Solving Lewis and Hart’s Problems with Normativity, Evolution, and Disagreement

Introduction

David Lewis and HLA Hart have very similar problems with their accounts of social conventions in Conventionand The Concept of Law, respectively.  They both want to capture an obligatory character of social conventions. They both see an obligation to comply as being an essential feature of social conventions.  They both think that the only way to capture this obligatory character is if there is a single rational thing to do in a situation that gives rise to a social convention, which is to conform as everyone else in one’s population conforms.  They both think that if there is a single rational thing to do, then you have something very closely akin to an obligation to do that one rational thing.  But, these Lewisian social conventions only arise in situations wherein everyone is particularly motivated to align their beliefs and expectations with one another and to coordinate their behavior and achieve unanimity of conformity.  The problem with Lewisian social conventions is that they must pop into and out of existence instantaneously, and they can neither evolve nor devolve over time. Nor do they allow for disagreement or dissent or pluralism.  This is the price that one must pay to capture the obligatory character of social conventions.  But, we know that social institutions evolve and devolve over time.  People disagree about what the law is and should be. And, despite this fluidity and pluralism, we still behave and speak as if there were a real obligation to follow the law and to comply with the social conventions of one’s social group.  

The first step to solving this problem is to adopt Margaret Gilbert’s brilliant insight that social conventions are social group constituting.  The fact that a population has a social convention constitutes that population as a social group for that reason, if for no other.  The second step is to recognize the role that practical authority plays.  This is the case, even for Lewisian social conventions, because of the risk dominance of the status quo position.  No one is going to conform to an alternate Lewisian social convention, unless she knows that her entire community is going to do likewise.  The worst possible outcome is for anyone to fail to coordinate. Practical authority solves this problem, because the authority makes it known how everyone in the social group will behave, allowing the social group to overcome the risk dominance of the status quo position and move between alternate Lewisian social conventions.  The third step is to recognize that all social conventions are step public social goods.  This is the case because the authority is a freeriding defector.  Now, we no longer require unanimity to generate normativity.  Once the step public social good of the social group itself has been generated, the social group rests upon an equilibrium point.  No one wants to defect, because the social group would collapse back to the status quo.  And, the authority has no incentive to conform, because she fares far better by continuing to defect.  Thus, there is a single rational thing to do in this situation, which is to continue to conform to the social convention.  Now that we no longer need unanimity to achieve normativity/obligation, we can have sub social groups within a larger social group with their own sub social conventions that wax and wane over time.  Thus, we can have evolution and disagreement. And, we have solved Lewis and Hart’s problems.  

The Obligatory Character of Social Conventions?

Both David Lewis, in Convention, and HLA Hart, in The Concept of Law, are trying to capture what they construe as an essential feature of social conventions, namely, their obligatory character.  Hart wishes to capture a real obligation to comply with the law, and Lewis wishes to capture a real obligation to comply with the social conventions of one’s social group, and, in particular, the linguistic social conventions of one’s social group.  In Convention, Lewis has a particular interest in what it means for a certain population to have a language, for a language to be thelanguage of thatpopulation.  In The Concept of Law, Hart speaks to a legal system being a union of primary and secondary social rules, rather than conventions, precisely because he is intent on capturing the obligatory character of the legal rules that comprise a legal system, but he is in fact speaking of a union of primary and secondary social conventions.  This point of fact does not escape Hart.  He is neither invoking a mysterious supra-individual normative / obligatory character, nor one that emanates or arises from the aggregation of the intentions / agency of the individual social group members.  He knows that he has only the preferences, expectations, and actions of the individual social group members with which to work.  Neither Lewis nor Hart are under any illusions about this point.  Both Lewis and Hart are fully aware that they have only an individual instrumental (means-ends) rationality from which to generate something akin to an obligation to comply with the social conventions of one’s social group.  As Margaret Gilbert would and does say, in “Social Convention Revisited,” both Lewis’ and Hart’s accounts of social conventions are individualistic rather than holistic.  

But, this is also why Margaret Gilbert would and does say that both Lewis’ and Hart’s accounts of social conventions fail to capture the essential feature of obligatoriness.  Can there be a real obligation to comply with the social conventions of one’s social group, particularly with respect to Lewis’ and Hart’s accounts thereof?  And, moreover, is there a real obligation to comply with the social conventions of one’s social group, regardless of whether Lewis’ and Hart’s accounts thereof are able to capture this allegedly essential feature?  The simple answer is no; there is no real obligation to comply with the social conventions of one’s social group, nor the law, for that matter.  There are no normative facts.  There is no fact of the matter about how people should reason; there is only how people do reason.  Or, at least, there is only how most quasi-rational individual agents do reason most of the time.

I don’t think asking whether there is a real obligation to comply with the social conventions of one’s social group, be they linguistic or legal or what have you, is a very interesting question.  I think the answer is obviously no, there is no such thing as a real social obligation. Attempting to manifest a social ontological entity that is a real obligation to comply with the social conventions of one’s social group is a metaphysically suspect exercise in futility that obscures the real nature of human and social psychology that is vastly more interesting and useful.  If we want to talk about how we should build our social institutions, we must first understand how we do build and maintain our social institutions.  

I’m far more interested in explaining why most people most of the time behave and speak as if there were an obligation to comply with the social conventions of one’s social group. An account of social conventions should capture whatever it is about social conventions that motivates most people most of the time to behave and speak as if there were an actual obligation to comply.  This something will be something closely akin to obligation, or, at least, normative character, or as close as Lewis and Hart can hope to get under the constraint of individual instrumental rationality.  An account of social conventions like Gilbert’s, one that enshrines a real obligation into its social ontology, obscures whatever it is about human and social psychology that compels most people most of the time to behave and speak as if there were this real obligation to comply.  Both Hart and Lewis think that they are able to capture something closely akin to obligation, but they pay a heavy price for it.  The price they pay for it is only being able to give an account of what Margaret Gilbert calls “Lewisian social conventions.”  

