1 Institutions and norms: what shall be done

Institutions structure human behaviour and enable groups to achieve their goals. Standard examples of institutions include phenomena such as marriage or private property. These can be written laws, e.g. patents and copyrights, as well as spontaneous behavioural patterns and social practices, e.g. traditional haggling over prices.

Two main accounts of institutions prevail within the philosophical literature: rule-following and equilibrium theory (see Greif & Kingston, 2011; Hogson 2015), supplemented by various integrative or hybrid theories. I shall briefly introduce all three possibilities, however the latter shall be my main point of reference. Specifically, I shall refer to the hybrid theory proposed by Frank Hindriks (2023) based on the account of social norms formulated by Cristina Bicchieri (2006, 2016). In brief, it portrays an institution as a spontaneously generated coordination of beliefs and expectations within a collective, backed by motivations for the proper behaviour of its members as well as sanctions for infringement.

The phenomena of coordination, cooperation and behavioural normativity have also attracted scholars working in the field of embodied/enactive cognition. I shall mostly refer to a set of conceptions belonging to the tradition established by Francisco Varela, Evan Thompson and Eleanor Rosch (1991). In particular, I have in mind the groundbreaking concept of participatory sense-making introduced by Hanne De Jaegher and Ezequiel Di Paolo (2007) and a more recent proposal by De Jaegher (2021).

This paper attempts to show, first, that the available theories of institutions are limited in their accounts of the normativity and motivational force of institutions, and that is because they lack a more in-depth account of an agent, or subject. The latter, I argue, should be considered as a specific system coupled with its environment. In this context, the paper argues that using some of the concepts elaborated within the enactivist paradigm can help in overcoming these limitations.

On the other hand, since enactivismFootnote 1 has entered the domain of social theory quite boldly (e.g. beginning with De Jaegher & Di Paolo, 2007; see also Gallagher & Crisafi, 2009 and De Jaegher, 2013), it now becomes necessary to engage with the existing literature, at the very least so as to avoid the possibility of barging through an open door and creating concepts which have already been introduced. An additional meta-philosophical aim of this paper is thus to draw a conceptual link between these two contemporary research programmes which both deal with cooperative behaviour and the normative constraints placed upon it, but which have not yet been– at least to my knowledge– robustly connected (although Petracca & Gallagher, 2020 already provide a promising ground for this conceptual bond; see also Gallagher et al., 2019; Gallagher & Petracca 2022). There could be potential mutual benefits from a collaboration between the two, yet the contribution made by enactivism to the philosophy of institutions (a part of “social ontology,” as it is frequently called; see Guala & Hindriks, 2015) is the one I shall focus on.

The final result of the paper shall be a relatively new account of institutions approached as answers to specific collective demands. These local demands emerge in complex problem domains which require behavioural flexibility and therefore, I argue, also some additional behavioural patterns that do not directly address the problems, but govern the problem-oriented activities. They are therefore meta-patterns, or patterns of patterns. Institutions are therefore defined as higher-level problem domains brought forth by these meta-patterns of problem-solving (for a different treatment of the bond between institutions and problem domains see Werner, 2023).

The paper consists of two segments. In Sects. 2–4 I examine the proposal put forward by Hindriks (2023) and characterize a place where it needs some additional conceptual work. Sections 6–8 outline the contribution brought in this context by enactivism. Section 5 serves as a link between these two rarely cooccurring domains of inquiry, answering the question of why enactivism is supposed to do the job.

2 Two + one accounts of institutions

According to the rule-following approach famously advocated by Douglas North, institutions are “the rules of the game of society” and their job is to “establish a stable structure to human interaction” (North, 1990: 6). In other words, they are “codes of conduct, norms of behavior, and conventions” (ibid.: 36), and as such they can either be culturally-based and not necessarily written (informal institutions), or explicitly devised by various bodies entitled to do so within a strictly delineated ‘territory’, such as parliaments, unions, committees, employers, school boards, and the like (formal institutions).

However, the focus on norms per se usually employs a top-down perspective and may not be conceptually well-equipped to deal with the local circumstances which determine whether, and why, a given rule is implemented or not, in a particular place, and whether or not it provides the expected outcomes there. Institutional (implementation) failures are cases in which an institutional solution, designed from the top down and backed by powerful organisations such as the World Bank or the IMF, turned out to not be workable, in part due to its blindness to actual problems in specific parts of the world (see Chang, 2001, 2011).

When faced with implementation failures, the sole answer provided by the rule-following theory is that the people on the ground failed to learn the rule appropriately, or did not understand it, or not follow it sufficiently closely. Moreover, there might be some other factors such as interest groups, distribution of power, etc. which actively stymied the implementation (see e.g. Acemoglu, 2006). In principle, this is not an inadequate model, but as e.g. Hindriks (2023) points out, the rule-following approach tends to emphasise the normative character of institutions, while failing to state why compliance is supposed to be preferable from the standpoint of an actual agent. This implies, as Hindriks articulates it, that the rule-following conception says little about the motivations of those supposed to adopt a rule.

Intuitively, it is quite clear that motivation does play a role, yet I would like to portray the problem of implementation in a different manner, based on Chang (2011). Namely, whether or not a rule is adopted may relate to how people perceive their local needs, and therefore what their actual demands are. We all act within specific, contextual environments. Thus the conditions for successful action, be it economic activities such as growing apples, hairdressing, producing microchips, or some different kind of activity, such as cycling through a crowded town, are always determined by their local specificities. If that is so, then when theorising about institutions, we should at least supplement the top-down, rule-following approach with a bottom-up analysis of how not merely norms, but also demands for specific norms, are generated within these local conditions (see Ostrom, 2005). This means that we must pay attention not only to rules, but also to the actual environments in which they emerge, and where they can be expected to be implemented.

The alternative theory of institutions makes no mention of rules whatsoever, but instead focuses on behaviour itself, as well as agent preferences manifested through behaviour. The choice of a particular behaviour is understood there as a move within a game. The basic idea is that some group behaviours have a specific tendency to create patterns composed of individual, interdependent behaviours. Interdependency means that what one agent does has an impact on what another agent will do, by constraining the behaviour of others, creating specific conditions (e.g. incentives) for it, etc. and this relationship creates a complex web of constraints within a group.

The classic example in this respect is road traffic. If we have a group of drivers, each motivated by their own interest and planned destination, and each having certain expectations as to the behaviours of the others, sooner or later, a certain state of equilibrium will emerge, based on these expectations, causing all those driving in one direction to stick to a single side of the road (whether left or right). This conceptualisation dates back to the Nash equilibrium and draws both on game theory as well as dynamic systems theory. Institutions are identified with these spontaneously-generated equilibria, manifesting themselves through behavioural patterns (see Aoki, 2001, 2007; Binmore, 2010, Greif & Kingston, 2011; Hodgson, 2015; Hedoin, 2017).

The problem I would like to address emerges from the following recognition. First, as stated above, the rule-following view says nothing about how abstract rules relate to local circumstances and what links them to actual behaviour, be it motivation, driven by local needs and demands, or something else. Second, while the equilibrium approach does provide an explanation of the origin of behavioural patterns, which means that it provides significant insights into the question of regularities, coordination and cooperation, yet it says nothing about how these patterns are transformed into rules or norms, i.e. how genuine, normative rule-following, considered as a consciously-adopted behavioural strategy, emerges on the basis of spontaneous behavioural regularities. As Hindriks points out, the only answer provided by equilibrium theory refers to sanctions incurred by those agents who fail to comply. In sum, “equilibrium theory conceptualizes institutions as interdependent regularities backed by sanctions. (Hindriks, 2023: 1370).

