Prior to laying out the "psychological construction" in detail Rawls first indicates that questions of justice arise in situations where people who have no position of authority in relation to each other have to settle or acknowledge the rules of the institutions they participate in and determine shares in benefits and burdens. This leads to stating the two principles of equal rights to the most extensive liberty compatible with like liberty for all and that inequalities are treated as arbitrary unless it is reasonable to expect they will work to everyone's advantage. The former principle is very Kantian in formulation. Explicating the principles leads Rawls to restate the "analytic construction" of the original position again.
After going through the "analytic construction" Rawls returns to instances in which the principle of equality applies, indicating that they relate to institutions as part of the definition of them. This is slightly odd since the definition Rawls earlier gave of institutions only specified that they were "publicly recognized systems of rules" which doesn't seem to imply the principle of equality unless there is a presumed analytic link between publicity and equality. The second level at which the principle of equality is said to operate is in terms of the structure of an institution or social system and this requires seeing them as bound by the two principles derived from the "analytic construction". This neatly erases the difference between the principle concerning inequality and that concerning liberty since the latter included the notion of "equal liberty" but that seems rather too neat a move. Finally, the principle of equality is said to apply to the original position itself and that point is what is used to indicate the need for a general account of moral psychology as it presupposes a view of how cognition relates to morality (or a principle of practical reason).
Rawls' second question concerning what accounts for the fact that we and others do what justice requires indicates a need to find a view of how action takes place that is not, on Rawls' view, purely understood through the concept of rationality though what he seems to mean by this is that is it is not grasped by theoretical reason. Having placed these preliminary points in place we move on to the "psychological construction" which is described by Rawls as being something that is not to be taken empirically though it is to be thought of as "reasonably plausible".
The "psychological construction" is grounded on an idealisation of Piaget's view of the development of guilt in children where Piaget distinguished between authority guilt, association guilt and principle guilt though again the place given here to guilt is said by Rawls to be only "a matter of convenience". This use of guilt has, it should be said, rather large overtones, since it is, after all, central to Christian morality in particular. Authority guilt is characterized easily in relation to positions of children who basically accept that there are people - parents - in constituted places of authority whose injunctions are meant to be obeyed and so authority guilt is what is felt when we do not obey. This requires elaboration of a notion of rational self-love which requires taking seriously someone who, whilst in authority over you, nonetheless manifests some serious care for you. It does, however, assume that the one authority is exercised over does not possess serious standards of their own and is thus rightly governed by judgments they are incapable of making themselves.
The second form of guilt, association guilt, relates not to the standards of authority figures but rather peers with whom one has formed bonds of relation such as friendship and mutual trust. Breaking bond with these others produces association guilt. It is with regard to this that Rawls first makes manifest the whole point of this discussion is to enable thought about the "stability" of just relations since if others will not follow bonds or are generally thought likely not to then this loosens the whole association. Intimations of game theory are clearly present at this point of Rawls' analysis as becomes clear when he uses the discussion to interpret Hobbes:
One way of interpreting the Hobbesian sovereign is as an agency added to unstable forms of cooperation in such a way that it is no longer to anyone's advantage not to do his part given that others will do theirs. (104)
Friendship and mutual trust are seen, on this pattern as kinds of games of which the Hobbesian social contract is a supreme example.
The third form of guilt, principle guilt, requires separating ourselves from ties to others of any particular specified type and seeing how injury of anonymous others still produces in us feelings of guilt. This is taken to arise from the schemes of cooperation having been extended beyond their point of origin (a kind of Rawlsian "genealogy"). It requires then we accept and make explicit general principles of justice. If everyone in a system is known to have such feelings of justice then we have a situation that Rawls terms "stable".
The whole account relies, as Rawls himself states, on the notion of "moral feeling" but, whilst this notion can be found in Kant, Rawls has not referred at all to Kant in expounding it and his next move is, rather, Wittgensteinian in general flavour as he now speaks of characteristic linguistic expressions, behaviour manifestations and sensations that accompany moral feelings rather than trying to give any kind of transcendental account of moral feeling.
The account that follows of why we act on our duty of justice in particular cases is thus one that assumes the idealised anthropological account to broadly hold. Rawls sketches the idea of someone who does not feel a sense of justice but it is evident that this idea is merely a limit-case and not one that he takes seriously as something that could really arise.
The final discussion of who the obligation of justice is owed to revisits again the original position in order to describe such a situation as a "natural lottery". Finally Rawls effectively identifies the sense of justice with Kant's notion of a good will which is done in a preemptory way. The general argument has been intended to show that the capacity for a sense of justice is the key aspect of moral personality in the theory of justice but it is fairly evident that the argument of this piece is less than convincing.