IV. Commitments and rules *
The distinction between the "natural" conditions of social interaction that are beyond the reach of human intervention and the "artificial" ones that are created within social interaction itself is a classical one (cf. Hume 1948). It was not only of great value for the "moral science" of the British Moralists (cf. Raphael 1969, Mackie 1980). It can still be fruitfully applied to some fundamental issues of modern utilitarian (in the descriptive or sociological sense of the term) approaches to social interaction and the social institutions emerging from it.
This becomes particularly clear if one concentrates on game theoretic discussions of social institutions within modern utilitarianism. The real world counterpart of a cooperative game -- i. e. a game permitting enforceable agreements (cf. Harsanyi and Selten 1988, sec. 1.2) -- is a set of institutions, or, more precisely, it is an artificial set up that is created by social institutions. According to the classical utilitarian research program the existence and performance of these institutions have to be explained in ("natural") terms of purely individual goal-oriented action. From a traditional game theoretic point of view this amounts to the same thing as showing how a cooperative game can be created within a basically uncooperative one. (I sidestep here the issue of transforming a cooperative game into an equivalent uncooperative bargaining game via explicit formulation of all strategic possibilities.) Creation of anything comes about at a cost only and thus the question arises, why somebody should be willing to incur the costs of organizing social games -- in particular those of the cooperative variety. This question amounts to what critics of the individualistic and empiristic approach to social institutions have called the Hobbesian problem of social order. Therefore, if contemporary scholars try to solve the Hobbesian problem of social order with game theoretic means the proposed solution will encounter, at least in part, the same problems that it is meant to solve.
Though modem game theoretic analysis added many valuable insights to the discussion of social institutions the claim that it actually solved the problem of how social institutions can emerge and can be maintained and, thus, the Hobbesian problem of social order, should be treated with caution. It will be argued subsequently that a defense of the claim will require certain modifications of classical consequentialistic notions of expedient rational choice on which utilitarianism and game theory are based otherwise.
Surprisingly even today some people still seem to misunderstand the meaning of the modern notion of utility. The basic misunderstandings seem to stem from deeply rooted modes of thought that take utility as a reason for preferring certain states of affairs. Intuitively people tend to think of utility as something of emotional quality. Of course, the classical notion of utility did have this connotation. It was so to say something in its own right. Contrary to this, the modern notion of utility is strictly derivative. It is derived from preference. In particular, utility is not one of the reasons for preferring or ordering states of affairs. Utility as such is no motive guiding action. Consequently, utility is representing preferences rather than explaining them.
This in turn raises the problem of how to specify explicitly those motives that give rise to the preferences themselves and thus indirectly determine the shape of the utility representation. The classical Hobbesian assumption of selfishness evidently provides such a specification. Human beings have a natural inclination -- and according to Hobbes are even under a natural obligation -- to pursue their narrowly defined self-interest and nothing else.
If we do not dub any motive "selfish" simply because it is the actor's own motive then the assumption of universal selfishness is a bold conjecture. However, empirically this assumption about human nature seems to be far from self-evident. All people act selfish sometimes. Some people may act selfish all the time. But, to assume that all people act selfish all the time, seems to be patently absurd (at least if the term selfishness can discriminate between different forms of human motivation at all; cf. for a recent study of the theme, Sen 1987).
Of course, one could turn to learning theory and social psychology for a more realistic specification of human motives. However, contrary to what most utilitarians seem to assume, it is not completely clear how far learning theory can be integrated into the utilitarian research program without violating one of the basic assumptions of the latter (cf. on learning theory and its relationship to opportunistically rational choice, Vanberg 1989, who independently of the subsequent analysis reaches roughly the same conclusions though from a somewhat different point of view).