This situation naturally allows for integrative or hybrid theories combining these two theories in the hope of delivering a more comprehensive view. Hindriks (2023) builds his own hybrid proposal, drawing on Bicchieri’s (2006, 2016) theory of social norms. The latter also tries to integrate game theory with normative discourse.

I shall briefly examine Hindriks’ proposal and then move on to how, in my view, it should be supplemented. Here is why a supplementation should be beneficial. I will be arguing that the hybrid theories do not explain where the normativity of institutions comes from as such: i.e. what exactly provides abstract rules, as well as spontaneous patterns backed by expectations and sanctions, with a genuinely normative character. A more detailed conceptualisation of the acting subject as a complex system coupled with its environment, which is what enactivism provides, may fill the gap.

3 Norms that motivate: from expectations and obligations

To begin with, two kinds of expectations are defined by Bicchieri (2006). One is called ‘empirical’ and refers to situations in which I expect others to act in accordance with a given rule. The other is dubbed ‘normative’ and addresses cases in which I believe other members of my group expect me to act in a certain way. If we have these two kinds of expectations, we can extract the notion of a social norm, as nicely summarised by Hindriks:

…in order for a social norm to exist, people have to know that a rule exists that applies to a particular kind of situation. Furthermore, it must be the case that, if people were to have empirical and normative expectations with respect to a behavioral rule, they would prefer to conform to that rule. Thus, all it takes for a social norm to exist is that people have preferences that are conditional on both kinds of expectations (Hindriks, 2023: 1376).

Meanwhile, Hindriks questions whether the acceptance of a rule is required for that rule to become a social norm, and comes to a negative conclusion. To back his intuition, he introduces several thought experiments, of which I shall list just one here.

Imagine a totalitarian state called Oppressia, Hindriks proposes, where conformity is ensured through severe sanctions. Oppressians must manifest their allegiance to the state whenever a state official is present. Because of severe penalties, the Oppressians “conform and expect others to conform. Furthermore, having been brought up in this society, they take it as given that others believe that they ought to do so. But secretly all of them condemn the rule” (Hindriks, 2023: 1379).

Now, Bicchieri would have to deny that the rule of allegiance is a social norm, which according to Hindriks, goes against our basic intuition. Rather, he claims that this fictional case reveals that “normative beliefs are not required” (ibid.). However, he states, normative expectations are. We could say that Oppressia is a belief-free place, while its entire social system is built on expectations backed by sanctions.

To articulate his proposal clearly, Hindriks introduces the following definitions:

A normative rule features the notion of an obligation. Its basic structure is ℜ: ‘It is obligatory to do A in S’, or ‘Everybody ought to do A in S.’ The notion of a normative rule can be used to explain what a normative belief is. This is the belief that everybody ought to perform a particular type of action in some situation. In other words, it is the belief that ℜ. This notion can in turn be used to introduce a new conception of a normative expectation. To have a normative expectation is to expect many others to believe that ℜ (ibid.: 1378).

With that in mind, he argues that a theory based on Bicchieri (2006) considers that “a social norm is a generally accepted normative rule or principle”, meaning that people must have normative beliefs in ℜ if ℜ is supposed to be a social norm. On the other hand, his view is weaker in the sense that “[…] it requires normative expectations instead. Normative beliefs are optional” (ibid.).

According to Hindriks, on this basis we obtain:

A social norm ℜ exists in population P exactly if a substantial number of members of P expect others to believe that ℜ (ibid.)

In other words, if a relatively large number of people in a given community expect that others will believe that ℜ, this gives rise to a certain behavioural pattern, even if nobody actually does believe that ℜ, which suffices for the functioning of ℜ as a social norm.

Yet crucially, being a social norm in this sense is not sufficient for ℜ to become an institution. At this point, Hindriks adds another element. Namely, he says that social norms may by themselves be able to motivate agents to comply, which is an additional and essentially normative factor. In my reading, this means that accounting for the emergence of a social norm based on equilibria of behaviours and expectations is one thing, but there is a further step that has to be made, i.e. explaining how the emergent norm contributes, in a kind of feedback loop, to the behaviours and expectations of the agents (or more broadly, their preferences).

My interpretation of Hindriks is thus that this feedback loop can be seen as an institution, but not the norm itself. We read that “in order for a social norm to constitute an institution, it must affect people’s behavior, or at least their motivation” (ibid.: 1381). And in this context “institutions can differ in terms of strength. They can be weak or strong. An institution is weak exactly if its norm motivates many of its participants but not enough for them to comply” (ibid.). Given the latter, we should be able to say how institutions differ in terms of their strength. Moreover, each theory of institutions can reasonably be expected to explain not only how an institution came to exist, but also how it continues to persist.

It is therefore clear why Hindriks’ approach is indeed a hybrid one. While working within the vocabulary of game theory, in essence descriptive, he then introduces normativity, in new attire so to speak, by referring to what motivates specific moves within the game, and quite rightly claiming that regularities themselves are not the whole tale. In short– institutions are both regularities backed by motivations and regularities that motivate. He explicitly introduces normativity by means of the modal operator “It is obligatory to do A in S.”

However, the problem is that Hindriks fails to address a more fundamental issue: where normativity as such comes from, how it emerges on the basis of behavioural regularities, thus on the basis of coordination and cooperation, or on the basis of motivations. Note: motivations do explain obligations; we are supposed to be already familiar with what the operator “It is obligatory to do A in S” means, since it is clearly introduced as a primitive concept. Thus we must understand what being obligatory stands for as such.

Various deontic modal logics define and formalise this pregiven sense of the concept, and in a sense, Hindriks follows this path. Yet he neither gives any hints where this intuitive sense of normativity comes from (in other words, if we have this intuitive understanding of obligations, where does this very intuition come from?), nor does he try to unpack this intuition in an axiomatic manner. The question thus arises of which factors make us creatures that understand obligations. Addressing this problem shall form the crux of my subsequent remarks, as well as of my own proposal as to how institutions should be understood.

4 Commitment and motivation

Let me start with the question of how beliefs themselves are thought of in Hindriks’ (2023) proposal. He says that while belief in ℜ (a rule) is not necessary for ℜ to function as a social norm, normative expectations are nonetheless necessary. Meanwhile, the latter is the case when others expect me to believe in ℜ, which still makes the character of this belief a key concern.

Recall that ℜ stands for “It is obligatory to do A in S”. I argue that believing that ℜ can articulate two different attitudes. Imagine two travellers in a Muslim country wondering whether they should take off their shoes before entering a mosque and one of them says: “I believe it is obligatory to take off your shoes”. We therefore have a situation in which a belief is merely an assessment of certain facts about the norms in a given society, but not an expression of this person’s commitment to that norm. However, when that very same statement is expressed by the mufti in the mosque in question, it is clearly an articulation of his commitment.