The Obligatory Character of Lewisian Social Conventions?

In “Social Convention Revisited,” Margaret Gilbert hews very closely to Hart’s understanding, in The Concept of Law,of what it means to have an obligation to comply with the social conventions of one’s social group. You have an obligation to comply with your social group’s social conventions, if the other members of your social group are in a position to demand such compliance and sanction you for non-compliance.  But, if we’re only working with an individual instrumental (means-ends) rationality, then it doesn’t seem that we can have a real obligation.  If I am an individual agent, subject only to an individual instrumental rationality, and if I am taking my means to my ends, meaning that I am engaged in a course of action that I believe will secure whatever it is that I most desire (neglecting discussion of diachronic rationality), then I am behaving rationally.  No one, not even a member of my social group, is in any position to fault me for my choice of action, or to demand that I do otherwise, not on rational grounds, not as a requirement of rationality.  

This state of affairs has motivated many a practical and moral and legal philosopher to posit the existence of other branches of practical rationality, including prudential and social and moral rationality, that may or may not supersede an individual instrumental (means-ends) rationality, with morality generally construed as the paramount branch of practical rationality.  That this enterprise is metaphysically suspect, i.e., the positing of normative facts, has not been lost on many a practical and moral and legal philosopher.  The moral contractarians, like David Gauthier, in Morals by Agreement, have taken great pains to construct elaborate moral theories out of aggregations of individual agents’ preferences, expectations, and actions.  Margaret Gilbert engages in a similar exercise of constructing a real social obligation, i.e., joint commitment, out of the public demonstrations of the individual commitments of social group members to all other members of their respective social groups.  I think these are exercises in futility that, while not positing a supra-individual, metaphysically suspect social ontological entity of obligation, not only fail to hit upon their intended target, but also obscure what is true about the real nature of human and social psychology.  That it is true that most quasi-rational individual agents behave and speak as if there were a real obligation to comply with the social conventions of one’s social group most of the time has also not been lost on many a practical and moral and legal philosopher.  I think this is one of the great philosophical problems of our time. 

Both Hart and Lewis seem to think that there can be something closely akin to a real obligation when there is a single rational thing to do in a situation that gives rise to a social convention, which is what everyone else in your social group is doing.  If there is a single rational thing to do, then you have something closely akin to an obligation, even in terms of an individual instrumental (means-ends) rationality, to do that one rational thing.  And, if there is a single rational thing to do in a situation, then everyone in your social group has something closely akin to an obligation to do that one rational thing.

In The Concept of Law, Hart argues that this is what distinguishes social conventions (or rules, as he would put it, because there is an obligation to comply), from mere patterns of behavior, descriptive norms, and lesser, quasi-social rules, such as rules of etiquette.  A social rule is a social rule, for Hart, because there is a real pressing need for everyone in the social group to conform thereto.  No unilateral deviation from the rule can be tolerated.  Everyone in the social group fares far worse if anyone deviates.  The only rational thing to do in the situation that gives rise to the social rule / Lewisian social convention is what everyone else in your social group is doing.  This is what places your social group members in a quasi-position of being justified in demanding your compliance and sanctioning you for non-compliance.  This is why Hart says that a social rule (Lewisian social convention) is a social rule, because of the social fact that every single member of the social group takes the internal point of view towards the social rule and accepts it.  But, then, social conventions, Lewisian social conventions, only arise in situations wherein everyone in a population is particularly motivated to align their beliefs and expectations with one another and to coordinate their behavior to achieve unanimity of conformity.  And, Hart accepts or concedes as much.  And, this is the beginning of the downfall of his theory of a legal system, even though I think it can be saved, and my goal is to save it. 

The text book example of a Lewisian social convention is driving on the right or left hand side of the road.  If everyone drives on the right hand side of the road, then no one wants anyone to drive on the left hand side of the road, and everyone fares very poorly (injury, property damage), if anyone drives on the left hand side of the road.  The only rational thing to do is to drive on the same side of the road as everyone else in one’s community.  This is what places all of the social group members in the quasi-position of being justified in demanding compliance and sanctioning for non-compliance. This is why there is something very closely akin to obligation in situations wherein Lewisian social conventions arise.  We can certainly appreciate why social group members would behave and speak as if there were a real obligation to comply with the Lewisian social convention of driving on the right hand side of the road.  

The Problem(s) with Lewisian Social Conventions

With respect to a situation that gives rise to a Lewisian social convention, the worst possible outcome is a failure, any failure to coordinate.  Lewis allows for merely negligible deviation.  No one wants anyone to deviate.  But, this entails some very unfortunate consequences with respect to Lewis’ and Hart’s accounts of social conventions.  Their accounts of social conventions are accounts of Lewisian social conventions.  And, Lewisian social conventions are severely constrained.  Lewisian social conventions must pop into and out of existence instantaneously.  Lewisian social conventions can neither evolve nor devolve over time.  Lewisian social conventions do not allow for disagreement or dissent or pluralism.  This is because the worst possible outcome is any failure to coordinate; any act of unilateral deviation is grossly unacceptable.  Everyone in the social group fares far, far worse, if even a single social group member unilaterally deviates.  This is why the status quo is risk dominant, and risk dominant to such a degree, that no one in the social group would risk conforming to an alternate Lewisian social convention, unless she were absolutely certain that everyone else in her social group was going to jump instantaneously and simultaneously to this alternate Lewisian social convention.  But, that seems like a very high bar.  

Think again about the Lewisian social convention of driving on either the left or right hand side of the road.  Imagine a population wherein everyone drives on the left hand side of the road.  It matters little why, but imagine that there was some pressing reason for the community to switch to driving on the right hand side of the road, perhaps to bring themselves into alignment with a neighboring community.  Regardless of how pressing this need might be, no one in the social group is going to switch to driving on the right hand side of the road, unless she was absolutely certain that everyone in the social group was going to make this change, instantaneously and simultaneously.  This intuitively makes sense.  The worst possible outcome would be for even a single driver to make the switch unilaterally.  This could have catastrophic results.  This is why the status quo is risk dominant.  This is why Lewisian social conventions cannot evolve and devolve over time. This is why they must pop into and out of existence.  This is the price that one must pay to capture the obligatory character of social conventions.  If obligatory character demands unanimity of conformity at every moment, then social conventions that possess an obligatory character can never evolve or devolve over time.  There can be no disagreement about what the social convention is.  