Learning leads to a modification of behavioral dispositions. This either may be due to a modification of what individuals want to do or to a modification of what individuals can do. The first kind of learning is compatible with the utilitarian research program. It simply provides a specification of the utility function in that it explains the preferences the function represents. The second kind of learning leads to a modification of the "behavioral technology" of rational individuals. This may mean two things. On the one hand, rational individuals can do certain things they could not do before, like learning to ride a bicycle. On the other hand individuals may get locked in with certain habits, traits of character or behavioral dispositions that seem to restrict their strategy spaces.
For a thoroughgoing utilitarianism, and in particular for its game theoretic variant, the last type of "learning" poses severe problems. Above all the assumption that individuals due to "learning processes" may develop certain kinds of inner commitments is of interest here. Such commitments as a matter of fact will not make certain modes of action strictly unviable. Nevertheless, it is mistaken to model these commitments simply in terms of a reevaluation of actions and thus in terms of the first mode of learning. Contrary to this there seems to be involved a deviation from the basic utilitarian and game theoretic assumption that actions are evaluated and chosen according to the consequences they causally effect.
Adhering to a broadly utilitarian approach myself and accepting that in general explanations of human action should be given in consequentialist or teleological terms of individually rational choice I can imagine that the latter point may alarm utilitarians and arouse their resistance. Therefore, it may be helpful for illustrative purposes to put this crucial point into game theoretic perspective. This will be done in a twofold way. First, some implicit assumptions of certain utilitarian, in particular game theoretic ways of modeling social interaction will be made more explicit (2.2). Then, second, it will be shown that certain applications of game theory in modeling the emergence and maintenance of social institutions either will not succeed completely or will have to violate some basic assumption (2.3).
The utilitarian model of the rational chooser comes in two basic variants. In one variant the choosing individual is modeled from a purely external point of view whereas the other variant takes an internal point of view and looks at rational choices with the eyes of the choosing individual and his perception of the situation (cf. for the somewhat similar distinction between substantive and procedural rationality, Simon 1985).
In theories that take the external point of view, behavioral laws connect objective features of a situation with certain "objectively" rational responses to explain the latter. Such explanations typically occur in modem evolutionary biology and in some fundamental models of the working of social institutions like those that look at the market process as a selection process. They assume that individuals behave "as if" they were rational choosers. But, it is a telling fact that these models justify the "as if" assumption by the further one that an objective selection criterion exists and that according to this criterion relatively less successful individuals are eliminated.
In biology this "objective objective function" is provided by relative shares in a gene pool whereas for instance in Armen Alchian's (cf. 1950) model of the market as a selection process profit -- objectivized by double entry book keeping -- serves the same function in the determination of relative success or failure and thus brings about a race leaving behind the less efficient entrepreneurs. In both cases it is a general property of the interaction system that provides the legitimation of the utility maximization assumption. The process of elimination according to an objective selection criterion is crucial and not deliberate maximization on the side of the participants of the race. The argument works precisely because it is independent of individual preferences. Substantive or objective rationality is enforced by objective features of the process or the logic of the situation so to say.
Selection processes presuppose that there is some entity that is fixed such that selection can operate on it. The assumption of a fixed behavioral program can easily be incorporated into a theory that approaches human behavior from an external point of view. However, it cannot easily be adapted to the second variant of utilitarian theories which take an internal point of view to the decisions of the rational chooser.
The non-utilitarian social scientist who starts from fixed individual values that have been internalized once and afterwards govern behavior is closer to selection arguments than the classical utilitarian. If we compare the classical utilitarian model of rational behavior, for instance, with such models as so-called homo sociologicus then its most characteristic element is that the rational actor is not preprogrammed by internalized values, norms etc. From an internal point of view, the notion of a fixed behavioral program contradicts the assumption of opportunity taking individual behavior.
This can easily be illustrated within the context of game theory. In classical game theory we look at the choice situation with the eyes of the chooser. We refer to his perception of the situation and model choices as conscious decisions that could be otherwise. These decisions may be perfectly determined by the laws of nature. Still, it is assumed that from the point of view of the chooser they could be otherwise. The chooser does not predict how he will choose. He makes his choice and this in turn itself is theoretically explained by referring to the model of the situation that the rational chooser himself is supposed to entertain.