There is a good way in which to articulate this difference: we can provisionally distinguish between de re and de dicto normative beliefs. The former directly refer to specific courses of action that a given rule discusses, and consider these actions as somehow normatively not indifferent. This could include taking off shoes, shaking hands or not, giving back borrowed money, etc. On the other hand, de dicto normative beliefs refer not to these actions as such but to the rule itself. This means that while de re normative beliefs are about things and states of affairs, de dicto normative beliefs are about propositions, not things. Using Hindriks’ notation, it can be said that my de re belief that ℜ, as if through the lens of ℜ, refers to the behaviours ℜ addresses; while my de dicto belief that ℜ is focused exactly on ℜ, as an abstract entity.

Consider the following example. Suppose that a norm says “It is obligatory for a person to be married to only one person”, which articulates the concept of monogamy. We can refer to it in de re mode, by expressing our commitment to this particular understanding of marriage, but we could also articulate a somewhat less engaged de dicto stance, which could state something such as: “Yes, I do believe following the monogamy rule to be obligatory, less because I am personally committed to having just one partner, but that based on the relevant scientific literature, I find it most effective in terms of the distribution of wealth, job market situation, and overall productivity of society, etc.”. While the de re belief articulates a commitment to a particular vision of marriage motivated, for example, by someone’s religion, the alternative de dicto belief can be understood in purely functional, or perhaps technocratic terms: monogamy as a rule just works (assuming for the sake of our example that this is indeed so), at least given the current socio-economic conditions, but that there is nothing more to it (according to the individual in our example). To be clear, one may hold beliefs of both kinds at the same time, but interesting cases emerge when one rejects a de re attitude and sticks with its de dicto counterpart.

The word ‘commitment’, which has just made its first appearance, is key here. It is also present in Hindriks’ considerations, but unfortunately merely implicit, thus never fully brought to light and analysed. Think of Hindriks’ thought experiment depicting Oppressia. We are told that people condemn the rule of expressing allegiance, which means that they are certainly not committed to the rule itself, but this also means that first, they are committed to not expressing allegiance, even if they are too afraid to manifest it; and second, what the regime expects of them is commitment to the rule in question.

Therefore, instead of the distinction between de re and de dicto normative beliefs, we can also provisionally speak of personal commitment-involving and personal commitment-free beliefs. While presumably, all de re normative beliefs involve commitment, this is not the case for de dicto normative beliefs which perhaps can also involve commitment, but– for the sake of this investigation– I focus on those that do not.

This would all suggest that first, the concept of belief that Hindriks works with involves personal commitment, not just an observer-made assessment of facts or probabilities, so to speak; therefore– that this concept refers to de re beliefs. However, the issue becomes less clear if we look at Hindriks’ summary of the conditions under which a social norm, defined as a rule that people expect others to believe, is practised. We read:

A social norm ℜ is practiced within P exactly if the members of P:

  1. 1.

    expect everyone to conform to ℜ,

  2. 2.

    expect everyone else to believe that ℜ,

  3. 3.

    expect that everybody is disposed to sanction violations because of 2, and/or.

  4. 4.

    believe that ℜ partly because of 2,

  5. 5.

    prefer to conform to ℜ because of 1 as well 3 and/or 4 [cooperation rule].

  6. 5*.

    prefer to conform to ℜ because of 1 and are more favorably inclined to conform to ℜ because of 3 and/or 4 [coordination rule]

The three conditions that feature in conditions 5 and 5* capture the motivating role that norms, sanctions and expectations can play. (Hindriks, 2023: 1380)

We are told in 4 that people believe that ℜ because they expect everyone else to believe in it, which moves the focal point from the notion of belief to that of expectation. Now, an analogous distinction between personal commitment-involving and personal commitment-free expectations can be provisionally drawn here as well. Imagine the two travellers and the mufti mentioned above talking about their expectations as to whether or not people take off their shoes. Or take a different example: I can expect Russia to withdraw its troops from Ukraine, in the sense that I am optimistic enough to consider it a likely scenario, but I have no personal attitude toward it, and I can obviously expect them to leave Ukraine on moral grounds, thus with a strong personal commitment, while being fully aware of how unlikely this would be. That said, if expectation in 4 involved commitment, the whole would look cohesive– my personal commitment-involving belief is motivated by my personal commitment-involving expectation– both of which are parts of my moral stance.

Admittedly, one could insist that the implication be in the opposite direction, i.e. believing in ℜ implies the relevant expectation, but at the end of the day, we can have implications on both sides, so that the relevant beliefs and expectations reinforce each other, which would be a compromise position as to what the direction of the implication is. The important point is that we would then have a cohesive system of personal commitments.

Yet if Hindriks employs the personal commitment-free notion of expectation and the personal commitment-involving notion of belief, we must somehow grapple with the claim that the former grounds the latter. For example, my personal commitment-free expectation that others will comply with ℜ, which articulates only my assessment that their compliance with ℜ is more probable than lack thereof, is said to ground my personal commitment to ℜ. However, it seems that the former can influence only my behaviour, since I will adjust my actions so as not to collide with the actions of others, or to not receive sanctions from them– and therefore a certain equilibrium-based coordination indeed emerges– but this has nothing to do with my personal commitments. Such a coordination system has no element of commitment to it. Therefore, its motivational power also becomes problematic. Consider again our example: what motivates most of the travellers to take their shoes off in the mosque is likely the sanction they would receive from a larger group of people, e.g. those praying in it, but this has nothing to do with their personal commitment to taking off their shoes in the mosque, which they simply do not consider to be their own rule. If we have both the personal commitment-free notion of expectation, and the personal commitment-free notion of belief, the whole framework is also fine, but it could also be so without 4, which merely seems to muddle things.

To wrap up this part, if we have one notion that is personal commitment-involving (or de re) and another that is personal commitment-free (thus, it’s likely de dicto), then the link between our assessment of the likelihood of people doing A, and our commitment to A, becomes a puzzle. On the other hand, if we have both personal commitment-involving notions, Hindriks’ (2023) proposal is cohesive, and perhaps it should indeed be read this way, but then the more fundamental problem of the source of commitment, or the normative strength of norms, remains unexplained. The mere mechanics of expectations does not suffice.

This is important given that ultimately Hindriks argues that only those rules which are able to motivate behaviour constitute institutions, while how strong an institution is depends in part on how motivated people are in following the rules which constitute it. And this eventually means, in my reading, how committed they are. Institutions therefore inherently involve personal commitment, thus full-blown normativity. This is a claim I shall also advocate for, but as said previously, the source of this normativity is still unclear. In the meantime I would say that this discussion of motivation smuggles in the issue of commitment, without unpacking it. We read:

In order for a social norm to have motivating force, I propose, an agent has to regard it as legitimate. Furthermore, this fact must have a positive effect on his motivation to comply with it. This requires him to have a normative expectation as well as a normative belief and preferences that are conditional on both. If these conditions are met, the fact that he believes that he ought to comply with the relevant rule motivates him to do so, at least to some extent. This means that the social norm motivates the agent directly. (Hindriks, 2023: 1380)

Regarding something as legitimate is clearly a normative act. Therefore, if a theory of social norms is supposed to inform a theory of institutions, which is what Hindriks has in mind, then the former cannot get rid of the personal commitment-involving concepts of belief and expectation. However, these concepts are not ones that the game-theoretic approach could imbue with meaning on its own.

Let me be clear at this point– my contention is not that both the game-theoretic approach and Hindriks’ hybrid proposal are wrong in principle. Instead, I argue that something more is needed in addition to them. And this something more could be provided by enactivism. But why enactivism, one could ask.