For the reasons I mention immediately above, some feel that Lewisian social conventions don’t actually exist in the real world.  They believe that they are a mere idealization.  This matters little to me.  My intention is to modify Lewis’ account of social conventions so as to accommodate a whole range of social conventions, of varying degrees of conventionality and varying degrees of normative (and perhaps obligatory) character. My account will be able to accommodate Lewisian social conventions, whether they actually exist in the real world or no.  I will be able to accommodate these Lewisian social conventions of the highest degree of conventionality and the highest degree of normative / obligatory character, even if they are only a useful idealization.  

Ronald Dworkin famously offers a withering critique, in Law’s Empire, of Hart’s account of a legal system for just this reason, that for Hart, as for Lewis, there can be no disagreement about what the Lewisian social convention or social rule is, nor the law for that matter.  Dworkin was actually quite brilliant to notice that Hart’s account of a legal system, because his legal system is a union of primary and secondary Lewisian social conventions, despite the fact that Hart refers to these Lewisian social conventions as social rules, doesn’t allow for theoretical disagreement about what the law is or should be.  If the normative / obligatory character of a legal rule (that is a Lewisian social convention) is premised upon the social fact that everyone in the social group whose legal system it is conforms to that legal rule, expects everyone to conform to that legal rule, and prefers to conform to that legal rule, as long as everyone else in the social group does likewise, then a legal rule for which this were not the case, would not be a legal rule (a Lewisian social convention) with a normative / obligatory character. Dworkin astutely points out that, per Hart’s theory of a legal system, any disagreement with respect to whether a particular legal rule is a legal rule of whichever legal system, means that it isn’t a legal rule of that legal system.  (I address elsewhere how Hart attempts to salvage his legal theory in the most disastrous way possible – i.e., by alienating the citizenry / general public from their own legal system, which only serves to lead him into a quandary.)

Ronald Dworkin was quite brilliant to notice this flaw in Hart’s legal theory, a flaw which Dworkin exploited to great effect, and which gave us what is perhaps one of the foundational dialectics in legal philosophy, the Hart-Dworkin debate.  But, much as I admire Gilbert and fully recognize her brilliant insight that social conventions are social group constituting, while I am highly critical of her paradigmatically wholly normative account of social conventions, I also admire Dworkin’s brilliant insight into the flawed nature of Hart’s legal theory, while I reject Dworkin’s natural law theory. I reside upon the Hart side of this debate, as an exclusive legal positivist, regardless of the flaws in Hart’s legal theory.  Hart argues that a legal system is comprised of the union between primary, duty imposing social rules (Lewisian social conventions) of obligation and secondary, power conferring social rules of authority.  Dworkin argues that legal systems are not merely comprised of rules, but also principles, moral principles, which are not rules and cannot be rules.  I agree with Hart.  Dworkin’s principles are rules, and they can be rules.  Or, as I would say, Dworkin’s principles are social conventions, and they can be social conventions.  

In his attempt to capture an obligatory character as an essential feature of social conventions, and, in particular, linguistic social conventions, Lewis finds himself in a similar quandary to Hart.  In Convention, Lewis is particularly interested in establishing what it means or what it is for a particular population to have or to speak a certain language.  And, not surprisingly, he settles upon the notion that for a certain population to possess or to speak a certain language is for that population to have a social convention of truthfulness in that language.  But, just as we have strong pre-theoretical intuitions that our social conventions wax and wane over time and we disagree about what the law is and should be, we also have strong pre-theoretical intuitions that our languages, not only evolve over time, but that we also make statements and employ linguistic terms for many other reasons than to indicate true facts about the world. Hart thinks that in order for it to be the case that a certain social group has a legal system with an obligatory character, everyone in that social group has to take the internal point of view towards that legal system and accept it.  Lewis thinks something very similar in Conventionwith respect to language.  Lewis thinks that in order for it to be the case that a certain social group has a language with an obligatory character (it is an actual obligation to employ linguistic terms to refer to certain objects in the world), everyone in that social group expects everyone else in that social group to conform to the social convention of truthfulness in that language; everyone in that social group prefers that everyone in that social group conforms to the social convention of truthfulness in that language, as long as everyone else in that social group does likewise; and, in fact, everyone in that social group actually does conform to the social convention of truthfulness in that language. But, this seems intuitively wrong. Lewis seems to think that people only engage in conversation, in order to align their beliefs about what is true about the world.  But, I argue, in my non-ideal philosophy of language, that people actually engage in conversation for the purpose of imposing their worldviews upon one another. Just as I think Hart’s theory of legal systems can be saved, I also think that Lewis’ account of linguistic social conventions can be saved.  We can have a much more realistic account of what it means for a population to have a certain language, and we can still allow for the meanings of linguistic terms to evolve over time, and we can still allow for the conventional meanings of linguistic terms to evolve within the course of a single conversation, because a conversation is no longer only about aligning one’s beliefs with another person, with respect to true facts about the world.  

Can We Modify Lewis’ Account of Social Conventions?

As Ronald Dworkin so brilliantly pointed out, we disagree about what the law is and should be.  Likewise, we have strong pre-theoretical intuitions that our social conventions evolve and devolve over time, waxing and waning, arising and collapsing, as quasi-rational individual agents pressure the evolution and devolution of their social groups’ social conventions.  And, despite this fluidity and pluralism, we still speak and behave (most of us; most of the time) as if there were a real obligation to follow the law and to comply with one’s social groups’ social conventions.  

