What is an institution, that people may participate in it?

In a recent post, Christophe Heintz told us about “institutions that make us smart”. The posting was of great interest by itself, and got us thinking about institutions – our field still has its work cut out if we want to make sense of institutions.

We have all sorts of interesting tools and theories in cognitive anthropology (as that should be the name of our field – see discussion here), but precious little to say about institutions. That is, to put it mildly, rather unfortunate as many social phenomena seem to depend on the interaction between people and their institutions. Cognitive anthropology should be the place to understand why and how people are compelled by institutional arrangements such as money, marriage, law courts, schools and markets. So far, we cannot say we have made great progress in that direction.

The sketch of an epidemiological account

I had reached this point in my desultory reflections when I realised that I had been scooped, and that there actually is a rather interesting account of institutions, the best formulation of which can be found, again, in an article by Christophe Heintz. The model develops arguments originally presented by Dan Sperber in other texts. The main point here is that institutions can be studied in the framework of the deflationary ontology of cognitive anthropology. In the same way as “cultures” are patterns of distribution of mental representations in groups of people, “institutions” can be defined and studied without assuming that they are a separate level or domain of reality.

Sperber and Heintz characterise institutions in terms of, would you believe it, an epidemiology of representations. In particular, institutions are sets of representations that include normative representations about the distribution of representations. As Heintz puts it:

"Institutions involve at least two levels of representations: a higher level of regulatory representations and a lower level of representations whose content and distribution is in accordance with regulatory  representations. […] Understanding the role of institutions in cultural evolution then consists in describing the causal chains that link the mental representations and the public productions of the institutions together, and accounting for the causal effect of the regulative representations."

To take the example of law courts, these work because of distributed representations (about what the laws are, what the definitions of various actions are, what constitutes property, etc. etc.) as well as normative notions about who holds what representations (defining the various roles of plaintiff, judge, defendant, etc., as well as scripts for their interaction).

This account of institutions illustrates a reductionist approach to the ontology of social phenomena. Which is a Good Thing, of course. A standard anti-reductionist rejoinder is that there are many instantiations (low-level realizations) of a given higher level entity, institutions too. So there may be indefinitely many ways in which different participants mentally represent the workings of the law-court, with no overlap at all between these individual versions – so the judicial system exists independently of these mental representations.

Let us leave aside this question (which is actually no problem at all, and can be quickly solved by any social scientist worth his/her salt), and turn to cognition.

The standard economic account

The Sperber-Heintz account is quite interesting – and perhaps even valid! But it is also, to put it mildly, a tad abstract – it will need considerable fleshing out before we can say that cognitive anthropologists have a viable framework to understand institutions. In particular, it seems to me that this is a place where cognitive anthropology could start where economics left off, and solve complex issues that institutional economists consider out of their bailiwick.

The tradition of economic thought called “institutional economics”, founded by Douglass North and others, described institutions as “rules of the game”, either formal or informal, that regulate property rights, the enforcement of contracts and other such indispensable elements of economic growth. For instance, commercial law (formal) and scratch-my-back-I’ll-scratch-yours reciprocity (informal) are among institutions in this sense. The good thing about institutions is that they reduce transaction costs – given that the latter include the cost of contract-enforcement.

Since these rules benefit everybody but are also costly to design and enforce, who provides them? Institutions invariably include constraints on the behavior of participants, so how is that solved? Finally, free-riders should be avoided, and punishing them is also costly.

The various answers in the neo-institutional literature are in the form of historical development. The idea is that in some special circumstances, a slight change to previous institutional arrangements, e.g. in the direction of better property rights, can accrue prospective benefits to may agents – enough of them to agree to enforce the modified rules. This suggests that a long historical development can result in such arrangements. But economists generally agree that these suggestions do not amount to an overall account of institutional development.

Everyone gets institutions wrong

If the reductionist agenda is to be followed – and it is, it is – we will have to address the paradox, that most people who participate in institutions seem to have the wrong ontology! For instance, in many human groups we find marriage institutional arrangements, that we can perfectly well describe in the Sperber-Heintz way as combining shared representations about duties and rights, together with a set of normative rules about the distribution of these representations. Fine (although pretty difficult to achieve), but this account typically clashes with the participants’ intuitions. For most people in the world (except a few social scientists), marriage is something that is not just the occurrence of particular concepts and norms in people’s minds. Marriage is also… well, at that point people typically become rather vague, but they generally feel that the compelling nature of institutions shows that they are, somehow, in a mysterious way, “more” than the particulars of people’s specific social interactions.