5 Inviting enactivism to the table

With regard to the rule-following approach, I have already noted my belief that more attention must be paid not only to rules themselves, but also to the actual environments in which they emerge and where they are expected to be implemented. I now propose that the same applies to the equilibrium approach. This means that the focus must not be only on behaviours themselves and their regularities, but also on environments in which organised behaviour and behavioural patterns are possible. For there is no reason to assume that all environments enable the adoption of a single course of action and give rise to regularities.

To get a sense of what is at stake here, first consider the classic traffic illustration. Note that what first makes the behaviours of the drivers possible is pre-established infrastructure: such as roads, crossings, roundabouts, etc. Second, all the drivers know what their destinations are– they have cognitive access to what they want to achieve, and above all, they consider their targets to be in principle accessible. But these are by no means benign and trivial assumptions (see Werner, 2022, 2023). In fact, the very existence of an environment in which an organised course of action is possible, thus an environment filled with accessible things and with a certain behaviour-friendly architecture built into it, is an issue of its own. This also depends on socio-economic issues, since there are places in the world where such an architecture is absent or badly maintained.

The puzzle of cognitive access also refers to the most fundamental element clarified by James Gibson (1979)– namely that the domain of behaviour as such is not a mere set of things; rather, it should be considered a realm of affordances. The latter means a domain of specific “offers” provided to agents as to what actions are possible given a particular arrangement of things, their properties, and the agent’s unique capacities. For example, a highway “offers” itself as a means for driving, but only if you have an agent who has the appropriate skill; otherwise, it is just a vast expanse of useless concrete, a mere physical object. But the domain of behaviour is not filled with mere physical objects; it is filled with useful things.

This fundamental openness of the world to the pursuits a specific agent can have, to put it metaphorically, relates to the key issue being scrutinised here– normativity. In the realm affordances some entities are seen as not indifferent, morally, aesthetically or practically– a phenomenon aptly recognised by philosophers within the phenomenological tradition (see e.g. Merleau-Ponty, 1962; Jonas, 1966; or a collection of essays by Ingarden, 1983). This non-indifference is somehow ordered, so that something may be more or less salient. We are therefore discussing a realm in which actions/behaviours are also attributed a certain “meaning”, insofar as they lead the agent to the salient (negative or positive) thing, or state of affairs, in its environment.

Now, it seems that commitments can only be justified, or as Hindriks articulates it, “regarded as legitimate”, only within a realm where things, states of affairs and actions are accessible and attributed with a certain amount of significance or meaning in the sense just sketched.

This potentially relates to all living creatures, not only humans. Arguably, the specifically human commitment to norms/rules is based on how this realm of meaning or salience evolved into ever more complex forms, up to the values present in the human cognitive niche as Clark (2005, 2006) or Pinker (2003) would put it (for more on the concept of cognitive niche see also Werner, 2020).

This introduces another fundamental issue, one that is not satisfactorily addressed by any of the prevailing concepts of norms and institutions, namely the question of what kind of creature, and more specifically what kind of system, the agent directed and motivated by rules actually is. Questioning motivation, alongside Hindriks, cannot exonerate us from digging deeper into who is supposed to be motivated.

In sum, as already noted, the question is what this realm of salience or affordances is, as well as what kind of agent + environment bond produces it.

The question of “why enactivism?” has quite a straightforward answer: this emerging new paradigm in the philosophy of cognitive science explicitly addresses all the concerns we have mentioned: the acting system (agent) with its motives, the action-related, accessible and not-indifferent environment, and the agent-environment bond.

6 Unpacking what enactivism provides

This section briefly introduces enactivism. I would first like to emphasise its selective and somewhat subjective character. There are several subtypes of the enactive approach (see e.g. De Jesus, 2015 for an overview) which are themselves considered parts of an even wider movement referred to as 4E (embodied, enactive, ecological and extended cognition; see Rowlands, 1999; Chemero, 2009; Stewart et al., 2010; Gallagher, 2017). Faced with this complexity, I shall proceed in the most straightforward way possible, outlining and provisionally defining what I regard as the core concepts needed to make the enactivist contribution to the issues discussed above both understandable and clear.

To begin with, enactivism aims to make the case that the traditional, Cartesian conceptualisation of the mind and the mind-world relationship has reached the point at which it can no longer provide fresh ideas or explanations. This conceptualisation, referred to as Cartesian Cognitive Science by Rowlands (1999), draws on the fundamental ontological dichotomy between the internal, subjective domain of thoughts, ideas, beliefs, etc., in which cognition, thought of as a kind of computation, is supposed to take place, and the external world, populated with purely physical, measurable, thus allegedly objective entities. These internal vs. external and subjective vs. objective dichotomies are responsible for the most pressing conundrum of modern philosophy, namely the question of whether there is any external world at all, given that all we deal with is composed of subjective impressions or sense data, as well as the representations built from them. And if there is in fact an external and objective world, the next conundrum relates to whether or not it is knowable (see e.g. Russell, 1912 for a classic exposition of these puzzles).

Let us examine why these speculations are relevant. First, note that when modelled solely as a type of computational or representational machine, a subject capable of cognition happens to be placed in a given environment, but their cognitive activity could in principle have been the same, had there been no environment at all. Their cognitive functioning, when locked within some sort of Matrix-style machine, would not be different from what would take place were they located in physical reality. At least this is the way enactivists see things on the Cartesian side (see Thompson & Cosmelli, 2011 for the enactivist reading of the famous “brain in a vat” though experiment; and e.g. Milkowski, 2017 for a different reading of the enactivism – representationalism strife).

Arguably, the Cartesian understanding of the internal vs. external and subjective vs. objective divides, perhaps alongside other factors, brought the theoreticians of social norms mentioned above to forego any reference to how agents experience their surroundings, and to replace this matter with a discussion of preferences and expectations, as they manifest themselves through observable behaviour. Since if we are within the Cartesian paradigm, any discussion of experience automatically has a whiff of something not truly real.

Yet if Hindriks (2023) is right that in the end it is the motivational strength of social norms which determines their becoming an institution, and their persistence, we cannot absolve ourselves from introducing an account of experience, or at least some subject-oriented “perspectiveness”, into the model. Otherwise, the motivational strength postulated would have no relation to what is recognised as salient, or at least not-indifferent, in the agent’s environment.

I interpret enactivism as an invitation to rethink the mind– world relationship itself, at its most fundamental level, and to revise our Cartesian understanding of both subjectivity and objectivity. It starts with a claim that must seem quite intuitive these days, namely that cognitive subjects or agents are living creatures. However, this is in fact a very consequential statement. First, enactivism functions based on a particular conceptualisation of what a living creature is, which harks back to the classic works of Maturana and Varela (1980). Namely, the enactivist literature adopts the concepts of autonomy and autonomous system which can be characterised as follows for the sake of this investigation:

(Def 1) For a given system to be autonomous, the following conditions must be satisfied:

  1. a.

    It is realized within a specific domain, which I propose to call the base of the system.

  2. b.

    Its components recursively depend on each other in becoming components of the system.

  3. c.

    It maintains its boundaries and these boundaries create a distinction within the base between the interior of the system and its environment.

  4. d.

    It is an unique individual, in the sense that each system has some structural features which are unique to it (the DNA sequence being the paradigmatic example).

  5. e.

    It determines all possible interactions with entities outside of the system (see Thompson, 2007 or Thompson & Stapleton, 2009).

