Coordination as Naturalistic Social Ontology: Constraints and Explanation

Game Theory as a Model for Social Ontology

The tradition of understanding coordination as a source of social order is historically rich. It dates back to Hobbes (1651 [2016]) with his idea of social order emerging from the rational deliberation of individual agents and resulting into a jointly optimal social decision, an agreement to form a state. The so-called “Hobbesian problem” of social order has been popularized and given the status of the main theoretical problem of sociology by 1. The notion of social order presupposes a certain domain ontology given by the relations between agents. As Epstein (2018) points out, a notion of convention was first used as an alternative to agreement by Pufendorf (1673), to refer to language and law. His point was that conventions do not need to be explicitly agreed to and might exist and work without their intentional design. Later, Hume has famously contributed to the advancement of the notion of convention, which is now often referred to and quoted when coordination problems are involved: “Two men who pull at the oars of a boat, do it by an agreement or convention, tho’ they have never given promises to each other” (Hume 1740 [2003], Bk III, Pt II, Sec II). Hume offered analyses of many social phenomena like money, property, government, justice and others in terms of convention. After Hume, philosophers in the Scottish Enlightenment held that social order is an emergent product of individuals’ interactions, however, no such order has been specifically intended by individuals. As Ferguson (1767 [1995]) writes, “nations stumble on establishments which are, indeed, the result of human action, but not the execution of any human design”. In other words, tacit conventions as a fundamental part of the human social world have been discussed long before the rise of game theory, and it became especially convenient to treat conventions using game theory.

The study of convention gained particular momentum with Lewis’s seminal book Convention (1969 [2002]), which, according to Guala (2007), was the first attempt to apply rational choice analysis to the domain of social ontology. Lewis saw conventions as solutions to coordination problems in game-theoretic sense, a basic element of social ontology. But what are conventions and what relation do they bear to coordination?

According to Lewis, social conventions are behavioral patterns emerging from repeated coordination problems between two or more players. The distinctive feature of conventions is that players conform to these patterns, for they expect others to do so, and it is common knowledge that every player is expected to conform. Deviation from a conventional choice of action leads to lower payoff, so players do not have incentives to deviate unilaterally. As conventions are defined in terms of coordination problems, it is useful to elaborate on them.

Coordination problems are situations where agents have common interests and it is not evident how they can be satisfied. O’Connor (2019) distinguishes two classes of coordination problems, correlative and complementary ones. In correlative coordination problems, agents need to converge on the same choice to coordinate successfully. For example, consider a driving game, where two players drive toward each other and each can choose the left or right side to drive on. If they both are on the same side and no one swerves, they might crash, and if each of them chooses a different side, they will stay safe. One important feature of this and other coordination problems is arbitrariness, meaning that it does not matter on what side both players would converge. Instead, what matters is that they either coordinate by choosing the same action, for example, swerving to the right. On the game matrix, it is represented as two non-unique equilibria. It means that either of them solves the coordination problem (Figure 1).OPEN IN VIEWER

Complementary coordination problems, as opposed to correlative ones, require agents to perform different actions, or strategies, to coordinate successfully. As O’Connor (2019) points out, division of labor and of resources are examples of this class of games. For instance, two roommates want to organize a party and invite guests. To proceed, they need to tidy up the house and order pizza delivery. If they both do the cleaning, there will be no food when the guests come, and if they both order pizza delivery, they will have plenty of food but be embarrassed by the mess at the house (Figure 2).OPEN IN VIEWER

Regarding conventions, O’Connor (2019) draws two important distinctions: between conventions and social norms, and between more and less arbitrary conventions. First distinction means that not all behavioral regularities have normative force. For example, friends have a convention of meeting each Friday evening at a bar, and showing up is not what each of them strictly ought to do, for if someone does not come, it is fine for the rest of the friends. On the contrary, if two cars are driving on the same side of the road toward each other, the drivers are forced to swerve, for otherwise they might crash. They ought to swerve, for not only might one of them be fined but they might cause an accident. To clarify, as Bicchieri (2005) points out, conventions are different from social norms in the relationship between self- and common interest. They coincide in the former and do not necessarily coincide in the latter. In the case of friends at a bar, there is no or little tension between self- and common interest, while in the case of driving cars there is. O’Connor stresses that conventions and norms are the poles of a continuum along which the former acquire normative force.