So, both Lewis’ and Hart’s accounts of social conventions are terrifically and woefully inadequate. Both Lewis’ and Hart’s accounts of social conventions are merely accounts of Lewisian social conventions. Lewisian social conventions only capture one class of social conventions, social conventions with the highest degree of conventionality and the strongest normative (perhaps obligatory) character.  A “degree of conventionality” is Lewis’ term, but regardless of his circumspection, he still allows for merely negligible unilateral deviation.  He only allows for social conventions of the very highest degrees of conventionality to exist, i.e., Lewisian social conventions, because these are the only social conventions that may boast, according to Lewis, something closely akin to an essential feature of normative (and perhaps obligatory) character.  

Regardless of the inadequacies of Lewis’ account of social conventions, I still think there is so much that Lewis gets right about social conventions, including arbitrariness and common knowledge.  Arbitrariness, in Lewis’ sense, means that I want to conform as my social group conforms, and if my social group were to coordinate upon a different alternate Lewisian social convention, then, I would likewise prefer to conform upon that different alternate Lewisian social convention.  If I were to conform for the sake of an agreement, because morality compels me to do so, then this would not be conventional behavior.  (An agreement can serve as the basis of a social convention, but not for the sake of the agreement itself.)  But, this is also why, as Lewis makes clear in his account of social conventions, there must always be at least two alternate Lewisian social conventions in a situation that gives rise thereto, because, if there were only one, then conforming to that single option would not be conventional behavior.  

Contra Gilbert, I believe that individualistic accounts of social conventions are the only ones that avoid disqualification for being metaphysically suspect.  (I explain elsewhere why Gilbert’s attempt to concoct a social obligation, i.e., joint commitment, as its own branch of practical rationality, out of the individual commitments, and public demonstrations thereof, of the individual members of a social group, fails miserably.)  Lewis and Hart are both right to acknowledge that we only have the preferences, expectations, and actions of individual social group members with which to work.  (Gilbert claims that her account of social conventions requires neither conformity, i.e., an actual regularity in behavior, nor expectation of conformity, nor preference for conformity, but, nonetheless, results in a real social obligation to comply with the social conventions (or rules) of one’s social group, even when a verbal agreement is not the basis for the social convention. For Gilbert, the paramount essential feature of a social convention is its obligatory character.)  

But, they, Lewis and Hart, are also both right to notice that most people most of the time behave and speak as if there were a real obligation to comply with the law and one’s social conventions.  They err in thinking that the only way to accommodate both essential features (normativity/obligation and unanimity of conformity/a regularity in behavior) is for social conventions to only arise in situations in which there is but a single rational thing to do, which is what everyone else in one’s social group is doing. We want to capture normativity (and maybe even something closely akin to obligation) without demanding unanimity in conformity.  But, can we modify Lewis’s account of social conventions to capture evolution, disagreement, and pluralism, while also capturing something akin to normative (and perhaps obligatory) character?  I think we can fix Lewis’ account of social conventions, and, in so doing, fix Hart’s account of legal systems, as well as Lewis’ account of what it means for a population to have a language.  

How to Fix Lewis’ Account of Social Conventions

We can modify Lewis’ account of social conventions to accommodate all of the different classes of social conventions, of varying degrees of conventionality.  And, at one and the same time, we can continue to capture the quasi-obligatory character that Lewis captures, i.e., we can retain the essential feature that there is but a single rational thing to do for someone party to a social convention, which is continue to conform to the social convention, as long as everyone else party to the social convention continues to do likewise. Lewis takes himself to have captured as much of a quasi-obligatory character as he requires, if he is able to capture the fact that there is a single rational thing to do for someone party to a social convention, which is what everyone else in one’s social group is doing.  If there is a single rational thing to do, then Lewis takes it that you have something very closely akin to an obligation to do that one rational thing.  And, that’s as much as he requires, because this essential feature explains why most people most of the time behave and speak as if there were a real obligation to comply with the social conventions of one’s social group.  

How are we to go about modifying Lewis’ account of social conventions to allow for evolution and devolution, and disagreement and dissent, whilst there being a single rational thing to do?  The first step is to adopt Margaret Gilbert’s brilliant insight that social conventions are social group constituting.  The fact that a population has a social convention constitutes that population as a social group for that very reason, if no other.  The second step is to recognize the role that practical, and not merely epistemic, authority plays.  Practical authority makes it known how everyone in the social group will behave. Ultimately, I am going to say that all social coordination requires practical (and not merely epistemic) authority. The third step is to recognize that all social conventions (and social groups) are step public social goods. This is the case because the authority, the practical authority, is a freeriding defector who is not actually party to the social convention, but remains a member of the social group.  (So, if you can have a social group with two members, and I think you can, then you can have a social convention with one social group member party thereto, as counterintuitive as this might seem.)  Combining these 3 steps results in social group (and sub social group) constituting social conventions (institutions) that rest upon equilibrium points (meaning there is a single rational thing to do for one party to the convention, i.e., conform), but that can also evolve and devolve over time.  We have severed the essential feature of obligatoriness, or something closely akin thereto, from unanimity in conformity, allowing for negligible unilateral deviation.  Thus, we have normative (and perhaps obligatory) character without requiring unanimity in conformity at every moment in time.

Why is it so important to sever the essential feature of obligatoriness from the constraint of unanimity in conformity at every moment in time?  If Lewis’ and Hart’s accounts of social conventions demand unanimity in conformity at every moment in time, because there is a single rational thing to do, which is what everyone else in one’s social group is doing, so you have something closely akin to an obligation to do that one rational thing, as does everyone else in your social group, then the worst possible outcome is for anyone to deviate.  No unilateral deviation may be tolerated.  If this is what it means for a population to have a social convention, then, in order to do anything but maintain the status quo, the entire social group must always act as one, acting en masse, instantaneously and simultaneously. This means that no one is going to conform to an alternate Lewisian social convention, unless she is absolutely certain that her entire social group is going to act likewise, instantaneously and simultaneously.  This is why the status quo is overwhelmingly risk dominant, risk dominant to such a degree so as to make it impossible for the social group to do anything other than maintain the status quo, which is essentially negating the essential feature of arbitrariness.  And, as mentioned above, if there are no alternate Lewisian social conventions, to which the social group could jump, if they so choose, then continuing to conform to the status quo is not conventional behavior.  