In many cases, this intuition of “something more” adds great support to other assumptions, e.g. that the gods or ancestors created the social institutions, or that social norms are eternal and inviolable. But you do not need supernatural beliefs to stick to the non-reductionist intuition. In the US, for instance, a substantial number of people are all in favour of granting gay folks all the rights they want , including the standard form of legally binding union, except that that should not be called “marriage”. This nominalist anxiety is immune to all standard liberal arguments (e.g. “How does it change your marriage, or any other heterosexual marriage, that two same-sex people also call their union a marriage?”) because it is not just about words, I suppose, but (among other things) about something much more anxiety-inducing, the suggestion that institutional arrangements are nothing else than what we decide they are.

A similar form of intuition also applies to money. Even though money works only through the distributed trust of many people in the exchange value of some debt, it is often considered to have its own value beyond that. One could, I think, consider many other domains and find that people generally assume some form of non-reductionist ontology of institutions.

And they are, of course, very, very wrong. But why is that illusion so pervasive?

Do people have to be wrong about institutions?

There may well be functional aspects to the notion of super-organic institutions. One could speculate that this kind of cognitive illusion is necessary to coordination. The latter would in principle require that agents have accurate representations of each other’s goals, norms and concepts. In any group larger than bands of foragers, this would constitute a computational problem well beyond ordinary human capacities. People who represent norms and institutions as supra-individual can use them as a (usually very reliable) instrument to predict other agents’ future behaviors. The scientifically valid explanation of the workings of law courts is just massively non-intuitive and computationally expensive, while the notion of a “justice system” divorced from people’s singular circumstances is cognitively cheaper.

This would make institutional beliefs rather similar to other phenomena, like the pervasive anthropomorphization of social groups. It is almost impossible to discuss large-scale social interactions without using the idiom of intentional psychology (“the committee wanted this…”, “the government will not accept that…”, “the bourgeoisie has had to accept that…”, etc.), although any decent social scientist can tell you that this is both ontologically misguided and predictively very poor.

But a useful illusion has to come from somewhere. Since we evolved in the context of small groups without complex social institutions of the kind discussed here, what elements of our fitness-driven cognitive tool-box make it possible to think of super-organic social institutions?


  • comment-avatar
    Olivier Morin 3 March 2009 (15:47)

    Pascal asks how we, as a species, could have acquired the intuitions that we have about institutions, since we evolved in an environment without states, churches, armies, and such like. I think a possible answer might be found in the way we treat simple cultural objects like songs, games or recipes. These would not be considered institutions by economists, although philosophers of social science might accept them as such. People have puzzling intuitions about some cultural objects, and these intuitions also have to do with a belief that a cultural object is somewhat “more than” the total sum of individual representations of that cultural object. For example, people may claim that some versions of the story of Batman are more genuine or more valid than other versions, even while knowing that there is no reality whatsoever to the story of Batman, and consequently, no fact of the matter when it comes to deciding who’s right or wrong about Batman. I remember vividly a dispute I had with a Spanish housewife, concerning the essence of paëlla. She held the view that paëlla, as cooked by Spanish housewives outside Valencia, was not paëlla at all but rather some indefinite dish not worth naming. One reason was the fact that these “false” paëlla could sometimes include meat. Actually there is nothing to paëlla but the total set of practices and beliefs that people identify as paëlla. If one is an authorized member of the group of people recognized as competent practitioners of paëlla (for example, if one is a Spanish housewife), it is hard to see how one might get paëlla “wrong”. Yet people vehemently maintain that cultural practices, like paëlla, have essences, and these essences are distinct from the ideas that authorized people have about them. As luck would have it, the people who hold this view also happen to have excellent access to the essence of the practice they are talking about. You seldom encounter someone who claims that there is a One True Recipe for Paëlla, but does not claim to know the truth about it. In the philosophy of social sciences, this kind of (if I may) superstition is mainstream; it goes by the name of “constitutive rules”. Some social practices, the argument goes, are necessary features of paëlla, marriage, chess, baseball, etc. You just cannot decide that you are going to cook a paëlla, celebrate a marriage, play chess or baseball, etc. if you do not include them. Unfortunately, people can, and do, forget features that other people see as necessary and constitutive. There are paëllas with chorizo in it (anathema to my Spanish housewife), there are weird marriages involving same-sex couples, or lack of consent, and there has been as many twisted versions of chess and baseball as you can imagine in the history of these games. This comment is already too long, so I won’t speculate on what causes this habit of thinking. But I think it is credible as a precursor of our tendency to view institutions like states, churches, armies, etc. in a holistic light.