Further, within the context of an uncooperative game the rational actor of classical game theory cannot exclude any choices from his future choice sets. He cannot commit himself in advance. He has "full control" over alternatives of choice at any decision point, but he cannot control the set of choices open to him. Real world individuals may be different -- and I will argue below that they indeed are -- but, according to the assumptions of our game theoretic models the rational actor cannot deliberately give up full control. (For instance game theoretic discussions of credible threats would not make much sense if they could; cf. 3. 2. below.)
It is exactly here that Robert Axelrod's (cf. 1984) in other respects ingenious idea of conducting computer tournaments as models of human interaction may be quite misleading. A computer program may be as complex and as flexible as we want it to be but it can be as inflexible at the same time. From a fundamental point of view this amounts to smuggling in certain elements of commitment power into an analysis that seems to be framed entirely in categories of uncooperative game theory. In any case these notions are incompatible with the key concept of the consequentialist or teleological utilitarian model of human action which contains the assumption of opportunity taking behavior and the faculty to adapt to considerations of expediency (cf. Lindenberg 1983; Opp 1986; Meckling 1976).
The assumption of ("unalienable") full control may not be left out of sight in any game theoretic analysis of social institutions as long as an internal point of view is adopted. Of course, one could avoid the assumption by restricting analysis to an external point of view. Then it must be shown by arguments referring to the logic of the situation (like those of Alchian) why and how certain rational forms of conduct should be expected to occur -- or the rational choice model would not be applicable at all. One cannot have it both ways, on the one hand, eliminate the assumption of intentional rationality and, on the other hand, base the prediction that the outcome of social interaction will conform with rational opportunity taking behavior on intentional rationality.
I think that confronted with the latter alternative most utilitarian theorists and in particular those who are interested in explaining how institutions are created and sustained -- and not only what is going on within existing institutions like the market -- will opt for intentional rationality as underlying premise. Therefore, in the rest of the paper I will only discuss those variants of utilitarianism that stick to the assumptions of intentional rationality and full control. These assumptions raise strong obstacles to an analysis of social institutions. Quite ironically the fundamental character of these obstacles will become obvious if we take a closer look at some of the great achievements of applying game theoretic tools to the analysis of social institutions. (For convenience I will concentrate on analyses of the emergence of institutions solving the well-known prisoner's dilemma as for instance; in Taylor 1976; Schotter 1981; Opp 1983; Raub 1986; Raub and Voss 1986; Schüßler 1985; Voss 1985; Kliemt 1986 a; Kliemt and Schauenberg 1984)
Conditions favorable for overcoming the prisoner's dilemma can come about either naturally or artificially (i.e. institutionally). The first would typically be the case if behavior of a small group of individuals is scrutinized under the assumption that the participants of this small numbers' game know that they will interact for some time. The setting in which the game takes place, then, can naturally provide a functional equivalent to commitment power and artificial facilities of coalition formation.
In the second case, commitment power is bestowed on individuals artificially by special social institutions like those of the legal order. The individuals can use their "powers" to create new institutions.
The second variant of institution creation relies on preexisting institutions and thus is a secondary one whereas in the first variant institutions are based on a primary process that does not itself rest on preexisting institutions. For the problem presently under scrutiny the primary process is of primary interest and therefore I will concentrate on this process subsequently.