The second distinction concerns the arbitrary and historically contingent nature of conventions that they “might have been otherwise”. According to Lewis, this arbitrariness is one of the key distinguishing aspects of conventions. However, as Gilbert (1992) points out in her critique of Lewis’s work, not all possible solutions to a coordination problem are equally profitable for players. In cases where one way of coordinating is more preferred than another, convention will not be that arbitrary. In other words, arbitrariness is a feature of conventions that is a continuum between contingency and necessity. For example, signaling between vervet monkeys might well be modeled as a convention in the Lewisian sense of repeated behavioral patterns of solving coordination problems (cf. Harms 2004; Skyrms 2010). However, this convention is not historically contingent in the sense of several possible solutions being equally profitable, for there are evolutionary constraints breaking the symmetry between multiple equilibria. Agents might be hardwired to certain strategies. This distinction, as O’Connor underlines, illuminates some conventions as more functional and others as more conventional. ⁴

Another important concept, tightly connected to conventions and norms is that of institution, for it bridges the issue of coordination problems with social science. As famously dubbed by North, social institutions are “the rules of the game in a society or, more formally, the humanly devised constraints that shape human interactions” (North 1990, 3–4). They are also self-sustaining salient behavioral patterns (Aoki 2007) and norm-governed social practices (Tuomela 2013). However, it is not clear how institutions are related to conventions and norms.

Guala (2016) offers a definition of institutions as sets of correlated equilibria with normative force. With Hindriks, they propose a “unified social ontology” that views social institutions as rules in equilibria represented by theoretical terms like “money” or “marriage” (Hindriks and Guala 2015). “Rules” are the recipes that guide and prescribe certain behavior and that are used by the agents themselves, and “equilibria” are objective stable states of the strategic interaction between agents and population thereof. Rules-in-equilibria theory bridges the accounts of regulative rules, equilibria of strategic games and constitutive rules of the form: “X counts as Y in C”. The former two are complementary ones, and the latter one supplements it by providing a symbolic representation, or term like “money”.

The rule-based account conceives of social institutions as rules guiding and constraining behavior in social interaction. In sociology, the tradition of treating institutions as rules dates back to classical figures like Weber (1924)and Parsons (2015 [1935]). The equilibrium-based account sees institutions as behavioral regularities and, most importantly, solutions to coordination problems. The constitutive rules account, introduced by Searle (1995, 2010), sees institutions as systems assigning statuses and functions to physical entities.

According to Guala and Hindriks, the rule-based account is insufficient, for it cannot explain why some rules are followed and others not. To address this issue, an equilibrium account is needed to show the strategic character of rule-following. They illustrate this point by comparing the two paradigmatic games from game theory: the prisoner’s dilemma and stag hunt (Figure 3).OPEN IN VIEWER

Although mutual defection in the prisoner’s dilemma is a Nash equlibrium, ⁵ it is not a social institution, for it is not self-sustaining due to the independence of players’ strategies. In contrast, the mutual decision to hunt a stag instead of a hare, which are also both Nash equilibria, is actually an institution, for it requires a correlation of players’ strategies to achieve a bigger joint payoff. The latter means that the strategy is salient and beneficial for players. This, according to Guala and Hindriks, explains why some rules are followed and others not.