  • comment-avatar
    Christophe Heintz 4 March 2009 (23:49)

    I cannot but agree with Pascal that the epidemiological account of institutions is “a tad abstract – it will need considerable fleshing out before we can say that cognitive anthropologists have a viable framework to understand institutions.” However, a reductionist and cognitive account of institutions does not need to start from scratch. As I understand it, the epidemiological account of institutions adopts a reductionist standpoint by trying to understand institutions in terms of thoughts and behaviour of interacting people. Rather than taking institutions as social entities on their own, the epidemiological account would talk about institutionalised behaviour or practices. These practices are institutionalised to the extent that they are regulated by some well-distributed representations. Thus, institutions are analysed in terms of recurrent behaviour or practices and the representations that regulate them. Whence the research agenda: finding out which representations have a regulative effect, and how and why they have this effect. My optimistic note comes from the fact that there is already some work in the social sciences that is compatible with the epidemiological account and that provides some flesh to it. There are some works in the social sciences that look at the regulative effect of some representations. Here are examples: 1. Barnes and Bloor have developed a reductionist account of institutions and analyze one regulatory mechanism through which some institutions are maintained. It is the fact that representations referring to an institution often contribute to its maintenance. For instance, talking about the state of France contributes, even if minimally, to the existence of the state of France. Barnes and Bloor point out that our representations about institutions are, to some degree, self-referential. Take money as an institution, for instance. Representations about coins *as money* are constitutive of the institution of money. Without such representations, coins would just be physical tokens without symbolic value. So representations about coins as money are also representations about others and oneself attributing value to coins. The representations are about representations of coins as money and thus, to some extent, about themselves. Such self-representations have a regulative effect: representing coins as money encourages others to do so; talking about the state of France causes others to have representations of the state of France. 2. Think of organisations as institutions that regulate the distribution of labour. If that is the case, then we have a rich pool of work in the sociology of organisations, which is about the mechanisms through which organisations are created and maintained. For instance, leaders cannot always strictly implement their ideas about how to organise labour. In other words, the regulative effect of leaders’ design representations is limited. It is complemented by negotiations, day-to-day social interactions, tacit rules with non-hierarchical origins, and so on. Key authors in theories of organisations who are close to the cognitive approach include Simon, March and Crozier. 3. I agree that the New Institutional Economics is one place where to look. I believe that studies in this field have mainly used the model of the homo oeconomicus rather than some more psychologically realistic model of man. However, the recent book of Douglas North (Understanding the Process of Economic Change, 2005) explicitly goes in a more cognitive direction. For instance, structures of incentives determine agent’s behaviour only to the extent that agents perceive and understand which incentives there are. A law about commerce, for instance, have a regulative effect when most people obey the law because they have understood that one should obey it. As Pascal Boyer, I see here a call for cognitive anthropological studies. Structures of incentives vary across cultures, and these variations are themselves the result of human choices and influence further choices and the underlying cognitive processes. The epidemiological account of institutions is a theoretical framework that needs to be fleshed out with empirical case studies and more specific analyses about the nature and effects of regulative representations. The good news is that there is already some work out there that can be taken and used for the epidemiological research programme. The second good news is that there is still much to do. Among the questions that call for empirical research, Pascal raises the one that concern the origin of naive social holism.