If we imagine that there are only two Individuals, Robinson Crusoe and Friday, then what has been aptly called "the shadow of the future" (cf. Axelrod 1984, chap. 9) may enter the picture. Iteration of the basic or normal game modifies the incentive structure of the process. Moves of one player in earlier stages of the game may trigger moves of the other player at later stages. In his strategy choice every player has to take into account that his own present behavior may have causal effects on the behavior of his opponent and thus on his own future payoffs. From this it follows that -given an appropriate corditional strategy of the opponent -- each player can deliberately influence his future payoffs. As is well known supergame equilibria may evolve which lead to a sequence of so called "cooperative" moves in every normal game, Thus, even within a Hobbesian utilitarian approach the emergence of mutually beneficial behavior can be explained regardless of the fact that the underlying sequence of normal games is uncooperative and still contains the deviant or defection strategy that is dominant in every normal game taken separately. If we call the emerging regularity of behavior an "institution", which seems to be appropriate at least in a rudimentary sense of that notion, then in purely utilitarian terms a first and primary explanation of the emergence of institutions has been given.
Among the flock of game theorists and game theoretically minded social scientists these considerations are common or folk knowledge by now. Still, interpreting the way the result was reached in terms of natural functional equivalents of commitment power is not a folkway among social scientists. But it seems to be the appropriate way if we compare the analysis to that of a cooperative game proper (cf. Friedman 1986, 103 f.).
In a cooperative game commitment power comes about in two forms. On the one hand, a player can restrict his strategy space effectively. He can virtually eliminate certain "future" moves or strategic possibilities from the game (cf. Friedman 1986, 11 f.). He cannot pursue certain ways of action anymore regardless of the costs he might be willing to incur. This is an idealized version of commitment that often will be realized in social reality only in the limit, or under quite extreme circumstances. On the other hand, a player can modify his own future payoffs. Though the relevant modes of action are not effectively eliminated from the strategy space, their expected payoffs are (more or less) effectively altered. The player still can act in certain ways but he may be expected to forbear because it would be repelling for him to act otherwise.
The example of Robinson Crusoe and Friday contains functional equivalents to both ways of commitment. On the one hand, Robinson Crusoe and Friday are confined to their island. Therefore a fundamental restriction on substituting their interaction partner is effective. Each is committed to his partner in the first sense of that term, On the other hand, and related to the first, the interaction is a recurrent one. From this the possibility arises to influence later decisions of each other by former ones and thus to modify expected future payoffs deliberately by present decisions. This provides a functional equivalent to commitment power in the second sense.
Both forms of commitment are brought about naturally and not artificially by institutions. Under specific circumstances, as those outlined before, certain elements of cooperative games arise directly from natural circumstances or from "nature" so to say. They are provided without institutions. This indicates how in principle the realm of the artificial or of social institutions can be reduced to the sphere of natural conditions in general and in particular to "human nature" operating under natural conditions of interaction. However, it should be noted that the reduction holds good only insofar as the natural functional equivalents to commitment institutions prevail.
It seems that neglecting this qualification has led astray many utilitarian theorists of social institutions or at least that it has led them to incautious generalizations. It should be observed carefully again that being confined to an island with merely two inhabitants amounts to the same thing as being perfectly committed to a dyadic recurrent interaction. There is no exit option (cf. for a recent study of iterated PD with an exit option Orbell et al. 1984) and there are no opportunity costs in terms of interacting with another individual because there are no other individuals. This is a very strong premise which will be rarely fulfilled in reality. Actually, this condition is so strong that real world social institutions hardly if ever enforce equivalent commitments artificially. (Slavery is abolished and divorce introduced in most countries; cf. for "natural" mechanisms added to the artificial institution of forming long-term contracts Alchian 1984; Williamson 1983)
Recurrent interaction in the dyade without an exit option may indeed form an approximation of certain real world interactions like that between adjacent nations. These nations are stuck with their territory and thus with their neighbours. They know that they will be around with the same neighbours or neighbours for a while. Some immutable relationships of a biological kind like those of kinship may come to one's mind here too. However, these specific counterexamples notwithstanding, social interaction among humans typically deviates from the conditions faced in dyadic recurrent interaction of the Robinson Crusoe / Friday type. First, many interactions cannot be fractured down into dyadic interaction but instead of this are genuinely simultaneous interactions between more than two individuals. Second, and more important, almost all human interaction under modern social conditions takes place within a setting that involves many individuals. These may either be potential interaction partners who can enter an interaction, or "bystanders" who can influence the interaction from outside, or individuals who can adopt both roles.