When enactivism claims that cognition can only be found in living creatures, it makes a highly non-trivial claim, that cognition must be considered as the way in which an unique biological individual maintains its autonomy. Cognition is therefore not something which happens to the system, such as some useful tool or adaptation that the system has at its disposal in its various dealings with its environment. Instead, cognition belongs to and results from those functions which make the autonomous system what it is, which make it autonomous.

When thought through entirely, this claim leads to stark consequences for the Cartesian dichotomy between the internal and the external. In fact, it renders it obsolete. While maintaining the boundary between the interior of the system and its environment is a critical aspect, this has a dual function: making the two realms distinct, and simultaneously fundamentally connected. It is as if they were sewn together. Therefore, cognition does not just happen to be placed here or there– it is constituted by a specific location in a specific environment. This is a critical conceptual shift introduced by enactivism.

As for the very concept of environment, the difference with the Cartesian conceptualisation is also crucial. Here, “environment” does not refer to whatever is outside the sewed boundary, but rather to how a specific set of conditions for autonomy are satisfied within the system’s environment. Note that living creatures (more or less) differ when it comes to their conditions of autonomy (e.g. flies have obviously different conditions for Def1a– e than cats), which implies that those conditions determine what is relevant in the surrounding world. This is metaphorically implied by Di Paolo: “organisms cast a web of significance on their world” (Di Paolo et al. 2010: 39; this all derives from von Uexküll’s, 1926 famous notion of Umwelt).

This “casting” is part of a more complex process which enactivism refers to as structural coupling. The following is a characterisation I shall employ for the sake of this study:

(Def 2) An autonomous system is structurally coupled with its milieu iff the following two processes are realised:

  1. a.

    Sense-making: The system selects some segments of its environment, including other autonomous systems, thanks to which these selected entities are used by the system as part of the process of maintaining its autonomy.

  2. b.

    Co-determination: The selected entities determine the future states of the autonomous system. This means that structural coupling is not a linear process, but rather a looped co-determination of the subject and its system-specific environment (for more on this co-determination, see Varela et al., 1991: 150).

Given all this, cognition cannot be defined solely in terms of the internal processing of data coming from external sources, in particular due to the fact that the very notion of “external sources” makes hardly any sense without specifying the internal processes and conditions which determine a particular selection, in line with Def2(a). Therefore, as Francisco Varela, Evan Thompson and Eleanor Rosch put it in the book The Embodied Mind, which is a classic in the field, “a species brings forth and specifies its own domain of problems to be solved (…); this domain does not exist ‘out there’ in an environment that acts as a landing pad for organisms that somehow drop or parachute into the world” (Varela at al 1991: 198). The environment of a specific system capable of cognition is enacted, by virtue of being structurally coupled with the system.

(…) cognition is not the representation of a pregiven world by a pregiven mind but is rather the enactment of the world and a mind on the basis of a history of the variety of actions that a being in the world performs (Ibid.: 9).

The concept of enactment caused some frictions and misconceptions within the enactivist camp, with e.g. Pascal and O’Regan (2008) writing of its “idealistic underpinnings”, or Hutto and Myin (2013) referring to it as “extravagant.” There is no room to discuss this issue here (see Werner, 2020). Nonetheless a provisional grasp of the non-Cartesian character of enactivism is now hopefully clear, and it is necessary for a proper understanding of our discussion going forward. Enactment itself must be assumed as contextually understandable within the enactivist conceptual framework as sketched in this paragraph and in the subsequent one.

7 Participatory sense-making and the enactment of problem domains

Enactivism is also interested in group action. Hanne De Jaegher and Ezequiel Di Paolo start from the observation that coordination “is a ubiquitous phenomenon in physical and biological systems”, which is “typically easily achieved by simple mechanical means and, when cognitive systems are involved, […] it does not generally require any cognitively sophisticated skill” (De Jaegher & Di Paolo, 2007: 490). However, there are “degrees of coordination” in the domain of action (ibid.: 491). Therefore, there are also cases that cannot be reduced to these “simple” mechanics. Unfortunately, De Jaegher and Di Paolo do not further develop this idea of “simple” coordination, as opposed to a more complex one, but it seems that what they mean by the latter is the coordination of individual behaviours, producing patterns which may in turn have an impact, in the form of a feedback loop, on these very same behaviours. They state this happens within the social domain.

They therefore claim that we need to move “beyond a view that defines interaction as simply the spatio-temporal coincidence of two agents that influence each other”. Instead, we need to “move towards an understanding of how their history of coordination demarcates the interaction as an identifiable pattern with its own internal structure, and its own role to play in the process of understanding each other and the world” (ibid.: 492). The crucial point here is that within the social domain “patterns of coordination can directly influence the continuing disposition of the individuals involved to sustain or modify their encounter” (ibid.), which represents the feedback loop noted above.

The question then becomes how it is possible for this emergent pattern to influence the preferences of individual agents. I consider this puzzle as the counterpart to the one Hindriks (2023) brought to light, namely the motivational strength of norms, insofar as we consider norms to be patterns emerging from coordination.

Therefore, I shall argue that the solution proposed by De Jaegher and Di Paolo (2007) establishes a platform on which it becomes possible to also address the concerns which emerged during our discussion of Hindriks’ proposal.

First, De Jaegher and Di Paolo draw a distinction between coordination and cooperation. The former can be defined as “the non-accidental correlation between the behaviors of two or more systems that are in sustained coupling, or have been coupled in the past, or have been coupled to another, common, system. A correlation is a coherence in the behavior of two or more systems over and above what is expected, given what those systems are capable of doing” (ibid.: 490). This shared endeavour gives rise to what they call participatory sense-making. This signifies the “regulation of social coupling” which “takes place through coordination of movements”, and when “movements—including utterances—are the tools of sense-making”. This refers to situations in which “coordination affects individual sense-making” and “new domains of social sensemaking can be generated that were not available to each individual on her own” (ibid.: 497).

Here is how I propose to read De Jaegher and Di Paolo’s (2007) concept for the sake of this particular study:

(Def 3) A group of autonomous systems is involved in participatory sense-making iff each member of the group, acting for the sake of its own sense-making (as defined in Def 2(a)), coordinates its behaviour with the behaviours of others in such a way that the emergent group behaviour changes the course of its individual sense-making and impacts its efforts to maintain autonomy.

Now, going back to Hindriks’ (2023) concept of motivational strength, I propose:

(Def 4) A given behavioural pattern, which can be characterised as a state of equilibrium in the domain of action, has motivational strength for the autonomous systems involved in it, if the pattern is part of the systems’ pursuit of participatory sense-makingFootnote 2.

Being “motivated” is the cause for the perseverance of a pattern, which in turn, according to Hindriks, gives rise to institutions. In this case, motivation is not merely some mood, emotion or affection, but is grounded in the restructuring of the conditions for individual autonomy. To the effect that remaining in the state of equilibrium becomes something significant at the most fundamental level of the individual’s functioning. Simply put, systems are motivated to follow the pattern, risking disintegration by not doing so. This may not necessarily be a complete disintegration of them as individuals, since among the conditions for autonomy, there is likely some sort of hierarchy. The risk which can be attributed to the refusal to adopt a pattern, can therefore range from relatively benign perturbations to the process of maintaining individual autonomy, up to ever more consequential disruptions.