Moreover, Lewisian social conventions do not comprise all social conventions.  Lewisian social conventions are a single class of social conventions, of the highest degree of conventionality and the strongest normative/obligatory character.  Both Lewis and Hart have backed themselves into a corner in which they have to pretend that the only social conventions that exist are Lewisian social conventions. They are both pretending that what it means for a population to have a social convention is for the population to have a Lewisian social convention.  But, this flies in the face of our pre-theoretical intuitions that our social conventions wax and wane over time, casting a larger and smaller normative shadow over our lives as they do so.  

This also means that we don’t have to make Hart’s arbitrary and, to many philosophers, nonsensical distinction between lesser and, therefore, quasi-social rules, like rules of etiquette, and greater and, therefore, real social rules, like driving on the right or left hand side of the road.  This point has long been employed to disparage Hart’s account of social and legal rules, because the distinction is so obviously desultory. If we sever obligatoriness from unanimity of conformity, then we can consider non Lewisian social conventions of lesser degrees of conventionality and weaker normative/obligatory characters. A rule of etiquette is a social rule, a real social rule, just as much as driving on one side of the road or the other is.  

What we will have to do is make a distinction between legal rules and mere social rules that are not also legal rules.  But, Hart has already done this for us.  This was his most brilliant insight, which he betrayed in his quest to establish the obligatory character of law.  A legal rule is a legal rule, because it is part of a legal system, the union of primary, duty imposing social rules of obligation and secondary, power conferring social rules of authority, for which such authority is being conferred upon public officials, and not mere citizens.  Social rules become legal rules once the social group has conferred authority upon public officials (qua public officials) to promulgate, modify, and adjudicate these social rules.  The boundaries of the power conferred are the social rules of the social group whose legal system it is.  If a public official exceeds her power, then she has not violated an obligation; she has merely failed to make law.  Hart undermined his legal theory by alienating the general public / citizenry from its own legal system in his dogged pursuit for social and legal obligation. 

Step One – Social Conventions are Social Group Constituting

I disagree with a great deal in Margaret Gilbert’s account of social conventions in “Social Convention Revisited,” BUT I think her insight that social conventions are social group constituting is absolutely brilliant.  The fact that a population has a social convention constitutes that population as a social group for that reason alone, if for no other.  This is really important for my account, as you will see below.  When a population generates a social convention, the population generates a social group.  I refer to Gilbert’s account of social conventions (really social rules) as a paradigmatic wholly normative account of social conventions.  For her, the obligatory character of social conventions is the paramount essential feature, and she requires neither conformity (a regularity in behavior), nor expectation of conformity, nor preference for conformity. In my opinion, Gilbert’s account of social conventions is really an account of social rules, but in the vein of moral rules, as not requiring actual compliance, expectation of compliance, or preference for compliance to justify their obligatory character.  Gilbert is highly critical of Lewisian social conventions, which she refers to as individualistic, rather than holistic, and she argues that there can be no real obligation to comply with Lewisian social conventions.  Gilbert is arguing for a branch of practical rationality that is social obligation, that is an obligation to comply with the social conventions of one’s social group, and that is something apart from individual instrumental (means-ends) rationality, prudential rationality, and, even, morality, although she acknowledges that morality remains the paramount branch of practical rationality, superseding all others, including social/conventional rationality.  

Why would Gilbert include an essential feature of social group constitution in her wholly normative account of social conventions (really social rules)?  This seems, at first glance, counterintuitive.  Why would it matter to her that a population manifest an actual social group when she claims not to require any manifestation of a regularity in behavior, nor any expectation of or preference for such a regularity in behavior in a population?  It is actually precisely for this reason, particularly in the instance when there is no verbal agreement to serve as the basis for a social convention. I argue elsewhere that Gilbert does in fact rely upon, and must rely upon, a regularity in behavior, as well as an expectation of and a preference for such a regularity, particularly in the instance in which there is no verbal agreement to serve as the basis for a population’s social convention.  

Gilbert relies upon her concept of “joint commitment.”  As part of a population’s joint commitment to a particular social convention, all commit all to comply with the social convention, jointly and severally.  But, in the absence of a verbal agreement to comply with a particular social convention, how would this joint commitment take place?  Gilbert indicates that she requires only that each member of a population behave as if she were a member of the social group, in the public sphere of that particular social group, or, out in the open, as it were.  Once this occurs, and the joint commitment has taken place, all social group members owe their compliance to the social conventions of the social group to the social group itself, not the other members of the social group.  Once the joint commitment takes place, the social convention has been created, as the property of the social group, and the social group has been created, as the joint and several owners of the social convention, and no individual social group member may rescind her commitment to comply with the social convention, unless the social convention, and, perhaps, the entire social group, is jointly and severally rescinded.  But, this seems odd.  First and foremost, the threshold for making such a commitment seems intolerably low.  I commit myself, and others (!), to a great deal by doing very little, by merely behaving as if I were a member of a particular social group in their public sphere. Moreover, I don’t know how I would know how to behave like a member of a particular social group in their public sphere, unless I knew of their regularities in behaviors, and I knew that they expected their social group members to conform to these regularities in behaviors, because they had a preference for these regularities in behavior. 

