Tuesday, July 26, 2011

Limited conventionalism and the law

For the past few years I've had an on-going research project of trying to use the framework David Lewis developed for analysing conventions in the interesting meta-ethical case where whatever basic principles we might have (for this purpose it doesn't matter what they are - one problem at a time!) don't give us enough guidance and we end up with equally attractive but mutually exclusive options and no principled way to choose between them. By now I've polished my approach down quite a bit, so here is a short sketch of it I've written today applied to a few examples from the literature and a sketch to how it might account for at least some instances of the origins and authority of the law.

Limited Conventionalism and the Law


In this essay I introduce my ‘limited conventionalism’ as a naturalistic model for the derivation of moral obligations. Limited conventionalism is the application of David Lewis’s analysis of conventions applied to the ethical case where our moral code fails to give us clear guidance. The problem is that almost any moral standard, and all of those putatively derived from natural facts about human beings, lack the detail to specify unique best responses to every situation, but instead leave us with a set of equally attractive but mutually exclusive options and no way to choose between them. Limited conventionalism is the claim that in such cases we can establish a convention recommending one of these best candidates over the others, that we would be justified in doing what the convention recommends and morally unjustified in not following it. Those features of human beings on which naturalists hope to construct an account of ethics would, on limited conventionalism, be the criteria by which the best candidates get picked out would be. In this manner we are able, I argue, to derive an adequate moral system from purely naturalistic grounds.
As an illustration of this point, I apply my model to the ‘pluralistic relativism’ of David Wong, whereby there are natural constraints on the type of society in which humans could flourish, but that these limitations radically underdetermine the type of moral code people can follow, leaving room for incompatible but equally good moral codes specific to a society. I also sketch out how limited conventionalism would do with other approaches, like the more thoroughgoing anti-realism of John Mackie or more modest views where what conventions offer aren’t new principles but only responses to particular problem situations.

Lewisian Conventions and Limited Conventionalism

A convention is a structure of expectations about a recurring co-ordination problem such that people in such-and-such situations do this rather than that. Because you have justified expectations of how others will act, and they of you, you can reason towards mutually satisfactory ends. The set of best candidates come from comparing the preferences of everybody in the situation and choosing the alternatives where everybody can do no better if all other parties act as expected. For instance, in the prisoner’s dilemma the two prisoner’s both confessing or both staying silent are the two best candidates, while one confessing and the other not isn’t, and if two of us are trying to arrange a meeting, us both being at some particular place is a best candidate while one of us being at one place and the other somewhere else isn’t. For a convention to arise, these best candidates need to be more attractive to those involved than having the co-ordination problem continue. An interesting feature about Lewisian conventions, which carries over to limited conventionalism, is that if you end up choosing one of the best candidate options, it makes no difference how that choice is made. What matters is that it is common knowledge which course of action gets recommended. Two other pertinent features of Lewisian conventions is that none can arise which aren’t mutually beneficial, and once a convention exists, you are obligated to follow it.
To apply Lewisian conventions to the moral underdetermination case, we first arrange the available options based on how well they conform to some shared moral standard. Then we separate off a class of best candidate options which perform no worse than any other options in light of that shared standard. The convention then gets established to recommend one of these candidates rather than the others. Once again, the manner in which one gets chosen doesn’t matter, as long as it is one of the best candidates and the recommendation becomes common knowledge. Thus, limited conventionalism is a way to extend the set of shared moral obligations within a community.
Drawing up a convention is thus a two stage process: first at least some of the best candidate options are identified, and then (through whatever method) one of them is chosen and that choice is made common knowledge. This leads to an epistemic structure which allows individuals to navigate the underdetermination problem case: given that the chosen option is so as a matter of common knowledge, everybody expects everybody else to follow it, everybody knows that everybody has that expectation and that they are also expected to conform, and so on. Following the conventional option is thus something which can be done as a matter of course, thus avoiding the uncertainty which threatens underdetermination problem cases.