Going further, I propose that behavioural patterns with motivational strength, when approached in the enactivist manner, manifest an additional feature which I term “productivity” (ontological productivity): they enact a specific new domain.

To understand what this productivity means, let us first consider an example, although not the paradigmatic case of road traffic this time. This is because road traffic– as already noted– usually takes place within an organised space, containing infrastructure designed for traffic, and with goals determined by the agents prior to entering the game. Let us compare it to another example: a lawn surrounded by trees, located in an urban park. Now imagine a group of children playing football there. The important fact is that this lawn does not provide any proper infrastructure, either for playing football or doing anything else that is clearly defined. The domain of action is therefore as undetermined as it is possible to be in real-life cases (where it is always determined to some extent). First, this implies that the game of football itself, and thereby some of the actions performed by the players, establishes a defined space. This space is functionally delimited from the rest of the park, so that other people, who are not players themselves, will not intrude into the space in question, by coordinating their behaviour with that of the players. Second, what the players want to achieve is also set up or enacted as part of the game. Therefore, unlike in the traffic example, it is not pregiven. For example, four specific trees function as two goals. This makes something akin to “put the ball between those trees” the goal of both teams’ actions. The main point here is that the act of playing football in a park enacts a specific realm, with specific salient elements within it, but only for the duration of the game.

This exemplifies an interesting potential of group behaviour, yet one we can learn nothing about from the game-theoretic approach alone. Let us say that it is a productive potential, i.e. the possibility of creating not only a pattern but also a realm by means of coordinated behaviour. Equilibrium theory does not give a full account of how this enaction would be possible. On their own, coordination and cooperation are insufficient. Note that the other people in the park somehow also coordinate their behaviours with that of the players, in order not do disrupt the game or get hit by the ball, but they do not share the objectives or demands of the players, and therefore do not participate in the enaction of a new game domain.

De Jaegher and Di Paolo (2007) clearly suggest that participatory sense-making is productive, in the sense just described. They write that the process of interaction “emerges as an entity”, has “operationally closed organization”, and “constitutes a level of analysis not reducible, in general, to individual behaviors”. Due to these qualities, it has “a temporary form of autonomy” (De Jaegher & Di Paolo, 2007: 492).

The last point is particularly important, since it allows for a generalisation of the notion of autonomous system. With that in mind, I propose:

(Def 5) A higher-levelautonomous system is a system that.

  1. a.

    satisfies Def 1;

  2. b.

    includes autonomous systems as proper parts;

  3. c.

    the participating autonomous systems are involved in participatory sense-making.

If so, I also propose:

(Def 6) A behavioural pattern with motivational strength defined in Def 4 (thus partaking in the process of participatory sense-making of some systems) is productiveiff it is a necessary condition (at least one of the necessary conditions) of the emergence of a certain higher-level autonomous system in the sense defined by Def 5.

The set of behavioural patterns constituting the football game in our example enact a higher-level autonomous system of goals, problems to be solved, demands to be satisfied, etc. For example, since the goal of the game is to put the ball between these specific trees, scoring a goal is the enacted demand, but a problem domain also emerges relating to the question of how to put the ball between the trees.

This notion of problem domain turns out to be quite crucial. However, problem domains are usually assumed to be understandable, hence requiring no extra explanations. For example, Clark defines cognitive niche construction as “the process by which animals build physical structures that transform problem spaces in ways that aid (or sometimes impede) thinking and reasoning about some target domain or domains. These physical structures combine with appropriate culturally transmitted practices to enhance problem-solving, and (in the most dramatic cases)to make possible whole new forms of thought and reason” (Clark, 2005: 256–257).

Problem solving is the subject of numerous publications, yet the very existence of a problem is usually taken as a given. In a similar spirit, Hindriks and Guala (2019) write that institutions “constrain and enable”, meaning that they “enable societies to cooperate in ways that benefit all or at least many of their members.” Strictly speaking, “the function of an institution is to promote cooperation and thereby to generate cooperative benefits.” (Hindriks & Guala, 2019: 2). Moreover, a “particular type of institution is defined by the particular coordination or cooperation problems it solves” (ibid.: 7). However, “cooperative benefits” from solving “coordination problem” are possible only when there is a domain in which the societies already have certain goals, thus certain shared tasks, which usually means fixed problems awaiting collective solutions.

These unthematized references to problems and problem domains sound as if the latter were already there, waiting for the agents to “parachute in” as Varela et al. (1991) would put it. This certainly cannot be the case from an enactivist point of view. What a problem domain consists of depends on the conditions for autonomy, and therefore on participatory sense-making.

Going back to the game of football in the park, note that once the demand for putting the ball between the chosen trees exists, thus creating a problem domain at a given moment in time, there is never just a single way of solving it. Players can always chose one of several possible scenarios, some of which turn out to be successful. Furthermore, some degree of flexibility seems to be implicit in the very idea of a problem domain. If a particular course of action in a given situation, e.g. for a particular arrangement of players, was the only course of action possible, then the very idea of solving a problem would make no sense whatsoever– what “problem solving” could there be if there is only a single thing the agents can do? It is constitutive to a problem domain that acting subjects enjoy some level of flexibility within their environment.

Taking the above into account, I propose:

(Def 7) Suppose there is a higher-level autonomous system understood in line with Def 5 and Def 6, and there is a demand d within it, determined by the system’s autonomy conditions (thus indirectly by the pursuit of participatory sense-making involving the participating systems). This higher-level system is aproblem domain, if there is more than one possible pattern of behaviour within it that leads to satisfying d.

Of course, if we consider just two possible courses of action satisfying Def 7, we could say that the problem domain is minimal (analogously to minimal cognition as defined by van Duijn et al., 2006), but this is just a theoretical figment. The point here is that what defines a problem domain is the fact that there is some level of flexibility within it with regard to strategies for satisfying the demand, and therefore there is, intuitively, something to “think” about– thus, a problem to be addressed. This “thinking” does not have to be human discourse specifically, but it does imply some relatively complex level of cognitive processing.

8 Institutions: meta-patterns and social norms of “letting be”

We can now get closer to our final goal, which is to improve the hybrid (game–theoretic + rule–following) understanding of institutions with the help of enactivism. For this goal to be attained, the concept of a social norm needs to be reintroduced.

But for this reintroduction to be successful, we need to employ a more recent proposal by De Jaegher, namely her concept of letting be. This is a specific pattern of behaviour which contributes to complex participatory sense-making (see De Jaegher, 2021) in situations in which agents “need to deal with the double-sided risks of determination. They need to deal both with being-determined (by the other, by the relation, and by themselves) and with determining (the other, the relation, and themselves)” (De Jaegher, 2021: 861).

While the nature of this “determination” is not fully explained, it could perhaps be defined in the following terms:

(Def 8) Suppose there is a set of autonomous systems acting in a fixed problem domain. We can say that the behaviour of the autonomous system a1 is determined by another autonomous system a2iff the range of possible courses of individual action available to a1 decreases as a result of specific actions taken by a2.

Arguably, some level of determination is inevitable if the agents are to coordinate their behaviour, to enact a problem domain in the first place, not to mention solving problems within it. And yet, in my reading, De Jaegher (2021) is profoundly right when she points to the risk of overdetermination. This can be understood as determination to the extent it becomes detrimental to problem-solving and even to the very pursuit of the participatory sense-making responsible for the emergence of the relevant problem domain. Note the subtle difference between the two: overdetermination can cripple the ability to solve problems, but it can also reach deeper and interfere with the very capacity to enact the problems to be solved. This last issue in particular would deserve its own separate study.