But, now it makes sense why Gilbert would include social group constitution as an essential feature of social conventions in her account thereof, along with social obligation, including all other social group members being in a position to demand compliance and sanction for non-compliance.  If a social convention is the property of those who jointly and severally created it, and who, therefore, jointly and severally own it, then, necessarily, those joint and several owners form a social group, for that reason, if for no other. Gilbert speaks in “Social Convention Revisited,” at length about how social groups talk about their conventions being theirs.  It is important for Gilbert’s wholly normative account of social conventions (social rules) that a population knows that they are a social group, and that they know what their conventions are and that those conventions are theirs, precisely because she does not require actual conformity, nor expectation of conformity, nor preference for conformity.  If you don’t owe your compliance to the other social group members, because Gilbert’s account of social obligation is holistic, and not individualistic, then you must owe your compliance to the social group as a whole, as a single social entity.  (This social entity does seem quite metaphysically suspect and mysterious at this point, though.)  This is why, for Gilbert, social/conventional rationality is a stand alone branch of practical rationality, superseded only by morality, but which itself supersedes both an individual instrumental and a prudential rationality.  This is why, for Gilbert, once the joint commitment has been concluded, individual social group members are no longer in a position to rescind their individual commitments, even if they have individual instrumental and prudential interests to the contrary.  However, Gilbert’s account of joint commitment fails, particularly in the absence of a verbal agreement, without regularities of behavior that are expected of social group members and for which social group members have a preference for that reason, at least for most social group members, most of the time. 

As I mention above, I find Gilbert’s account of social conventions and social obligation implausible and metaphysically suspect, despite her claims of avoiding any untoward social ontological entities.  You just don’t get to have your social ontological cake and eat it too.  If you don’t want to employ metaphysically suspect normative facts, such as a social obligation, not to mention a social group, that emanates mysteriously from the mysterious manifestation of a social convention in a population, then you have to do as Hart and Lewis do, and make use of individual agents’ expectations, beliefs, preferences, and actions.  But, you don’t get to make use of individual agents’ expectations, beliefs, preferences, and actions, while claiming that you are doing nothing of the kind.  Either social obligation is real or it isn’t.  (Hint – it isn’t.)  And, now we have the challenging task of explaining why most people most of the time speak and behave as if it were.  This is what Hart and Lewis did, and I think they made some brilliant moves, but I think we can do even better.  

Step Two – The Role that Practical Authority Plays

Social coordination requires practical authority.  I employ the esteemed legal philosopher Joseph Raz’s definition of authority, in The Authority of Law, as the power (or capacity or ability) to alter the protected reasons of someone else –a protected reason is a first order reason to do something and a second order reason to stop considering any alternate actions.  This is the case, that social coordination requires practical, and not merely epistemic, authority, even for Lewisian social conventions, because of the risk dominance of the status quo position.  No one is going to conform to an alternate Lewisian social convention, unless she knows that her entire community is going to do likewise.  The worst possible outcome is for anyone to fail to coordinate.  Practical authority solves this problem, because the authority makes it known how everyone in the social group will behave.  

An authority communicates the salience / optimality of a particular Lewisian social convention to the population.  Communication is required, but not necessarily verbal / linguistic communication, in order for it to be the case that the entire population knows how the entire population will behave.  A communicator always bears a cost for communicating, so being a communicating authority is risky, if one’s authority is not recognized.  But, it can be in an authority’s long term rational self interest to bear this cost.  Authority is always assumed.  Therefore, for an authority to arise in a population demands that there exists at least one risk tolerant, foresighted individual.  Coordination problems can go unresolved, if such an individual does not exist.  Additionally, once such authority has been assumed, it must also be conferred by the population.  Coordination problems can also go unresolved, if a population fails to confer such authority. 

This person who communicates to make salient a Lewisian social convention, be it optimal or no, be it already salient or no, is an authority.  How does this person become an authority?  This authority is conferred by the other social group members out of necessity, in order to solve the coordination problem and overcome the risk dominance of the baseline / status quo position (or the state of nature). In the case of Lewisian social conventions, the population is particularly motivated to coordinate, to align their beliefs and expectations and behaviors with one another.  Consider the motivation to coordinate upon the Lewisian social convention of driving on the right hand side of the road (or the left). If there is nothing particularly salient about either proper coordination equilibrium point / regularity in behavior, then it will be near impossible for the population to coordinate, instantaneously and simultaneously, en masse, upon either social convention.  Additionally, in order for the population to jump as a group to an alternate Lewisian social convention, the population must overcome the risk dominance of its current state.  Therefore, a secondary social convention will arise out of necessity, a social convention to confer authority upon some individual or another.  

But, if all social coordination requires the existence of an authority, then a secondary social convention that identifies an authority to make salient a primary social convention, in order to solve a coordination problem, seems to be no solution at all. For now we require an infinite regress of higher and higher order social conventions to identify ever more authorities. Fortunately, authority is always assumed, which isn’t to say that an authority always arises when needed. Unfortunately, coordination problems sometimes remain unsolved.  Recent empirical work supports the theory that authority is always assumed.  This assumption of authority stops the regress. No social coordination to identify an authority need take place.  Only a secondary social convention is required to solve the primary coordination problem.  The social group manifests its conferral of authority by adopting the primary social convention made salient by the assumed authority.  The normative / obligatory character of the secondary power conferring social convention piggy backs on the normative / obligatory character of the primary Lewisian social convention.  There is a single rational thing to do, conform to the primary social convention to which one’s population conforms, so there is a single rational thing to do, which is to confer authority as one’s population confers authority, since this is the only way to solve the coordination problem and to overcome the risk dominance of the baseline / status quo position.  

My view of authority is very much in line with that of the esteemed legal philosopher Joseph Raz, and I adopt his account of authority as my own.  Raz argues that to have authority over persons is to have the ability to change their protected reasons.  A protected reason is a first order reason to perform some action and a second order reason to discard (not consider) any first order reasons to act otherwise (to not so act).  The first order reason to perform some action (to conform to the Lewisian social convention made salient by the authority) is the communication of the salience of this Lewisian social convention / regularity in behavior.  This is also a second order reason to not consider any other first order reasons to act otherwise, i.e., the salience of the one Lewisian social convention is a second order reason to not consider any of the alternate Lewisian social conventions.  