The first objection may be answered, at least partly, in quite an easy way. Simultaneous small scale interaction will still allow for "trigger strategies" in the wide sense of that term. If any single individual should deviate from a cooperative mode of behavior then this may be plausibly regarded as "triggering" the reaction of other individuals. From this it follows that the behavior of a single individual can have causal effects on the future choices of other participants of the interaction, and not only on the public outcomes.
In a small group context, information about whether one player has deviated (not necessarily who) will be easily or (almost) costlessly available most of the time. Under these circumstances individuals may be expected to behave strategically (in Buchanan's 1965 sense of that term). It makes good sense to assume that they make their own reactions contingent on some threshold of the precise number of cooperators or non-cooperators (which, of course will cover trigger strategies in the narrow sense as a limiting case). They will react on alternative forms of behavior of a single individual under some well-specified circumstances (at the threshold) with different alternatives from their own choice set.
If such conditions prevail explanations of the primary evolution of social institutions that are based on individually rational behavior are viable. They show within the confines of an individualistic and consequentialistic approach how institutions can emerge and can be maintained in a setting that does not yet contain institutions (either in time or on a higher foundational level). However, at least to my reading of the relevant literature, these explanations of the primary evolution of social institutions do not only support the utilitarian approach but also tend to undermine it.
The explanations provided are strictly confined to small number settings. Small groups and large groups differ in a fundamental way as far as social institutions are concerned. In particular, Michael Taylor's (1976) analysis shows that the original emergence and selfsupporting maintenance of social institutions without external interference should essentially be regarded as a small group phenomenon. Because human individuals as a matter of fact live together in large groups we must then still explain how those "large scale institutions" that render viable this form of life could emerge and can be maintained.
The fact that small scale interaction is embedded in a larger context of social interaction may not be neglected. It is of crucial importance even if the type of interaction under consideration can be fractured down into dyadic interactions. This does not only give rise to the exit and entry option and the possibility of external reactions influencing the internal structure of the interaction. It also raises the question of how the insulation of small scale interactions within a larger interaction can be accomplished. How can barriers to entry and to exit be erected? Without these barriers the 11 natural" or non-cooperative functional equivalents to artificial commitment powers in cooperative games could not exist. There would be no insulation of a small group from the larger context and thus there would be no long-term expectations to interact with the very same individuals repeatedly in a strategic context where individual actions can trigger responses of others. In light of the fact that interaction partners can be substituted by other ones that could offer equivalent terms, the shadow of the future vanishes.
In fact we try to organize markets in a way that allows for the substitution of transaction partners. However, in market settings the institution of contract that is enforced by the legal staff is already presupposed. The legal staff is locked in a small number game external to the ongoing game between the market participants. The larger game can be created as a side effect of rational behavior in the smaller one. In our models of the market, members of the legal staff are in another game so to say. Still, on a more fundamental level of analysis all individuals are in the same more comprehensive game.
We must analyze the more comprehensive game, if we want to understand the game of life and the role of institutions within that game. We must explain in utilitarian terms of individually rational choice how stable "subgames" (not in the common technical sense) of a larger interaction can form regardless of the omnipresence of potential offers and counter offers to form new coalitions. The game theoretic analysis, based on the ends that are pursued by individuals in the more comprehensive game, has to show how small stable coalitions of rule enforcers can emerge and persist. Only then the "natural functional equivalents" to commitment institutions will prevail. Without this, analyses of supergames which show how cooperation can be self-enforcing in problematic social situations will have no bearing on real world problems of large interaction systems. For, as has been argued before, these solutions will be plausible only in small group interaction.