However, the notion of players’ correlated strategies, or correlated equilibrium, ⁶ as an explanans of the stability of institutions is insufficient, as the authors point out, for it is too loose. They provide an example of non-human animals solving coordination problems, but still not having institutions. For instance, male baboons, lions, swallowtails and other species exhibit a recurring behavioral pattern that can be described in terms of correlated equilibrium, when males patrol an area to mate with females and have ritual fights with intruders if encountered. The evolved pair of players’ strategies—“fight if own” and “surrender if do not own”—minimizes possible damage to both parties and lets the owner occupy territory and mate. This problem was originally represented with the “Hawk-dove game” by Maynard Smith (1982). This game is symmetric, meaning that each player has the same action profiles. This is why there is only one number in each cell (Figure 4).OPEN IN VIEWER

Maynard Smith calls this solution “bourgeois equilibrium” (represented as “B” in te game matrix above), for it describes animal territorial behavior: “Hawk if owner” and “Dove if intruder”.

Building in this idea, Guala and Hindriks argue that human and animal conventions differ only in the scope of actionable signals. The set of signals to which animals might respond is rather narrower than that of humans, due to the tight coupling of stimulus and response to achieve coordination. In other words, animals are hardwired to their strategic choices in coordination problems, and human are not. For example, the “hawk” and “dove” strategies in the case of territory ownership in animals is genetically inherited and not arbitrary. It is functional rather than conventional in O’Connor’s terms. However, as Sterelny (2003) argues, more complex creatures like humans are able to decouple stimuli and behaviour with the aid of representation of the environment that conditions behavior. In other words, humans are able to invent and follow different rules given the same correlation device—a source of signal that correlates agents’ actions. Moreover, rules are themselves symbolic representations of strategies of a given game. These representations not only serve as symbolic markers of the properties of equilibria, but considerably save cognitive effort by functioning as a shortcut. “Stop if red” is such a rule that guides behavior and at the same time serves a symbolic representation of an equilibrium. But this alone does not say much about social ontology.

To address this issue, Guala and Hindriks, drawing on Hindriks (2005), propose to bridge their rules-in-equilibria account of institutions with the constitutive rules account. The latter presents institutions as systems of statuses and functions paradigmatically proposed by Searle (1995) as the formula “X counts as Y in C”. Searle draws a sharp distinction between constitutive and regulative rules, emphasizing the difference in their syntax, for that of the latter is “if X, do Y”. As the authors note, constitutive rules are linguistically transformed regulative rules, aided with a new term to name an institution. Combining these accounts enables researchers to investigate Y-terms like “money” or “marriage” used by individuals in everyday life and analyze their internal regulative and strategic character, thus bridging explicit ontology of social science and implicit ontology of ordinary language.

The unified rules-in-equilibria account has several shortcomings. Although it mentions that social coordination as correlation of strategies with normative forces seems to have evolutionary roots, this approach does not address the constraints shaping animal conventions into social institutions and the role of cognitive mechanisms in it. However, Guala (2020) is concerned with a similar issue and asks, “what cognitive mechanisms establish coordination?” In addition, Bicchieri (2005, 2018) refers to social norms as being “activated” by scripts and grounded in cognitive schemata. However, its inner workings remain unexplained.

The main issue with game-theoretic explanations of social coordination in terms of equilibria, as Kaidesoja et al. (2019) argue, is that they do not explicate causal processes or mechanisms. Hence, the question to be answered is “what cognitive mechanisms are responsible for the evolutionary transition from animal conventions to human social institutions?” This would require deeper entwinement of game-theory, cognitive science and evolutionary theory.

The CNSO Framework

To start unfolding the “Coordination as Naturalistic Social Ontology” framework, it is important to note that the notion of coordination is relevant in two meanings, game-theoretic and cognitive scientific ones. In the former case, coordination (C1) is a successful result of agents’ strategies leading to a solution of a game, either optimal or suboptimal. In the latter case, coordination (C2) is also a successful result of a social interaction, either spatiotemporally synchronous or not, that is possible due to the cognitive capacities to represent goals and actions, to monitor and predict those and to adjust one’s own actions to the actions of another individual with whom coordination takes place (Vesper et al. 2010).