The individual who assumes authority is going to be a risk tolerant person with foresight. Regardless of whether the authority is making salient a new Lewisian social convention, in order to solve a new coordination problem, or whether the authority is making salient an alternate Lewisian social convention, for the social group to jump to en masse, the authority pays a cost for communicating salience.  In order for it to be rational for the authority to bear the burden of communicating salience, the authority / communicator recognizes that it will be in her long term interest to do so.  What does an authority have to gain in the long term for bearing the cost of communicating to her social group, allowing the social group to solve a coordination problem?  

First and foremost, if the alternate Lewisian social convention happens to be optimal (preferred by all social group members), then it will certainly be in the rational self interest of the authority to try and move her social group to this more optimal state.  But, it can be in an authority’s long term rational self interest to try and move her social group to any Lewisian social convention.  The reason being is that any Lewisian social convention is social group constituting.  (As I mention above, I strongly concur with Margaret Gilbert’s astute assessment that social conventions are social group constituting.)  A population with a Lewisian social convention is a social group. And, being in a position of authority within a social group confers benefits, including the benefit of being able to move one’s social group to alternate Lewisian social conventions.  

Step Three – Social Conventions (and Social Groups) are Step Public Social Goods

An authority is a foresighted, risk tolerant individual.  Authorities do not always arise in populations, even when we need them. Sometimes coordination problems go unsolved.

When an authority does arise, it is because someone was able to see that it was in her long term interest to communicate an instruction to a population, at a cost, in order to reap the benefits of the position of being an authority in a social group –someone who is a freeriding defector and not actually a party to the social convention, while still being a member of the social group, leaving her open to moving the social group to alternate social conventions.  An authority in a social group also benefits by not having to contribute to the generation nor maintenance of a social convention / institution.  This is why social conventions are step public social goods.  

I employ Hugh Ward’s brilliant account of step public social goods in, “Three Men in a Boat, Two must Row: An Analysis of a Three-Person Chicken Pregame.”  The social group rests upon an equilibrium point immediately upon generating the social group via the social group constituting social convention.  There is a single rational thing for the parties to the social convention to do, which is to continue to conform.  The authority has no incentive to do anything other than continue to defect.  Thus, we can have normative (and perhaps obligatory) character sans unanimity.  It is at least rational, if not obligatory, to affirm your social identity by continuing to confer power upon the authorities in your social group.  

I argue that Lewisian social conventions are step public social goods that are constitutive of social groups.  Step public social goods are goods that are generated when some minimum threshold level of a social group contributes to or participates in the creation of the good.  This is why step public social goods are social conventions, because everyone who contributes has an incentive to do so, as long as everyone else does so.  A textbook example of a step public social good is a group hunt for a large animal.  Taking down a large animal requires that a minimum number of social group members participate in the group hunt.  If the group hunt is successful, everyone in the social group will benefit, even those who had not participated in the hunt.  And, everyone prefers that a group hunt be successful, as they benefit far more than if each social group member hunts individually for small game. The group hunt will not be successful if this minimum threshold level of participation / contribution is not met. If the group hunt is not successful, then those who attempted to participate in the group hunt fare far worse than if they had simply hunted alone for small game.  Once this threshold level of contribution has been achieved, the population of contributors immediately finds itself at an equilibrium point, at which point the only rational thing for contributing members of the social group to do is to continue conforming to the social convention. Those social group members who did not participate in the hunt, but who benefit from a successful group hunt, have no incentive to contribute, because they fare far better by continuing to free ride, and those who participate in the hunt have no incentive to free ride, because if they stop participating, the group hunt will fail.  

Think of the generation of a step public social good, which is the social group constituting Lewisian social convention, as a successful group hunt.  How is it that the social group resides at an equilibrium point immediately upon generating the step public social good / Lewisian social convention?  This is because the authority does not herself conform to the social convention / contribute to the generation of the step public social good.  She is a free riding defector.  She benefits from the generation of the step public social good (which is the constitution of the social group), and she benefits from not having to contribute to the generation of the step public social good.  She does not herself conform to the Lewisian social convention (Lewis allows for negligible unilateral deviation).  

Because the authority is a free riding defector, the social group rests at an equilibrium point, immediately upon having generated the step public social good, which is the generation of the social group.  The pool of contributors who conform to the Lewisian social convention have no incentive to do anything other than continue to conform; otherwise, the population would collapse back to the status quo, as the case may be.  The freeriding defector / authority has no incentive to contribute, because she fares far better by continuing to free ride.

Once the step public social good / Lewisian social convention has been generated, this constitutes the population as a social group, and the free riding defector / authority is in the position of benefiting from, but not having to contribute to, the step public social good that is the social group itself.  The benefit that the authority always accrues is being in the position of being able to make salient alternate Lewisian social conventions, i.e., of being an authority, of having authority over the other social group members.  There is always some degree of risk involved, because communicating is costly, and if the population does not confer the assumed authority, then no social group constituting step public social good / Lewisian social convention is generated. The assumed authority requires some degree of foresight, because a cost is borne in the present for a payoff to be reaped only at the point at which the social group constituting Lewisian social convention is generated.  To be clear, because the authority is a free riding defector, the authority is a member of the social group, but not a party to the Lewisian social convention.  The parties to the Lewisian social convention have an obligation to conform to the social convention, because there is a single rational thing to do, which is to conform, because the social group sits at an equilibrium point.  If any one of the parties to the Lewisian social convention were to unilaterally deviate, the step public social good, which is the social group, would collapse back to the status quo.  

A Lewisian social convention is always a step public social good.  There must exist at least one member of the social group who is an authority / free riding defector.  As long as there is a free riding defector, then there is a member of the social group who has the power to communicate the salience of alternate Lewisian social conventions.  This leaves open the possibility of the social group jumping en masse to alternate Lewisian social conventions.  So, Lewis’s essential feature of arbitrariness is retained.  Likewise, as long as there is an authority member of the social group, we retain the normative / obligatory character of the social convention, because, for the parties to the Lewisian social convention, there is a single rational thing to do, which is to continue conforming to the Lewisian social convention.  And, since the authority is able to communicate authoritative instructions to the social group, we also retain the essential feature of common knowledge.  