As long as we stick to the assumption of full control, it is not an easy question to answer how, without a spatial separation of groups, relevant barriers to exit or entry could arise. We may be tempted to introduce notions like group loyalty or values shared by the members of the groups etc.. on an ultimate explanatory level. But, then we might be well out of our utilitarian ways.
The utilitarian who wants to explain why rational actors may be committed to interacting with a certain group of individuals will point out relevant costs that are equivalent to the costs arising from a spatial separation of small groups. Indeed this would be quite close to the well-known analogy between transaction costs and costs of transportation. Along the same line of argument the utilitarian could also dwell on the idea of sunk costs that are recoverable only in an interaction with specific individuals. He could finally refer to a kind of endogenous change in taste that regularly arises from interacting "successfully" with the same individuals, to efficiency gains that accompany repetition, or to some combination of both factors.
As far as applications of game theoretic tools are concerned the arguments suggest that the emergence of a coalitional structure may be explained by cost differentials. Even if decisions are made according to considerations of expediency, differential advantages will favor the "Stay option" in every instance. The formation of stable subgroups of a larger population can be explained, thus, in a way that is compatible with the assumption of full control.
The situation is different if we assume that the relevant costs arise from commitment to a group rather than the other way round. Can such commitments to certain courses of action be rendered compatible with the assumptions of full control and consequentialistic explanations of individual action in terms of expedient choice? Utilitarians try to address problems like this using some notions of higher order opportunistic considerations. I will argue that they are thereby either prone to violate their own basic assumptions or cannot reach a convincing answer to the question of how stability comes about.
The utilitarian will accept that successful individual strategies may involve restraints on straightforward maximization or straightforward opportunism. Utilitarian theorists typically assume that individuals can make a choice over whole patterns of their choices. Following Viktor Vanberg's (cf. 1988) suggestion we may call such a choice of a "behavioral disposition" a "constitutional choice" on the level of the individual. After rational actors have chosen their constitution of individual behavior they do not decide according to considerations of expediency alone. They have "predecided" the issues on a higher level. If individuals make their constitutional choices correctly it seems as if they are deviating from maximization and opportunism while, as a matter of fact, they are maximizing their long-run utility.
This view, though quite common, is hardly compatible with the utilitarian aim to explain the outcomes of social interaction in specific situations as the result of opportunity taking behavior of individual actors in these very same situations. For, if individuals could really "chain" themselves to certain general strategies or to an individual constitution of choice that furthers their pursuit of long-term interests opportunistic behavior will not be effective in any instance of choice. What has been an expedient choice before, now typically will be a choice brought about by accepting a rule, by an acquired behavioral disposition, etc.
The question at issue is not whether restraints on opportunity taking behavior in the short run may be soundly based on individual interests in the long run. That this is true is beyond doubt. The real issue is whether assumptions like that of full control leave clearance for such behavior.
Putting the problem into game theoretic perspective it should be obvious [hat, for instance, threatening other individuals by announcing future reactions is often in the long term interest of an individual player. The threat is regarded as incredible, however, if the execution of the threat will hurt the interests of the threatening actor at the point of decision. It is assumed that the decision of whether or not the threat is executed is a new one which is under full control of the actor and definitely not predecided by or when the threat is announced.
Contrary to that the assumption that rational actors can at will adopt long-term strategies, regardless of whether they hurt their short-term or direct interests, would allow for credible threats of any sort. Arbitrary restrictions of individual strategy spaces would be possible and thus strictly non-cooperative games could never exist. -- This is not only absurd from the point of view of every day life it is also incompatible with the basic model of the rational human individual that prevails in utilitarianism and game theory.
Rule obedient behavior, in general, does not fit into models which explain all choices by their expected consequences. In the very instance of choice, rule obedient behavior cannot be explained by the incentives operating on the individual at that decision point. Regardless of whether it is a rule of thumb or an "obligatory" standard, and even if the adoption of the rule can be explained on a higher or an ultimate level as the outcome of a constitutional choice furthering long-term utility maximization, the proximate explanation will be based on something else than maximization.