Naturally constrained social ontology might be understood as coordination—both a continuing process and its results recursively interacting with each other: small groups, social institutions and social structures. ⁸ Coordination itself can be understood as a causal mechanism of correlating agents’ strategies by their monitoring, predicting and adjusting to other agent’s actions. Coordination has the global scope of evolutionary time which is computationally tractable via dynamic models, and the scope of local social interaction that is empirically testable with the cognitive scientific experiments in joint action research (Knoblich et al. 2011; Török et al. 2019; Sebanz and Knoblich 2021). This description implies two natural constraints on social ontology as coordination.

The first natural constraint is evolutionary conditions for equilibrium emergence for distinctively human forms of social coordination, which is social institutions as described in rules-in-equilibria account earlier. Namely, if animals are able to solve coordination problems with “animal conventions”, which might be described formally as evolutionary stable strategies in coordination and cooperation games, what enables the emergence of social institutions as conventions with normative force? Evolutionary conditions for the emergence of uniquely human coordination constraint social ontology by naturalizing the problem of social order—that social institutions were not created ex machina, but emerged gradually from the other forms of coordination that had no full-fledged normative component. ⁹

The second, cognitive constraint is closely related to the first one. If social institutions evolved from animal forms of coordination, and humans are drastically distinct from animals in terms of cognitive capacities, these capacities must have influenced the emergence of social institutions. Although this sounds truistic, it nevertheless binds social ontology to these cognitive capacities, meaning that social institutions are ontologically dependent on them. As dependence might be causal or constitutive, and cognitive capacities might have had causal influence on the emergence of social institutions, it follows that institutions might causally depend on cognitive mechanisms. ¹⁰ Although it does not make social institutions a causal phenomenon per se, it says that they are only possible due to cognitive capacities of agents.

To further untangle the relationship between the cognitive and the social and to divide the labor, Sarkia et al. (2020) propose a framework based on mechanistic philosophy of science (Craver and Darden 2013; Glennan 2017). Its main idea is that mechanisms for phenomena consist of entities “whose activities and interactions are organized so as to be responsible for the phenomenon” (Glennan 2017, 17). Drawing on Bechtel’s (2009) exposition of mechanistic explanation in cognitive science, the authors suggest that such explanations might be given by answering four questions: (1) what is the phenomenon? (2) what entities and processes does it consist of? (3) what are the interactions of these entities contributing to the phenomenon? (4) what is the environment where the mechanism is situated, and how does it affect its functioning? The division of labor, the authors argue, is accomplished by answering different sets of Bechtel’s questions. Evolutionary anthropologists and developmental psychologists answer the first and the fourth of them by studying the key differences between human and great ape cognitive capacities and the development of uniquely human ones in the course of human ontogeny. Cognitive scientists answer the second and third questions by identifying particular cognitive mechanisms, i.e., entrainment and common object affordances (Knoblich et al. 2011). In addition, the authors put forward the idea that social scientists might address further questions outlying those suggested by Bechtel, i.e., complex social networks emergent from social coordination.

CNSO holds that coordination is the main ontological unit of the social—it is an ongoing process governed by a causal mechanism that leads (intentionally or not) to social institutions—normatively-driven self-sustaining behavioral regularities. Social institutions provide familiar sociological terms like roles, and the latter depend on the former constitutively, not causally: if there were no social institution of “marriage”, there were no roles of “wife” and “husband” that imply certain rights and duties. These are the first-order derivatives of CNSO, and the notion of social structures as patterns of relationships between agents is its second-order derivative.

This derivation is a preliminary way to address Kincaid’s discussion of social structures as representing conceptual and not causal relations (Kincaid 2008). If coordination is a causal mechanism, and its successful recurring result is a social institution as rules-in-equilibria, it means that social institutions are an inherently causal phenomenon with an appended theoretical term like “marriage”. This term generates conceptual relations between roles. This would show how causal relations regarding the social establish conceptual, or constitutive ones. Overall, naturalism about what Epstein (2015) calls anchors—instantiation conditions for what to count as a social fact—means that the main ontological unit of the social—coordination resulting into institutions—comes into being causally, and not constitutively. However, these implications must be addressed in a separate paper. In the next section I address a question of relating two different types of explanation involved in the CNSO framework: equilibrium and causal-mechanistic.