Morality, decision making, payoffs, and other games
Morality - inherent or not?
Durkheim spoke of morality in the sense of rules and regulations. But Weber was concerned with morality in in the sense of values acquired to elaborate fact. Both he and others realised that as morality is acquired, then it is subject to control by authority and reframing. What they did not address was that if morality is inherent, then such control needs to be differently applied - total reframing may not be possible. In other words, the inculcation of morality – or rather of a governing view or right action – must proceed differently if morality is inherited rather than simply conditioned.
Piaget studied groups of children from 1932 through to the late 1960s in order to better understand moral decision making concerning beliefs about right and wrong. From his work he came to believe that all development emerges from action. That is to say, individuals construct and reconstruct their knowledge of the world as a result of interactions with the environment, (note how closely this matches
Bourdieu's view of the effects of culture). Watching rules develop amongst children as they played, he became convinced that as the rules differentiated right and wrong, so too did a resultant morality develop. He concluded that morality stemmed from an inherited ground rather than from genetic predisposition.
He believed that young children were in what he termed a heteronomous state of moral reasoning. This state he claimed to be composed of massive egocentrism centrism due to the nature of a young child’s cognitive structure (genetics), and also due to the natural and uni-directional flow of authority from parent toward child. However as the child aged the heteronomous state was insufficient to allow for the cooperative decision needed for group interaction. Hence the change from a heteronomous state to a state of morality – where rules of social interaction converted to a ground of fairness. And thence to more universal notions of right and wrong.
Thus it may be said that Piaget believed that children did not simply learn and internalise normative behaviour for a group as Durkheim had proposed, but rather that a child defined morality individually during the struggle to arrive at fair interactions and solutions in group dynamics. Or said another way, Durkheim believed morality could be instilled; Piaget believed it was acquired through interacting in cooperative decision making.
As a result Piaget wrote that the state must emphasise cooperative decision-making through problem solving. The state would thereby create a ground of moral development through discovery of common, fairness based, rule sets. Within the early constraints of cognition and heteronomity, a child therefore should be free to choose a path to moral, and hence social, behaviour.
The empirical evidence that Piaget presented was largely observational and interpretive, and somewhat (very?) difficult to replicate by others interested in moral development. However later researchers, particularly Kohlberg developed Piaget’s ideas much further, with more solid empirical support. Kohlberg elaborated the concept of heteronomity and subsequent moral development into a multi-stage theory. It extended Piaget’s moral growth to development-linked stages in older children.
The stages of his theory he said where linked to implicit values, and were critical in showing how and when a child organises understanding of virtues, rules, and norms, integrating these into moral choice . Unlike Piaget however, his research led him to conclude that certain rules of justice and fairness were inherent moral principles in all cultures. This because moral development he concluded, was logical in terms of successful group cooperation. Not a doxa as such, but more a ground from which doxa may spring. And not inherited from genetics, but inherent in the very nature of cooperation.
The goal of the state therefore, was to help/guide/entice students in their growth toward moral agents within and for a community. The state was to promote rules and norms upon children which acted for community justice and fairness. But the primary responsibility of enforcing these rules should rest with the students themselves, although of course ultimately subject to the ’guidance ’of state authority.
The idea was to encourage freedom of choice and decision making – free will – so long as the choices presented were appropriately matched to the student’s cognitive stage and to a fundamental ground rule of not harming the community.
Needless to say, outliers - the very bright, very slow, or very violent - would not fit this model and would have to be controlled by some other means. Means which usually meant use of force.
Many other schools of moral development have sprung up, but most are variants of the two themes above: 1) moral action is the result of inculcation; 2) moral action is developmentally based. In the latter view, the role of the sate is to emphasise reflection, perspective taking, conflict resolution, and autonomous choice. In the former view, the role of the state it is to direct through hidden and overt rules which do not exclude force.
Proponents of both viewpoints suggest that the state is vital in the creation of moral behaviour, where moral behaviour is implicitly defined as being co-operation and operation within the rules laid down by the state, of which ones individual community is but a representative part. The goal was to produce an individual who cooperates with the society and social group for which she was destined, and whose ontological view is precisely that of the group.
Situational rationality and decision making
Suppose that while grading an essay it is discovered that undergraduate students have produced virtually identical papers. The students are arrested, placed before a firing squad, and shot... oh no, sorry - that’s just what their professor’s wishes would happen. Sigh... instead he is forced to go through a tedious time-consuming process set up the the Board of Governors, which begins with interviewing both students separately.
From the student’s point of view during the interview process, there are two choices: 1) Confess to cheating and implicate the other student, or 2) do not confess to any wrong doing. The possible outcomes for this situation are as follows:
1. Neither student confesses: In this case a mild punishment (M) of a 10% reduction in grade point average will be given.
2. Both students confess and implicate one another: Here a fairly severe punishment (FS) will occur. That is, both students will have to redo the course when it is next offered.