In this context, De Jaegher (2021) writes that participatory sense-making is a “continual, ongoing balancing act between too much and too little determination” (ibid.). We now arrive at the crucial notion of “letting be” as a course of action, which boils down to an individual agent not overdetermining or abstaining from the overdetermination of the behaviour of other agents, thus allowing them to choose a behavioural patternFootnote 3.

This concept is particularly relevant when solving relatively complex problems. It may sound quite familiar to those with some knowledge of the literature in management science, for example on “Agile development” practices and strategies (see Appelo, 2011; Mitchell, 2016). It provides examples of instances in which the only viable way of tackling a highly-complex problem, e.g. in programming, is to let all members of a team work freely by exploring various scenarios, including out-of-the-box ones. This strategy, which at first glance might merely look like lack of control, in fact makes the best possible use of individual talent.

In my reading, the crucial though somewhat underrated point is that for the strategy of “letting be” to be implemented, there needs to exist an environment in which abstaining from overdetermination is feasible, thus one in which the individuality of problem-solvers can manifest itself in a way that is beneficial to the overall process of problem-solving. We therefore have the issue of what social factors make it possible for problem-solvers within groups to navigate between underdetermination and overdetermination, while avoiding both extremes. This can prove quite challenging in cases of larger groups, in which agents do not know one another, unlike smaller task teams where people rely on their knowledge of one another.

I would like to suggest that to avoid the under- or overdetermination of behavioural patterns in complex problem domains, some meta-level patterns need to emerge (a similar idea can be found as part of the enactivist account of language proposed by Di Paolo et al., 2018; see also Werner & Kiełkowicz-Werner, 2022 for an application of this idea in value theory). This implies the need for patterns of letting be. Such meta-patterns would stabilise the strategy of “letting be” as a social practice.

Therefore, I propose, there exist behavioural patterns at a higher level (meta-level) governing the behavioural patterns of the lower level (object-level). The lower-level patterns deal directly with the fixed problem domain, meanwhile higher-level patterns are patterns of dealing with these lower-level patterns. The idea is thus that meta-patterns are crucial if the complex balancing between underdetermination and overdetermination is to be achieved and– which is particularly important– stabilised. This stabilisation means that the balance is not momentary, nor an achievement limited to a short duration, but a relatively robust part of social practice; that it is indeed a pattern of patterns. Precisely in this context, I propose to speak of social norms.

Let us now proceed step-by-step with the reintroduction of social norms. I first propose:

(Def 9) In addition to behavioural patterns devoted to solving problems in a fixed problem domain, which may be regarded as object-patterns, there are meta-patterns established over the course of participatory sense-making, whose function is to make it possible for autonomous systems to avoid over-determination (to balance overdetermination and underdetermination) in their pursuit of problem-solving. These meta-patterns are called patterns of “letting be”.

However, what is not included in Def 9 is an agent’s awareness of these meta-patterns. In principle, there may be factors regulating the pursuit of solving of complex problems and adopting the “letting be” strategy, but which are not recognised as such by the agents themselves. I would rather not speculate as to whether this is the case in some cognitively advanced non-human animals, although I find it likely.

But when it comes to humans, we can speak of recognised or acknowledged meta-patterns of “letting be,” which going back to the conceptualisations proposed by Bicchieri (2006) and Hindriks (2023), can be defined in terms of beliefs and– crucially– social norms.

I propose:

(Def 10) Suppose we have a problem domain and a complex problem p within it, such that solving p requires a “letting be” strategy. The meta-pattern ℜ (this notation refers back to that of Hindriks, 2023) stabilises the realisation of a “letting be” strategy aimed at solving p and therefore functions as a social norm iff:

  1. a.

    every autonomous system active in this problem domain– every conscious agent– is motivated in the sense defined in Def 4 to solve p;

  2. b.

    every agent believes that ℜ contributes to solving p indirectly, which means that ℜ specifies a set of potentially viable behavioural patterns aimed directly at solving p;

  3. c.

    every agent believes that all others are also motivated in the sense mentioned in point a;

  4. d.

    drawing on the motivation and beliefs listed in points a– c, every agent is ready to sanction violations of ℜ (since violations bring disruption to the process of solving p and the overall process of participatory sense-making in the problem domain).

[A note:] Sanctions can be defined in terms of overdetermination: at its most basic level, sanctioning means a purposeful overdetermination of the behaviour of an individual agent who refuses to adopt ℜ voluntarily.

  1. e.

    every agent applies the conditions c and d to itself (which stands for the agent’s preference to comply with ℜ).

  2. f.

    the agent’s preference to comply with ℜ, characterised in e reinforces– when it is successfully realised– the agent’s realisation of the conditions a– d, which means that it reinforces the relevant motivations, beliefs and readiness.

Let us note several points here. Def 10 is clearly based on Hindriks (2023), I therefore propose it can indeed be taken as an enactivist contribution to his proposal and to institutionalist literature in general. However, unlike in Hindriks’ original list of conditions, motivation is not merely left unexplained, in the background of the concept of social norms, but is introduced explicitly in Def 10a referring to Def 4. Thanks to this posit, we can better determine what it means for an agent to be motivated. Namely, in the final analysis, it can be cashed out in terms of the agent’s maintenance of its own autonomy and of the very existence of the problem domain.

With this in mind, we can then introduce the agent’s recognition of the norm. Not in terms of commitment-involving beliefs or expectations, whose justification comes from outside the model, but in terms of beliefs as to how ℜ contributes to the shared attempt at solving the problem. The normative character of these beliefs is internal to the situation described in Def 10, meaning that it is grounded in the very nature of participatory sense-making.

Furthermore, sanctions, also left with their intuitive meaning in the standard approach, and defined here in terms of purposeful overdetermination, are mechanisms for avoiding disruption in the pursuit of participatory sense-making. In this sense, they are not ad hoc punishments, but again something grounded in participatory sense-making.

An agent’s preference is also a kind of belief. Yet this time, according to Def 10(e), a self-referential one (which accounts for the fact that an agent may fail in fulfilling its own expectations).

Finally, Def10(f) introduces the feedback loop, which in principle draws on the successful application of ℜ. This means that once the meta-pattern ℜ is realised and actually contributes to solving p, it reinforces its further use. In the end, the persistence of ℜ, emphasised in particular by Hindriks (2023), and rightly so, is also explained within the proposed framework and it boils down to ℜ’s contribution to the process of problem-solving. Crucially however, since the entire looped process plays out in time, if ℜ is no longer beneficial to the pursuit of solving p, there may so to speak be delays in the recognition of this fact. Therefore, even a meta-pattern that becomes completely useless can still persist for some time (depending on additional factors that are not part of the model). This last observation improves our understanding of the persistence of inefficient or dysfunctional institutions, studied by economic historians, among others (see Acemoglu, 2006; but also e.g. Fricker, 2007 on epistemic injustice, which proves relevant, too).

That said, one easy way to introduce institutions would be to state that they are social norms with sufficient motivational strength, or sets of correlated social norms, or something else along these lines. I propose a different approach though: institutions are not norms; they are realms enacted by social norms. Therefore, they are higher-level autonomous systems; specific environments or specific problem domains.