  • comment-avatar
    Christophe Heintz 5 March 2009 (00:03)

    Pascal Boyer points towards one important direction of research that starts with the observation that people have intuitions about institutions. These intuitions are often strongly essentialists, as is nicely illustrated by Olivier’s Spanish housewife. For the epidemiological project, there are two research questions: a) How do these essentialist intuitions about institutions contribute to the formation and maintenance of institutions? b) Through which cognitive processes do the intuitions about institutions arise? In particular, which are the cognitive abilities involved and what is the evolutionary history, if any, of these abilities? With regard to the question a), I would say that these essentialist intuitions about institutions contribute a lot to the formation and maintenance of institutions. The work of Barnes and Bloor (point 1 of the previous post) can be interpreted as saying just that: essentialist representations of institutions are constitutive of institutions. In most cases, thoughts and behaviour regulating institutions seems to originate in mental representations about the regulated institution. For instance, it is because a director of an organisation has an idea of his own organisation that he will take actions for regulating the behaviour of people working in it. Could there be an institution such that the social actors participating in it do not have a (essentialist) representation of it? Well … I have no example in mind at the moment, and I’d be curious to know if anyone could suggest one. The epidemiological characterisation of institutions makes no mention of representations of institutions. So showing that they are always present would be a nice empirical result within the epidemiological framework. I’m looking forward to more research being done about b)! One possibility is that naive social holism comes for free with psychological essentialism. Then the theory of mind is recruited to reason about social entities. But note that it is also recruited to reason about other non-human entities: scientists often tend to attribute, informally, intentions to the object they study (e.g. “the electrons want to / try to …” ). Admittedly, we also have the ability to ‘perceive’ emergent properties and it seems we can reason quite well about social phenomena.

  • comment-avatar
    Christophe Heintz 5 March 2009 (00:25)

    The erroneous argument according to which agents’ naive social holism implies scientific social holism is nicely formulated by Thagard in ‘Societies of minds: Science as distributed computing’ Studies in History and Philosophy of Science, 24:49–67, 1993. (available at [url]http://cogprints.org/676/0/Societies.html#anchor04[/url]). Arguing against reductionism and methodological individualism, Thagard quotes Durkheim on social facts, then writes: (Do not be confused: in this paragraph, Thagard talks about the sociology of science. So when he talks about scientists, he talks about the social agents studied by the social scientists. His scientists are our laymen who have a naive social ontology – not to be confused with the scientific view of social entities). [quote] Much of the plausibility of methodological individualism comes from the ontological point that societies consist of their members, but this does not imply that we can explain the operation of societies by attending only to the behavior of their members. Consider the explanation of why a particular scientist thinks and behaves in certain ways. I presume that the explanation will take the form of a description of the computational apparatus producing the scientist’s thinking, which includes a representation of what he or she knows. […] much of the scientist’s thinking will require the representation of social organizations […] Thus if we were building a simulation of the cognitive processes used by the scientist in decision making, we would have to represent information about social entities such as agencies, departments, fields, and conferences. Thus psychological explanations of scientist’s thinking are in part sociological. My point here is supplemental to the one about social facts: a full model of science must not only take into account the ineliminable organization of science, but also the mental representation of that organization by scientists. [/quote] I think this cannot be an argument against reductionism or methodological individualism. One can study the myth of Santa Claus without postulating the existence of Santa Claus. One [b]can[/b] study people’s ideas about, say, a scientific department, without having the scientific department in the explanans. Why did an individual come to have this representation about the scientific department? Because he saw a building and a sign on it where it was written “department of …”, because people talked to him about the scientific department, etc. Only representations about the scientific department may need to figure in the explanans. Of course, this does not mean that institutions should never figure in the explanans of sociological explanations. Just that it is, in principle, possible to replace them.