3. Only one student confesses and implicates the other: University policy is such that it encourages students to turn in cheaters. This includes those who cheat themselves. Hence, a very mild (VM) punishment (having to redo the paper) will occur for the cooperating student, in order that other potential whistle blowers will be encouraged to come forward in the future.
4. However the student who did not confess and did not cooperate but who has been implicated by the other student in (3) will receive the very severe (VS) punishment of expulsion.
There are many ways to study the decision making and moralities involved: Optimistic payoff decision trees could be used. These fall within the general field of graph analysis and are very useful for evaluating problems involving minimisation – such as when seeking to understand opportunity gain or loss for adapting a strategy of acquiescing to authority in an undergraduate environment. Another technique is sensitivity analysis. This can be used to see how estimates for the states of nature affect or alter the recommended decisions – for example how the expected value of multiple disparate choices in an undergraduate career effect the expected value of the probability function for success. But the most commonly used is a simple payoff matrix, which we can represent as follows, where the number to the left of the comma represents the the payoff for student Mary and the number to the right of the comma the payoff for student Alice:
|
|
|
Alice
|
|
|
|
|
Confess
|
~Confess
|
|
Mary
|
Confess
|
1:FS,FS
|
2.VM,VS
|
|
|
~Confess
|
3: VS, VM
|
4. M,M
|
This is a version of the popular prisoner’s dilemma first mentioned by Tucker during a lecture at Stanford. But it should be noted that this matrix is not the decision table commonly found in say, economics books. Because this matrix consists of interactions amongst the cells, as well being bound by the strict rules of composition and interpretation according to game theory.
Of interest here is what occurs when a student confronted by a firing-squad daydreaming professor attempts a rational decision that will maximise her payoff (i.e. land in the VM zone). Mary will probably reason quickly and rationally as follows:
"Two things can happen: Alice can confess or Alice can keep quiet. If she confesses, then I will be expelled if I don’t confess (VS) [cell 3], but only have to redo the course if I do (FS) [cell 1], so clearly it is best to confess. However if Alice doesn’t confess, and I don’t either [cell 4], I’ll only get a bit off my grade (M) which I can live with. And if Alice doesn’t confess but I do [cell 2], I get to rewrite the paper (VM) and can probably get a good grade on it. Obviously, it’s best for me if I confess. So that’s what I’ll do.”
This is sound and proper reasoning, so it is highly likely that both Alice and Mary will come to the same conclusion. If they do so, their sound and independently reasoned individual choice in decision making, will result in a fairly severe punishment (FS) for both, rather than the easy result they had reasoned toward. Or said another way, individual rational decisions in non-zero sum environments (i.e. where the rewards are not the same as the punishments) results in both individuals being worse off than if they had acted ’irrationally’ and not confessed.
Now clearly there are artificial constraints here: the two students are not allowed to communicate; there are only two students involved – there could have been many more; one student could plead temporary insanity; and so on. It has been shown by Axlerod, Bendor, and many others that when when some or all of these constraints are removed different decisions can be made. In particular, it can be shown that should the students have been allowed to cooperate and iterate their choices (test the results of their decisions) they would have most probably made different decisions.
Where situations have clear outcomes, rules, and precursors – such as in the case just described, one can show that rational choice generally produces unwanted and unforeseen results. Said another way, basing moral decision upon a ground of common sense reasoning may not provide a good outcome. And it may not do so most of the time.
Anatole Rappaport – one of the principle architects of game theory – has said that there is no rational strategy that will lead to an optimal outcome in a non-constant sum game, such as the one between Mary and Alice. This is regardless of the decision strategies used. He believed, and mathematically supported, the notion that only in iterative cooperative games where participants communicate and have several opportunities to view results of their strategies, was there a chance for a truly rational optimal outcome.
It can be shown further, that every game has an equilibrium position - John Nash was awarded the Nobel Prize in part for his proof of this fact. At the point of equilibrium risks and rewards are balanced for all players. Obviously therefore moving the equilibrium point impacts profits. The question is, what variables should be changed and by how much in order that profits are maximised for the selected players. To calculate this techniques developed by Genichi Taguchi can be used. Taguchi’s methods allow for the testing hundreds or even thousands of variables without the need for more than a few experiments. Essentially his methods applies a codified reverse-engineering where only two variables are needed to define any black box system (in theory at least). Particularly though, for linear systems wherein: I) Parameters in which level or quantity affects process variation; II) parameters in which process variation is unaffected by level or quantity.
Grossly simplifying Taguchi’s methods, the technique is to set the first types of parameters at a level which minimise total process variation, while using the second types of parameters to control/adjust the total process. The Taguchi expectation then, is that rational behaviour will result in maximum return on investment by any participant or player. (Note the moral implications here.)
Now, in a traditional Nash Equilibrium experiment all bids are known. But in one where noise (unknown parameters) effect the outcome to a significant extent – weather forecasting for example, or the role of lobbying in deciding which oil-rich country to invade next, etc. – Baysean probabilities enter to push the Nash equilibrium in a non-linear fashion. Taguchi methods can use orthogonal arrays to isolate some of these noise factors from others in a cost effective manner; Bernoulli-Nash equilibria optimisation can do the same.