Why we are obligated to follow these conventions

The normative force of these moral conventions come from the fact that they are benign in light of the existing obligations, no individual can have a moral reason to go against the convention, and by going against the convention you would undermine other people’s depending on your co-operation in order to reason towards their desired ends. Conventions are benign respective to the existing moral standard because the candidates among which get chosen are themselves the most attractive in light of that that standard. This means that we can’t arrive at a convention which goes against our existing moral standard. Furthermore, nothing moral is to be gained by disregarding the convention, because no other option can appear more attractive respective to the moral standard. Finally, since the conventions are common knowledge, every party has the well-established expectation that everybody else will conform to the convention, and reason accordingly. To go against the convention would be to undermine others’ ability to reason to their desired ends, dumping everybody back in the less preferable, pre-convention co-ordination problem. In this respect the harm to others is very much like that which follows from lying, and not conforming is similarly unjustified, whatever the content of the convention (if it is a genuine Lewisian convention).

Applying limited conventionalism to examples in the literature

A position in the literature favourable to such an analysis is David Wong’s pluralistic relativism. Wong holds that there are certain natural constraints on the type of societies we could live in, with concomitant constraints on moral standards. However, these constraints radically underdetermine the content of morality. To model Wong’s relativism as an example of limited conventionalism, take the shared moral standard of some community to track those natural constraints. There are now a range of best candidate options corresponding to all the moral codes which fit within those constraints. The choice of any such a code counts as a Lewisian convention. Considering the case where there are many societies each facing the same underdetermination problem, which Wong plausibly argues is the actual case, the possibility of different codes being chosen by different societies cements the possibility of relativism. Notice that it is possible that single convention could establish an entire moral code in one go, but that it is far more likely that the code is constructed piecemeal, each new obligation derived conventionally limiting the ones that follow.
We can model more radical theories like J.L. Mackie’s and others wherein all moral obligations are conventionally determined by considering the case where there are no pre-existing moral obligations. Then the set of best candidates would be all the possible candidates. We could establish a moral code by choosing a starting point on some indeterminate standard of salience, most likely something basic and immediately useful like a limitation on killing, and then choose a best candidate limited by that one obligation, and then one by the set of two conventionally established obligations, and so on. We can also account for less thoroughgoing conventionality by considering the case not where we use conventions to choose new general principles, but rather responses to particular co-ordination problems. An example might be the choice between an adversarial or inquisitorial court system. It is in this weaker application, addressing examples of underdetermination case-by-case, that I believe conventions might be the most prevalent and useful.

Conventional Law

We can perhaps extend limited conventionalism informatively to cover some instances of law-making. Examples of underdetermination abound in legal scenarios, and there exists a range of mechanism to try and cut down their scope, like guidance for judging the intent of legislation when the letter of the law allows ambiguities. Nonetheless, it seems clear that there are vast arrays of cases where there is genuine uncertainty about what the appropriate judgement would be, whether it be a verdict, a judgement on some point of law, or passing sentence. Given the weighty reasons we have to the law be applied consistently, legal judgements have a dimension wherein they are co-ordination problems. The system of widely publicising court judgements and ensuring that practicioners of law are expected to know of relevant precedents does as much to establish common knowledge as could be in any such technical and contingent domain. As discussed above, given the shared interest in avoiding co-ordination problems (here, manifest in the inconsistent application of the law) and the fact that only conventions could prevent them (since any effort at co-ordination would necessitate the strcture of expectations which constitutes a Lewisian convention) the laws would have the necessary normative force to not be dead letters, as long as all the preconditions are met (especially that the option chosen is genuinely one of the best candidates and thus as consistent with the principles as any other). Correspondingly, in at least some significant cases, the law could be conventional in the Lewisian sense, and thus be examples of limited conventionalism in my sense.


Limited conventionalism provides a robust and informative model for the forming of moral obligations. Many significant applications of the model can be made to positions in the literature and in actual life. This mechanism carries the necessary normative weight to give us the reasonable expectation