18 mars 2015 Répondant : Mikael Cozic In classical rational choice theory, an agentís choice behaviour is explained by means of a preference relation (assuming certainty). By contrast, a reason-based choice model explains behaviour in terms of the motivationally salient properties of the options and/or the context. Such a model specifies what matters to the agent, and how it matters. More precisely, it specifies : -* - the properties which motivate the agent, captured by a motivational salience function assigning to every context the set of properties which are motivation- ally salient in this context, and -* - the way in which these properties motivate the agent, captured by a fundamental preference relation or weighing relation which ranks property combinations. What can rational choice theory contribute to moral philosophy ? A first and only apparent obstacle is that, in its standard interpretation, rational choice theory is concerned not with moral choice, but with actual choice. Indeed, choice models are usually interpreted positively, in view of explaining and predicting the agent’s actual choice behaviour. But one may adopt a normative, and more specifically moral interpretation, which turns a choice model into a model of moral choice, concerned with what the agent ought to do. Thus reinterpreted, a reason-based model concerns normative reasons rather than motivating reasons. More precisely, a reason-based model in its moral interpretation specifies : -* - which properties matter morally, as captured by a moral salience function as- signing to every context the set of morally salient properties, and -* - how these properties matter morally, as captured by a (moral) weighing relation between property combinations. This paper introduces reason-based models in their moral interpretation, and shows that such models can illuminate various classical debates in moral philosophy by offering a unified framework in which rival metaethical positions can be captured, compared and perhaps better assessed. A standard moral theory – for example classical utilitarianism or a particular deontological theory – gives rise to a corresponding choice function , which to every choice context assigns the set of actions deemed morally permissible ; but it also gives rise to a corresponding reason-based model , which captures what properties matter and how they matter according to the theory. For in- stance, classical utilitarianism gives rise to a particularly simple reason-based model : in short, only total pleasure matters, and more of it is better. In a nutshell, the following is shown : -* - Whether or not the moral theory is consequentialist is reflected in whether or not only properties of the consequence ever matter. This can be read o§ from the moral salience function : the theory is consequentialist if and only if the moral salience function is not context-regarding (in a technically well-defined sense). -* - Whether or not the moral theory is relativistic is reflected in whether or not what matters varies with the context. This is again reflected in the moral salience function : relativism holds if and only if the moral salience function is context-variant (in a technically well-defined sense). -* - Whether the moral theory is monistic or pluralistic is reflected in whether only one or more than one property can be morally salient in a context. This is again immediately a feature of the moral salience function. -* - Whether the moral theory is atomistic or holistic , and whether it is generalistic or particularistic , can be cashed out in terms of features of the weighing relation. In short, atomism and generalism translate into certain separability features of this relation. We also present axiomatic characterizations of the choice-behavioural implications of consequentialism and relativism. It will turn out that every relativistic theory is choice-bahaviourally indistinguishable from some non-relativistic (and usually non- consequentialist) theory.