Given that this is the case, and given that Lewis does reformulate his definition of social conventions to allow for negligible unilateral deviation, I am able to capture the normative / obligatory character of Lewisian social conventions sans unanimity. I no longer require that it be the case that there is a single rational thing to do for every single social group member, which is what everyone else in one’s population is doing, en masse; now, I am able to achieve a normative / obligatory character of Lewisian social conventions, even in the case where there is a single (or potentially more) free riding defector(s).  This is so, because the equilibrium point, which is the step public social good that constitutes the social group as a social group, makes it the case that there is a single rational thing to do for the minimum threshold level of cooperators who actually generate the step public social good.  At one and the same time, there is a single rational thing to do for those party to the Lewisian social convention, and it can also be the case that there are social group members who deviate from the Lewisian social convention (the free riding defectors / the authority(ies)).  My normative / obligatory character of Lewisian social conventions no longer requires complete unanimity.  And, as discussed above, all Lewisian social conventions demand that there exists at least a single free riding defector / a single authority, not only to move between alternate Lewisian social conventions, but also to coordinate at all, to escape the state of nature.  This has the interesting result that the authority who makes salient the one or the other Lewisian social convention, is not herself a party to the social convention, nor is she under any obligation to conform to the social convention, in the sense of an individual instrumental rationality.  

So, authority is assumed, then conferred out of necessity to solve coordination problems.  Social coordination takes place, not in deference to authority for its own sake, but because the authority has made salient one Lewisian social convention or another.  Authority is necessary for a social group to move between alternate Lewisian social conventions.  And, the authority is the free riding defector that makes the step public social good that is generated by a minimum threshold level of cooperators stable, because the step public social good only rests upon an equilibrium point, because of the existence of the free riding defector / authority.  It is because the step public social good rests upon an equilibrium point that the Lewisian social convention possesses its normative / obligatory character, because there is a single rational thing to do – for the pool of cooperators / parties to the Lewisian social convention.  

Conclusion – We Fixed Lewis and Hart’s Problems with Normativity, Evolution, and Disagreement

Now that we have an account of social conventions that recognizes that social conventions are social group constituting step public social goods (wherein the social group is itself a social good generated, even if there are others) including freeriding defector-authorities with the power to move the social group amongst alternate social conventions (providing arbitrariness and common knowledge).  Now, we can have normative (and perhaps a quasi-obligatory) character sans unanimity.  There is a single rational thing to do, which is to continue to conform to the social convention, if you are a party thereto.  If we have normative (and perhaps a quasi-obligatory) character sans unanimity, we can have social groups and sub social groups evolving and devolving over time (growing and subsiding), along with disagreement and pluralism.

Lewisian social conventions are social group constituting step public social goods that are only possible via the conferral of authority.  The authority in a population makes it possible for the social group to overcome the risk dominance of either the state of nature or an existing Lewisian social convention.  Because the authority is a free riding defector who is not herself subject to the Lewisian social convention, but benefits from it, the social group constituted by the Lewisian social convention rests at an equilibrium point. Therefore, there is a single rational thing for parties to the Lewisian social convention to do, which is to continue conforming to the Lewisian social convention; they have an obligation to continue conforming.  Because the authority reaps the benefit of being a free riding defector, which is the authority to make salient alternate Lewisian social conventions, Lewis’s essential feature of arbitrariness is retained, even in the case of the social group jumping en masse from a more optimal Lewisian social convention to a less optimal Lewisian social convention (social devolution), which would not otherwise be rational.  

Future work will include revising Lewis’s definition of social conventions to include social conventions other than Lewisian social conventions, social conventions of varying degrees of conventionality and greater and lesser normative / obligatory character.  This is now possible, due to the conferral of authority, because we no longer require unanimity to obtain a normative / obligatory character of social conventions. Because we now think of social conventions as step public social goods, requiring a minimum threshold level of contribution to generate the good, which constitutes the population whose social convention it is as a social group, we can think of step public social goods that require varying threshold levels of participation, constituting sub social groups. This opens the door for an account of how social conventions evolve and devolve over time, sometimes slowly, sometimes precipitously, and, yet, those party to the conventions are still under an obligation to conform to them.  

Work Cited

Gauthier, David (1986). Morals By Agreement. Oxford: Oxford University Press.

Gauthier, David (1997) Resolute Choice and Rational Deliberation: A Critique and a Defense. Noûs 31(1):1-25.

Gilbert, Margaret, 1989. On Social Facts, New York: Routledge.

Gilbert, Margaret, 2008. “Social Convention Revisited,” Topoi, 27: 5–16.

Hart, H.L.A., 2012, The Concept of Law, 3rd edition (first edition 1961), Oxford: Clarendon Press.

Lewis, David, 1969. Convention, Cambridge: Harvard University Press.

Ranehill, Eva, Schneider, Frédéric, & Weber, Roberto, “The unrealized value of centralization for coordination” (unpublished manuscript dated January 1, 2017 received from authors).

Raz, Joseph, 1979. The Authority of Law, Oxford: Clarendon Press.

Ward, Hugh, 1990. “Three Men in a Boat, Two must Row: An Analysis of a Three-Person Chicken Pregame,” The Journal of Conflict Resolution, 34(3): 371-400. 

If you wish to support me, as I seek justice, it would mean the world to me. Here are my PayPalMe and GoFundMe links. 

PayPalMe:  https://www.paypal.me/SarahBraasch

GoFundMe:  https://www.gofundme.com/sarah-braasch-legal-fund

You can follow me on twitter here:  https://twitter.com/sarahbraasch1?lang=en

Please subscribe to my YouTube Channel here: https://www.youtube.com/channel/UCz4xV2R6mTVJhAu9OQzwp5g