently in masses or large teams, when a group size is sufficiently large to cause division of “labor” among its members, as in the cases of an orchestral performance or a demonstration (Paternotte 2012; Smith 2014). Automatic imitation, if it occurs at all, is scarce in what I called the perturbation group, but it dominates in the followers group. Naturally, those differences are not ontologically strong in the sense that if an agent happens to belong to one group, she or he could not belong to another group. ¹⁸ The division of micro-level activities is stochastic to a significant extent, and it depends largely on the contingent place agents occupy in the stadium. Agents-spectators who happen to appear in one group or another have the cognitive skills required to perform micro-activities in both of them. In other words, differences in those aspects of their psychological profiles that are relevant to performing La Ola are important only in relation to the group to which they happen to belong.