To see the intuitive rationale for this move, let us revisit the football game example again and compare it to the standard traffic one. As already mentioned several times, traffic takes place within a pregiven infrastructure, and in fact, this infrastructure almost dictates to the agents what to do. The important part of the football game illustration is that while the players do build upon the pregiven infrastructure of the park, they also introduce a game-specific environment. Their behaviour is therefore productive, in the sense defined in Def 6. Of course, this particular domain is very simple and ceases to exist once the players go home, but I propose that what applies to such simple cases also applies to much more complex instantiations of behavioural patterns and relevant meta-patterns, including the “letting be” one. This implies that insofar as a particular “letting be” pattern is part of participatory sense-making, it can also be productive, i.e. contribute to the enactment of a given problem domain.

This point deserves to be emphasised. The process of solving a specific problem in the already-existent (one might say enacted earlier) problem domain gives rise to the “letting be” strategy, but this very strategy– the “letting be” meta-pattern– is also productive, in the sense that it also can give rise to a given (perhaps still “higher-level”, so that we have a more complex hierarchy of these “levels”) problem domain.

Given this fact, we can define an institution, based on the social norm ℜ, in the following way:

(Def 11) An ℜ-institution is a problem domain D*, enacted on the basis of an already-existent (previously enacted) problem domain D by the process of participatory sense-making and problem solving, such that the necessary conditions of emergence and persistence of D* (its conditions of autonomy), include the meta-pattern ℜ of “letting be”.

What do we gain from Def 11 when compared to the idea of identifying institutions with social norms that merely satisfy some additional conditions (e.g. motivational strength, as emphasised by Hindriks, 2023)? We gain three things, each of which contributes to ongoing debates and therefore constitutes a separate research topic. I will therefore mention them as potential developments of my proposed framework.

First, Def 11 does indeed draw a connection between an institution, emerging or implemented in a specific place, and the specific local conditions existing there, which can prove quite problematic when approached in the standard manner. For example, if we say that institutions can be reduced to behavioural patterns or norms, such as the ones governing road traffic, we ignore the specificity of the local infrastructure where this example of traffic takes place. Meanwhile, the unique qualities of a particular infrastructure can significantly impact behaviour. Therefore– crucially– different local conditions can create different demands for institutions, thought of as norms. Meanwhile, if institutions are defined as domains enacted on the basis of some pre-existing local conditions, created by local processes of participatory sense-making, then on the basis of specific problem domains existing there, they must respond to these local conditionsFootnote 4. Otherwise they would not have been enacted in the first place (which does not imply that they answer these conditions successfully; this is an entirely different issue). Hence, Def 11 enables us to descend from the domain of allegedly universal abstract patterns which maintain no connection to local circumstances (or this connection is somewhat undetermined, as in the case of the rule-following approach), to a more suitable domain of abstraction (since speaking of patterns is necessarily abstract as such)– closely tied to concrete local conditions. Apart from mere illustrations such as road traffic or the football game, this can also apply to real-life examples of institutional policies and contributes to the debate on informal institutions that support or stifle economic growth in various countries or cultures (see Acemoglu & Robinson, 2013; Mokyr, 2016), or whether universally applicable formal institutions, necessary for economic prosperity, can exist, and how these universal “recipes” for economic growth (if any) get along with local patterns of behaviour (see e.g. Ostrom, 2005 on the dangers of “institutional monocropping”, as she calls it).

Second, Def 11 clarifies the concept of the demand for an institution. The point being that it is not just the need for a behavioural pattern, quite ubiquitous within the animal kingdom, but a need which arises in problem domains, in which some level of flexibility within the problem-solving process is required for the task to succeed. In other words, institutions emerge where the “letting be” strategy is required to get the job done. This is not supposed to mean that humans are the only creatures capable of the latter, yet it does mark a significant threshold– institutions emerge as part of very complex environments created through structural coupling. This provides the opportunity for an evolutionary account of social phenomena which is not reductionist. Yet even more importantly, the conceptualisation I propose, composed of De Jaegher’s “letting be” and its meta-patterns, allows us to note a particular dynamic within human problem domains. In particular, I’d hypothesise that by referring to the issue of a proper balance between underdetermination and overdetermination, we can account for inefficient institutions, or dysfunctional forms of institutions (such as those in totalitarian regimes). Namely, it seems possible and potentially plausible to characterize them as having gone too far in either direction, given the specificity of the problem they were supposed to solve. That said, the issue of institutional resilience, widely discussed today (see e.g. Aligica 2013) can also be apprehended anew, where this resilience could be seen as stemming from a proper balance between the two extremes of determination.

Third, still in connection with the problem of locality, Def 11 finally does justice to an issue barely noted in the standard literature, namely that institutions can be seen as specific loci– one can be placed in or within a certain institution. For example, if I am in the United States, I not only occupy a specific location within the vast territory of North America, but I am also placed within the boundaries of a specific political entity, thus a complex web of institutions which constrain and enable (as Hindriks & Guala, 2019 would say) my behaviour in specific ways, which are different from the ways in which my behaviour is framed within the institutional boundaries of China, Italy, or France (or the EU as a whole). To obtain a sense of this locative aspect of institutions we can think of the phenomenon of migration. First, while a person who decides to leave one country and move to another may have a variety of reasons for doing so (a job offer, the climate, family issues, etc.), some of these reasons do involve social norms and behavioural patterns, and therefore institutions.

9 Conclusion

Rules + equilibria theories of institutions, in particular that proposed by Hindriks (2023), were diagnosed as having issues, described in the first part of the paper with regard to normative beliefs and expectations, relating to the normativity of institutions. To address this, a distinction between de re, de dicto, personal-commitment-free, and personal-commitment-involving beliefs and expectations was introduced. This paper proposes that these models of norms and institutions do not engage seriously enough with the issue of what kind of being (or more precisely system) an agent is and what the relationship with the environment, offering agent-specific opportunities for action, of that agent is. For that reason, I proposed that current theories of institutions should work in tandem with the emerging enactivist paradigm in the philosophy of cognitive science, which conceptualises a subject capable of action and cognition as an autonomous system, structurally coupled with its environments and involved in the pursuit of participatory sense-making.

First, institutions were approached as answers to specific problems. With that in mind, problem domains were defined in terms of higher-level autonomous systems composed of selected, agent-specific parts of the world. Moreover, the concept of a “letting be” strategy, applicable to the pursuit of solving complex problems in a shared environment, became the backbone of the proposed understanding of social norms as behavioural patterns. These are patterns which do not directly address a given problem, but govern the problem-oriented behavioural patterns. They are therefore meta-patterns, or patterns of patterns. Such meta-patterns are necessary, I argued, if there is a certain level of flexibility at the level of problem-oriented patterns, with individual agents to a degree free to choose their own ways of solving problems. Institutions were therefore defined as higher-level problem domains, thus domains of action, enacted by social norms, considered as meta-patterns of ‘letting be.’

All the conceptual posits in this paper aim at fostering a dialogue, which is currently quite lacking, between the literature on institutions within the philosophy of social science on the one hand, and enactivism, or the embodied cognition movement more broadly, on the other. In addition to purely theoretical gains, this dialogue is particularly required given the three areas of empirical research: the problem of the locality and universality of institutional solutions devoted to economic growth; institutional resilience, efficiency and persistence; and the institutional aspects of migration. I presume that a more suitable concept of institution would be beneficial in addressing these issues.