Does the fundamental tension in moral cognition between partiality and impartiality make resource-rational contractualism unfalsifiable?

The tension between partiality and impartiality is a fundamental feature of human moral cognition. On the one hand, we are undoubtedly partial creatures. We favor ourselves over others, our kin over strangers, and our in-group members over out-group members. On the other hand, we also value impartiality, perhaps especially in highly reflective or deliberative settings. Tenets of impartiality can be found in many religions, including Judaism, Christianity, Buddhism, and Islam. Impartiality is also at the core of many contemporary legal systems and influential philosophical theories, such as utilitarianism and various formulations of deontology. Indeed, taking an impartial stance – ideal observer theory (Smith, Reference Smith and Haakonssen2002), the point of view of the universe (Sidgwick, Reference Sidgwick1907), the view from nowhere (Nagel, Reference Nagel1989) – is normatively appealing, and we expect judges, umpires, and statesmen to demonstrate impartiality.

This conflict is also central within the contractualist tradition in moral philosophy. Following Hobbes (Reference Hobbes1651/Reference Hobbes2018), contractarians like Gauthier (Reference Gauthier1986) are mostly concerned with agreements among self-interested rational agents. They recognize inequalities between the contracting parties and put relatively less emphasis (but not none) on impartiality. Following Rousseau (Reference Rousseau1762/Reference Rousseau2004) and Kant (Reference Rousseau1785/Reference Kant, Hill and Zweig2003), contractualists like Rawls (Reference Rawls1971) and Scanlon (Reference Scanlon1998) put relatively more emphasis on impartiality, stressing the equal moral status of the contracting parties and procedures designed to prevent parties from taking advantage of their actual positions (e.g., the veil of ignorance).

This tension is also apparent in resource-rational contractualism, which borrows from both the contractarian and contractualist strands. Thus, Levine et al. are very explicit that there are many circumstances under which resource-rational contractualism predicts partiality. For example, the Nash bargaining solution and other game-theoretic models are sensitive to asymmetries in bargaining power. Similarly, the weights that we attach to others’ welfare are highly parochial. In this regard, one key cognitive mechanism posited by resource-rational contractualism, virtual bargaining (“What would we agree to do in a negotiation?” (Chater, Zeitoun, & Melkonyan, Reference Chater, Zeitoun and Melkonyan2022)), is fundamentally partial. Simultaneously, Levine et al. also explicitly acknowledge the importance of impartiality in their theory, mentioning “society-wide” bargains (p. 21), preferences for “equal, universal moral rights” (p. 8), and egalitarian rules (p. 21). Crucially, the theory gives a central role to universalization processes – “What if everybody did that?” (Levine et al., Reference Levine, Kleiman-Weiner, Schulz, Tenenbaum and Cushman2020) – which are fundamentally impartial.

In empirical tests, resource-rational contractualism can thus be compatible with both partial and impartial moral judgments. This makes it flexible but at the risk of being unfalsifiable. Take a simple example: what is the fair way to split a pie of fixed size between two agents with unequal bargaining power (Le Pargneux & Cushman, Reference Le Pargneux and Cushman2025; Le Pargneux, Roberts-Gaal, & Cushman, Reference Le Pargneux, Roberts-Gaal and Cushman2025)? Under a contractarian interpretation of resource-rational contractualism, the split should favor the agent who is in a better bargaining position, as predicted by virtual bargaining and the Nash bargaining solution. Under a contractualist reading, an equal split may be preferred, as it seems to coincide with what rational agents unaware of their personal circumstances would agree to under a Rawlsian veil of ignorance. When does resource-rational contractualism predict partiality vs. impartiality?

One immediate solution could be to use the resource-rational component of the theory to derive precise predictions. Perhaps cognitively cheaper processes favor partiality and cognitively more expensive processes favor impartiality, or the other way around. This solution seems unlikely to succeed. First, some (cheaper) model-free heuristics are impartial (e.g., “split resources equally”), but others aren’t (e.g., “split resources in proportion to bargaining power”); some (more expensive) model-based processes are impartial (e.g., universalization), while others are less so (e.g., virtual bargaining). Second, any workaround directed at partiality and impartiality also needs to convincingly address other versions of the same fundamental problem, which reappear when considering further conflicts within the moral mind: equality vs. equity, individual vs. collective property, duty vs. consequences, etc.

A better solution instead views resource-rational contractualism as a theoretical framework aimed at mapping out the cognitive organization of morality: how the subprocesses and mechanisms involved in moral cognition are organized and interact with one another (Le Pargneux, Reference Le Pargneux2025). Under this approach, the key empirical question is not “Are moral judgments in this task on average partial or impartial?” but rather “Which cognitive processes are involved in producing moral judgments in this specific context?” In other words, the focus moves from the content of moral judgments to how they are produced. Indeed, consider that observing a general preference for equal splits in an experimental paradigm is quite uninformative on its own, as it may result from very distinct processes: quick heuristics, deliberative reasoning processes (universalization, aggregate expected utility calculations), explicit application of religious (“love your neighbor as yourself”) or philosophical (the categorical imperative) teachings, etc.

This shifts the aim of empirical investigation from mere directional predictions to understanding how people make moral judgments and decisions. Are people using heuristics or rules, and which ones? Are they implicitly weighting the welfare of others based on kinship or in-group favoritism? Are they explicitly asking themselves, “What if everybody did that?” or implicitly looking for arrangements that ensure mutual benefit? To assess this, we can complement measures of choice with (individual-level) computational modeling (what precise quantitative predictions are made by each candidate process?), process-tracing methods (think-aloud, eye-tracking, mouse-tracking), analysis of free-response data (“how did you make your choice?”) and response times, and manipulations aimed at disrupting the tradeoff between accuracy and cognitive costs (stakes, time pressure, cognitive load). Recent empirical work by the authors illustrates the insights that can be gained by adopting this approach (Levine et al., Reference Levine, Kleiman-Weiner, Chater, Cushman and Tenenbaum2024; Levine et al., Reference Levine, Kleiman-Weiner, Schulz, Tenenbaum and Cushman2020; Trujillo et al., Reference Trujillo, Zhang, Zhi-Xuan, Tenenbaum and Levine2024).

Far from being unfalsifiable, resource-rational contractualism provides an illuminating framework, which is both cognitively realistic and evolutionary coherent, to better understand how the subcomponents of moral cognition are organized and interact with one another to produce moral judgments and behaviors. If we view resource-rational contractualism as focused on how instead of what, there may be no need to resolve the conflict between partiality and impartiality within the theory itself.