And so there is always an equilibrium point were risks and rewards can be balanced; and there are methods of optimising the decisions – both game theoretic and others such as Taguchi, concerning payoff maximisation to selected players.
This has many implications for decision making, and for morality in general. Because rather than being merely the nature/nurture concept implied by Piaget and others, moral decision making is also boundaried by the mathematics which applies and a person’s ability to calculate the payoffs involved. To give a single example, a subject’s core beliefs can be altered by application of the mathematical evaluation of plausibility in Nash equilibrium, as demonstrated by Bicchieri and others.
Now consider the well known ethnomethodological experient of H. Garfinkel from several decades ago. He had a number of medical students interviewed on the pretext of discovering why medical school entrance interviews were stressful. The experimenter posed as a representative from a highly prestigious medical school. During the interview session, the students were played recordings of a pompous, rude, poorly mannered, interviewer verbally abusing medical school applicants. The experimenter made it clear that he not only approved of the interview, but felt that this was in fact an ideal interview for his school. The medical students were then asked for their opinion and analysis of the recorded interview. As is typical of findings in the studies of authority (particularly Milgram’s famous work regarding torture), the results showed that most of the medical students worked very hard to reconcile their opinion with that of the experimenter. It was argued that this in turn depended upon the degree to which an internalised conformity to authoritarian views has occurred in the participant, an acculturation imbued from the beginning as absolutely essential to success in most medical schools .
A simple cooperative payoff matrix may be useful here. (S) represents a subject under stress (S) reduction, and the payoff to the experimenter is cooperation (C) from the subject.
|
S= stress level. C= cooperation
|
|
Experimenter
|
|
|
|
|
Agree with tape
|
~Agree with tape
|
|
Med School Candidate
|
Agree with tape
|
1:<S, >C
|
2.>S, <C
|
|
|
~Agree with tape
|
3: >>S, <<C
|
4. <S, >C
|
This matrix can be read as follows:
If the Medical school candidate agrees with the tape, but the experimenter overtly disagrees, then the there is increased stress for the subject and a perceived decrease in cooperation by the experimenter (cell 2). But the stress is much greater for the subject if she disagrees with the tape while the experimenter clearly agrees with it since she is now presenting two non-cooperative behaviours: (1) she is disagreeing with the tape and (2) she is disagreeing with the experimenter (cell 3).
Of course there is some obvious arbitrariness here. Yet what is of interest is that this matrix that what is being described are payoffs for degree of acculturation. I believe it is quite valid to apply this type of analysis toward both the decisions stemming from core beliefs and the payoffs from acculturation. The resultant process of decision making it can be argued, may not be as much effected by logical considerations as much as by the need to automatically map an action strategy to the hidden ground of acculturation. Moral reasoning as it where, by means of payoff.
Human decision making is not normative (rational) but is rather subject to risk apprehension and anticipated regret (viz. Kahneman’s work). Morality, being merely another form of decision making, is similarly subject to such apprehension.
Studies of complex n-person games and games of risk suggests that co-operation is inherently productive. When individuals or groups perceive incentive (gain) through strategic action (risk taking) these mathematical theories show that the best strategy is often one of conformity. The conformity strategy optimises payoffs for the majority of group members. (Note the similarity to the post-modernist’s ideas regarding symbolic capital.) Such conformity can be obtained most efficiently if there is some signalling mechanism between players in the game – common practise, common language, common action all are such mechanisms. When the signals and strategies mesh into consensus, I would suggest that not only is the payoff likely to be higher for the majority of non-competitive games (i.e. most social actions in group environments), but additionally social norms are internalised. Optimising payoffs leads to more complete acculturation. Or to put it another way, decision making becomes aligned with the wishes of the group. Since this maximises payoff, one might go out on a limb and suggest that the efforts of all states to mandate conformity through shared moral viewpoint (“my country right or wrong”) has more to do with the mathematics of risk assessment than of true moral choice.
Further, those who internalise a norm cooperate even when the payoff for cooperation is smaller than the payoff for dominance. This has been called principled conformity to a norm by R.D. Cooter. Those who fail to so internalise, cooperate only when some objective result (payoff) for cooperation is at least as high as the payoff for dominance. This is known as adventist conformity. For example, in studies of changing environmental attitudes, it can be shown that cooperation is enhanced in groups where participants share a common ethos (eg. by J.M. Baland et al).
There are therefore despite the normal philosophical approaches to morality, clear choices (both overt and covert) to be made: whether or not to conform; the degree of conformity; whether or not to replicate the doxa, ... and so on. These choices are in essence risk assessments which can be represented by 1-N person payoff matrices, which are IMHO, part and parcel of how the brain works to compute choice. Whether that choice be moral or simply upon which foot should be socked first, is largely irrelevant.