  • comment-avatar
    Dan Sperber 9 March 2009 (11:36)

    How nice to see old ideas of mine on institutions (first toyed with in the early 80s, including with Pascal Boyer) that have failed to draw attention over all these years now being interestingly expounded by Christophe Heintz and defended, together with relevant questions and caveats, by Pascal! Just one point of clarification. Pascal writes: [quote]”For most people in the world (except a few social scientists), marriage is something that is not just the occurrence of particular concepts and norms in people’s minds.”[/quote] I am sure Pascal got it right in his mind, but this formulation might lead to minsunderstandings. The epidemiological approach involves, to begin with, a reconceptualisation of the social and proposes new concepts that are generally in the vicinity of folk and standard social science concepts but that involve a different ontology. So, we just do away with “marriage” as understood by the common folks (including by ourselves in our non-profesional life) and by fellow social scientists. Simple put, nothing in the world instantiates what folk or standard sociologists understand by the term. We are interested in a different kind of object, one that [i]is[/i] instantiated and the occurence of which gives folk and standard sociologist the mistaken idea that there is an institution of marriage, that some people are married (and other divorced, widowed, and so on, all these predicate that presuppose prior marriage actually lacking instances). The object we are interested in is not properly decribed as “the occurrence of particular concepts and norms in people’s minds.” It is rather [i]the causal chains[/i] that distributes not only concepts (of marriage in particular) and (representations of) norms (about how one becomes married and what follows, and about specific instances of people being now married, and so on) in a social network, but also practices and artifacts. I insist, our objects are primarily the causal chains themselves, not the representations, practices and artifacts they distribute and the occurence of which they explain. Now, it takes some getting used to truly think of cultural causal chains and even more to use them in scientific work, and yes, it is all deucedly abstract at first, but that is the way to go (or so I believe), so, if you like this approach to institutions (or just would like to understand it properly), work on it! The causal chains that, inter alia, distribute representations of marriage in general and of specific marriages in particular have to a large extent effects similar to those folk and standard sociologists wrongly attribute to the institution of marriage and to specific marriages. So, we can easily help ourselves to the limited but real competence and knowledge accumulated by folk and standard sociologists and reformulate what is of use to us in it in epidemiological terms. But note, this is not a matter of changing the vocabulary, it is a matter of getting rid of the folk and standard sociologists’ ontological superstitions and of genuinely reconceptualising the field. And what is the use of taking all this trouble? Well, that is where and when the serious causal-explanatory work begins.

  • comment-avatar
    Maciek Chudek 16 April 2009 (14:07)

    Humans do a lot of interesting things, one of them is naming things. There are infinitely many things in our environment we could name, but we pick out just a (relatively) few interesting ones. Pascal’s question about institutions may be putting the cart before the horse – that we essentialise institutions after we’ve named them isn’t so surprising, essentialising any named thing may be a simple constraint to make communication tractable. The really interesting question is how, if institutions are such complex constellations of distributed representations whose characterisation is “non-intuitive and computationally expensive”, we come to pick out any particular institution and share a name for it in the first place. This is a two part problem. The first step is, of all the variants of all the representations/norms/concepts in people’s minds, identifying one that is consistent and stable enough that it’s worth naming. The second part is getting other people to understand what the heck you’re talking about. If Pascal’s suggestion is right and we’re cognitively adapted to a Pleistocene past riddled with institutions, we may just *see* institutions intuitively – picking out and essentialising their pattern the way we pick out discrete objects in our visual field. This would certainly solve both problems: if everyone’s effortlessly perceiving all the extant institutions it’s easy (or much easier) to identify the most important ones, name them and point them out to others. This just doesn’t jell with my experience. Unless I put on my academic hat and try really hard, it’s very difficult for me to perceive any institutions but those we already have names for, and those I perceive very easily. I may just be strange, please let me know if this isn’t true of you. I’m afraid I don’t know of any more rigorous investigations of this question. In either case, there may be a simpler solution that doesn’t involve custom cognitive adaptations. If children intuitively try essentialising any named thing, and only keep using words for things they find useful, culture can quickly fill with labels for institutions which deserve names (i.e. naming them is useful) and which make handy heuristics when essentialised (i.e. essentialising them is useful). The second problem is solved because these, as long as they’re heuristically useful, are easy for the next child to pick out. The first problem might be solved, as in genetic evolution, by noise – such as the constant slew of new labels and concepts being generated by nuts and crackpots. Many of these will be useless, but the few that essentialise to handy heuristics will be picked up and integrated into culture. Accounting for institutions may not require its own adaptive, cognitive account – just the economic definition of institutions, plus an account of why people essentialise any named thing and why named things tend to be heuristically useful.