Though utilitarians are right in insisting that there may be a latent tendency to suspend accepted rules or behavioral dispositions all the time, the adoption of the rule can only have a motivational effect, if its application, at least in part, is independent of case-by-case calculations of situational gains. To admit this, already implies that certain forms of behavior cannot be explained by direct reference to expedient choice alone. The same may be stated analogously for behavioral dispositions.
Again, the utilitarian may agree that the phenomena Pointed out exist while at the same time he may insist that they form what he regards merely as tiny crab grass in the green lawns of the utilitarian theory of social institutions. But, we are not talking about an unimportant or negligible aspect of institutional reality. Without some stable inner commitments to some rules, the amount of institutional stability which we, as a matter of fact, observe would be completely surprising. Even groups -regardless of some inertia stemming from costs of collective decision making -- could. in principle, follow a -juite erratic behavioral schedule (cf. on that Brennan and Buchanan 1985, chap. 5) and they indeed would in a changing world if all individuals in the group decide according to considerations of expediency all the time The gradual and stable evolution of institutional traditions, especially legal ones, would be a riddle without some restrictions on individual strategy spaces by adopted rules.
It may be seen quite easily, and sociologists were aware of this fact all the time, that such activities as blaming and praising other individuals either motivated by some natural behavioral disposition or by an accepted rule or standard of behavior are of essential importance for an understanding of tie maintenance of social institutions (cf. from a philosophical point of view, also Mackie 1985 a). Legal philosophers and sociologists of law like Herbert Hari (cf. 1961) have pointed out that the same holds for rule obedient behavior and the maintenance of formalized legal institutions.
To take but two examples of both realms, the institution of promise keeping and the institution of contract may serve as an illustration here. The first can only exist, if individuals apply the rules involved in the social practice underlying the institution of promise keeping as a guidance of own and for criticizing deviant behavior of others. They will feel retributive emotions if somebody deviates or complies with the standards of promise keeping. Such activities as praising and blaming others typically will not only be expected by the individuals directly concerned but also of "bystanders". Further, the: activities, at least in part, will not be explainable by the expected causal effects of an isolated activity of the class. But, if the gains of these activities are causally insignificant for the actor while the costs are not, why do they occur regularly along stable patterns?
The stability of the legal institution of contract also depends on the fact that at least some individuals sometimes deviate from opportunistic behavior. It would be absurd to assume that all judges (as "bystanders") evaluate all their decisions in terms of the causal effects that they expect for themselves. Of course, some judges may sometimes be motivated this way. But, a stable legal order depends on the fact that some participants will sometimes act simply because they accept certain rules while the causal consequences of their isolated acts are insignificant for them as evaluated in terms of their own altruistic or egoistic interests. (This alone explains why some individuals are deliberately put into the detached position of the judge.)
It seems obvious that in all human institutions, besides observable regularities in opportunistic behavior, there are some elements of inner commitment. It is a mistaken strategy to deny that inner commitments exist. Neither is it appropriate to take care of them simply by assuming different utility functions. Though this is possible all the time it will diminish the empirical content of the theory and it will only distract attention from the real phenomenon that human beings can develop inner commitments. The issue is not whether there are inner commitments but rather what their scope and limits are and how we car, integrate this factor into a general consequentialist and utilitarian model of social institutions.
The foregoing analysis may be summed up as stating that, as a matter of fact, in every game there are additional strategies that usually are neglected by game theoretic analysis. These strategies are "in the game" because they belong to the behavioral technology of human beings. We may take account of these strategies in two ways.
First, we may characterize the human individual along the lines of a multiple-self model. The rational strategist is, so to say, a higher self. This higher self will be confronted with additional strategies that arise from the "inner structure" of the human actor (or "acting body") that is governed by the decisions of the higher self. From this point of view, game theoretic models merely get more complex (cf. for a parallel model Margolis 1981, 1982). The drawback of such a view is that we either come dangerously close to Gilbert Ryle's ghost in the machine or that the actor of our theories is merely a theoretical construct, aid no real actor at all. In any case, it is not the human being whose behavior can be observed.
