Part II, "Of Merit and Demerit; or, of the Objects of Reward and Punishment", considers what may be properly rewarded or punished. It introduces Smith's conception of justice, and differentiates justice from the other virtues, setting it aside as the basis of law. In particular, this section of Smith's book distinguishes between the two similar but distinct virtues of justice and beneficence.
First, Smith discusses reward and punishment in the abstract. Reward, he says, naturally follows from the sentiment of gratitude; punishment naturally follows from resentment. Reward and punishment are like opposite sides of a single coin: gratitude and resentment are both sentiments brought about by someone acting upon us, and their difference depends on whether the act was positive or negative.
In addressing the question of what may properly be rewarded or punished, Smith takes the logical next step: whatever is the proper object of gratitude ought to be rewarded, and whatever is the proper object of resentment ought to be punished. To determine what makes resentment or gratitude proper, Smith returns to the notion of sympathy: whatever resentment every impartial observer would find agreeable is proper, and the same is true of gratitude. This is nothing more than an extension of Smith's very definition of propriety.
Smith emphasizes that this theory is contingent upon the propriety of whatever act of harm or beneficence led to gratitude or resentment. If an act of goodness was committed frivolously, we cannot approve of the gratitude of the person who was the object of the act. Similarly, if a supposed injury was committed with perfect propriety, as in a gentlemen's argument, we cannot enter into the sentiment of justifiable resentment. When taken in its totality, then, Smith's notion of the proper objects of punishment and reward is contingent on a combination of what he calls "direct sympathy" with the person who acts (the agent) and "indirect sympathy" with the person who is acted upon (75-77).
Though reward and punishment seem similar, there is a significant difference between them: namely, reward follows an act of beneficence, and punishment follows a violation of justice. Beneficence must be given freely by definition and cannot in any justifiable way be enforced, whereas justice is the fundamental right of all and is enforceable by force. Justice is necessary for the security of society. Therefore, justice and punishment are the rightful components of law.
A punishment must be proportional to the violation of justice that is being addressed, the ultimate violation being murder, which can be punished by death. Smith also adds that the agent who violates justice will undoubtedly experience remorse for his action because, once his unlawful inclination is satisfied, he will have no choice but to reflect on his violation with the same coolness that is exhibited by the society around him. It is in this same way that he who acts in gratitude and deserves reward reflects as an impartial spectator might on his actions, and is conscious of his merit.
Lastly, Smith argues that this organization of the principles of justice and punishment is useful to society as a means of preserving order and security. This, he says, is the work of the beneficent and omniscient God who has designed all parts of the universe to function harmoniously, much as the parts of a clock do. To this end, God has designed our behavior and sentiments to support the proper functioning of society. Thus, though we do not always reflect on the good which punishment does for the preservation of society, we always approve of just punishment -- because, Smith argues, approval of just punishment is a sentiment designed by God to sustain a harmonious society.
A potent question here is whether Smith's argument regarding the utility of justice actually depends on the concept of God. Though later passages of the treatise involve God more heavily, it is not hard to replace "God" with "natural selection" here and reframe Smith's argument in terms of modern science (i.e. in the passage on 87-88). If we accept the principle that evolution selectively guides human development to enable the propagation of the species, it seems reasonable that typical human psychology would be designed so as to reinforce a harmonious society. An inherent approval of justified punishment could work in such a paradigm.
Smith is concerned with justice as a negative virtue -- that is to say, as the virtue of not acting in a way that injures other people. From this analysis, it seems that the idea of being a Good Samaritan would fall into the category of beneficence, which Smith believes cannot be enforced. This means that bystanders who witness injustice would not be guilty of injustice -- a stance which, while perhaps not very controversial in Smith's time period, would require significantly more justification in a post-World-War-II world. The current state of the world is more aligned to the idea that people who stand by and let evil happen are also guilty of injustice, though that injustice is not as reprehensible as the actions of those who actively promote evil.
An extension of Smith's conception of justice is the issue of injustices done to oneself -- for example, not taking care of oneself, hurting oneself, or otherwise behaving in a manner other than what one deserves. What actions ought to be taken in such situations? This is a more complicated matter than the issue that Smith chose to address, because we must at the same time consider the same person as the agent and the person acted upon. Moreover, any resentment that the person feels toward himself or herself would probably perpetuate self abuse. As a result, we would not be able to justly approve of this resentment in the way that Smith describes. Such an example appears to underscore the limits of Smith's framing of justice.
These cases notwithstanding, Smith's theory of justice in the abstract is very appealing, particularly regarding the balance which he says must be struck between direct and indirect sympathy. A practical difficulty here, of course, is that people are required to imagine themselves in two very different roles -- that of the agent and that of the person acted upon -- in order to judge a single situation. This is a skill which requires significant practice and even expertise. Again, Smith is able to get around this by resorting to the pedagogy with which he approaches moral theory: moral behavior is inherent to us, so we do not always need to be conscious of how it is working as we act; rather, we only need to recognize and agree with the descriptions of how our behavior operates, when such proper descriptions are offered. This is a significant presupposition, but if it is taken for granted, the rest of Smith's approach is solid.