Abstract: Moral dilemma is a type of decision-making situation in which moral principles are in conflict. Under this condition, a decision maker may violate certain moral principles whether he or she endorses or rejects a behavior proposed for a given situation. As research on this topic has advanced, scholars have developed a variety of research paradigms of moral dilemmas. Thus, a systematic analysis of the development process of moral dilemma research paradigms is needed. Such work would not only provide a more accurate methodological understanding of moral dilemma research but would also provide a methodological reference for research on similar psychological conflicts. This paper systematically described the advantages and disadvantages and theoretical value of four current research paradigms in the study of moral dilemma: the classic dilemma (CD) paradigm; process dissociation (PD) paradigm; the consequences, norms, and generalized inaction/action preferences (CNI) model; and the consequences sensitivity, overall action/inaction preferences and norms sensitivity (CAN) algorithm. CD paradigm makes the utilitarianism principle of maximizing the consequences of conflict with the deontology principle of pursuing the normative behavior itself, making it possible to explore moral principle preferences in moral choices. However, it cannot independently measure preference for the utilitarianism principle and preference for the deontology principle, which makes the interpretation of results ambiguous. Based on this paradigm, PD paradigm considers both utilitarianism and deontology principles against a proposed behavior and independently measures these preferences in moral decision-making. PD paradigm has helped resolve many research controversies. However, it does not completely dissociate moral decision preferences. For example, it cannot demonstrate the inclination to generally endorse/refuse a behavior while not considering any moral principles. Particularly, it confounds the inclination to generally refuse behavior proposals while not considering any moral principles in the preference of deontology principle. In this way, it systematically overestimates deontology preference. To overcome this limitation, the CNI model applies a multinomial processing tree model with a factorial design of moral scenarios (norm: prescriptive/proscriptive × consequences: benefits greater/smaller than costs), so that the preferences of generalized action/inaction irrespective of moral principles can be estimated. However, it presupposes that the agent sequentially considers norms—consequences—generalized inaction/action preferences, which results in an overestimation of norms sensitivity and makes the generalized inaction/action preferences untenable. To address this important limitation, CAN applies a new algebraic method to estimate norms sensitivity, consequences sensitivity, and overall action/inaction preferences in moral decision-making. In the development of these four paradigms, later ones resolved the limitations of former ones. Different psychological processes are gradually uncovered with new developed paradigms. The development of research paradigms reveals some directions for future research. First, some controversies in moral decision studies might be due to fundamental limitations of earlier research paradigms. CNI and CAN can be used to solve these controversies. Second, the moral theories based on early research paradigms should be reevaluated using the new paradigms, to verify the tenability of these theories. Third, these research paradigms could be applied to explore other psychological conflicts. In sum, this paper systematically reviewed the development procedure of the methodological paradigms in moral decision-making. It provided a methodological reference for moral dilemma studies and related research.

Key words: moral dilemma, process dissociation, CNI model, CAN algorithm, moral decision-making