Second, we could add to our models certain strategies that do not model overt actions but could also take into account inner commitments as a kind of "meta-strategies". However, if we would assume that these inner commitments could apply universally to any strategy of the ordinary strategy spaces then we would again get into a position that seems to be patently absurd. In particular, it would become hard to see what purposes the central social institutions allowing for commitments could serve. Why should human beings incur the costs of organizing social cooperation by (artificial) social institutions if human nature provides them for free?
To avoid these absurdities, we have to assume that the rational actor can commit himself to a certain extent only. He cannot commit himself to the execution of future acts regardless of opportunity costs. If costs go beyond some threshold opportunity taking behavior according to considerations of expediency will prevail. Rational actors will switch from general behavioral programs or rule obedient behavior to a case-by-case evaluation of the expected causal effects, if costs get too high (and vice versa for lower costs).
The. latter remarks concur with some ideas proposed by Herbert Simon. Putting them into the perspective of classical game theory gives such ideas a different twist however. According to the assumptions of switching behavior any non-cooperative game contains some functional equivalents to commitment rules of cooperative game theory. If the payoffs in a game are high these functional equivalents will be negligible. They will be something like background noise and the normal game theoretic idealizations can, in general, lead to adequate models of social phenomena. In such situations of high costs, all forms of commitment power must be artificially or institutionally created, or come about by functional equivalents of commitment power that arise from high cost barriers to exit or to entry (cf. 2.3.).
If, on the other hand, not much is at stake then what may be called the technology of inner commitment (cf. for the parallel between moral commitments and " technology" Vanberg 1988, 11 ff.) will not be negligible. If low costs prevail a social interaction that otherwise would be modeled as an non-cooperative game may require a model that takes care of restrictions of individual strategy spaces. Not only overt actions have to be modeled as parts of the strategies of players but also inner commitments to specific courses of action or strategies. These commitments may considerably change the character of the game.
The latter suggestions may be easily applied to game theoretic analyses of social institutions. It seems to be quite clear that most of the time real world institutions contain more than one "social game". Typically, there will be a game of bystanders and individuals taking the role of rule or norm enforcers. In addition there may be a further game of enforcing the norms or rules of enforcement according to some norm or rule (cf. on this Axelrod 1986). Ultimately, on some stage of the hierarchy of enforcement games, individuals in the enforcement process interact under conditions of low costs (if they are not locked naturally in a small numbers' game as outlined in 2.3.). Parallel to. this there will be the normal cost game that, at least in part, is governed by the rules and norms created in the other game(s). Asymmetries in costs will play a key role. Individuals participating in the low cost game can exert externalities on others -- or even mutually on themselves -without incurring high costs.
This is essentially an extension of Coleman's (cf. 1983) stimulating argument stating that individuals, though not having a sufficient motive to contribute themselves, may have a sufficient motive to make others contribute to the provision of a public good. Whereas Coleman sticks to considerations of expediency his basic notion is here taken beyond considerations of expediency. This is of crucial importance for extending the division of labor to a regular and stable enforcement of informal and formal institutional rules. This is possible because, on the ultimate level, the game of life naturally contains some elements of a (low cost) cooperative game that are so to say wired in in human nature. Thus, social institutions that are organizing social cooperation can emerge regardless of the costs of organizing social cooperation. It may be noted finally, if only in passing, that this result basically reproduces David Hume's theory of government. At least this, James Buchanan will like.
<< References >>
|
Kliemt, Hartmut: Papers on Buchanan and related Subjects. – Munich (Accedo Verlagsgesellschaft), 1990: p. 72 – 95. (Studies in Economics and Social Science (SESS), Vol. 1 (1990)) |
[Top]