Levine et al. argue that moral cognition is underpinned by a collection of computational mechanisms that approximate contractualist agreements between people. We support the core proposal, but we see an opportunity to strengthen its claims by addressing the role of social learning and cultural evolution in the construction of moral norms and their impact on moral reasoning.

What is the relationship between the moral norms we inherit culturally and the moral judgements we make in the moment as individual reasoners? The target article draws a clear distinction, viewing moral norms as frozen artifacts, and the cultural evolutionary processes that create them as blind selection: “Blind evolutionary processes can be powerful, but they take a long time and their results—the policies, structures or mechanisms they construct—are rigid.” Under this construal, culturally-evolved moral norms are separate from the cognitive processes described in the target article and therefore out of scope.

Is such a categorical distinction necessary? Knowledge generated over the course of cumulative cultural evolution can be abstract and flexible (Thompson et al., Reference Thompson, van Opheusden, Sumers and Griffiths2022). Consider the classic example of Micronesian navigators, who inherit sophisticated culturally-evolved strategies for approximating distances over open ocean (Hutchins & Hinton, Reference Hutchins and Hinton1984). It would be difficult to describe such complex forms of knowledge as rigid policies separate from reasoning. Instead, what is being acquired through social learning is a generalizable algorithmic concept that can be deployed in a variety of contexts.

In this example and in many other domains, the knowledge acquired as a consequence of social learning is perhaps better understood as a cultural program (D’Andrade, Reference D’Andrade1981): a structured yet flexible representation that can be adaptively revised, repurposed, and extended via the kinds of in-the-moment reasoning described in the target article. Are moral norms any different? Perhaps moral norms can be viewed as a specific type of cultural program: a joint program that prescribes procedural action not just for one person, but for a group.

One benefit of this viewpoint is that it accommodates the potential for structure, abstraction, and internal complexity in culturally inherited norms. A simple example is illustrative. Suppose you join a research lab that often runs low on beverages at lab meeting, but upholds a longstanding normative solution: “If there are fewer La Croix cans available than there are people who want one, then a subset of us will get one La Croix this time, and those who missed out get first pick the next time”. The norm expressed in this way makes use of conditionals, comparison operators (number of people > number of cans), and state tracking (whether or not it is currently my turn to get a La Croix). Recent work has begun to formalize this intuition and sought to model joint action using program-like representations (Yu & Thompson, Reference Yu and Thompson2024; Shum et al., Reference Shum, Kleiman-Weiner, Littman and Tenenbaum2019). This view foregrounds the flexible cognitive representations that can be constructed as a consequence of social learning.

A more cognitive view of cultural evolution and the knowledge it generates can also help to explain how people integrate socially-learned norms with in-the-moment reasoning processes. To see this, consider the norm of waiting in line - a culturally evolved, socially transmitted solution to the problem of how to manage space, time and linear order when people are sharing a resource (Kraftt & Griffiths, Reference Krafft and Griffiths2018). In the case of an unforeseen circumstance like a medical emergency, the normal principles of queuing must be revised, but need not be abandoned entirely. Reasoning over program-like norms can achieve effective program revision - the person suffering from the emergency can jump the line. However, the base “program” of waiting in line stays mostly intact - all the other people in line obey the generic principles. If there were multiple people with an emergency, the line might be converted to a priority queue at the beginning, where the most dire cases are at the front, while the rest of the line again obeys the generic principles.

In these cases, there is a meaningful interface between culturally-inherited norms and in-the-moment reasoning. Viewing norms as cultural programs offers a way to conceptualize how people arbitrate between community-level normative commitments and the demands of unfamiliar situations. While much more research is needed to understand the computational principles of this interface in detail, one clear possibility raised by such a “common currency” is that norms are integrated as part of the model in model-based planning; virtual bargaining over revisions to a shared “base program” would be one such instantiation.

More generally, a clearer connection between culturally inherited moral norms and in-the-moment reasoning also offers opportunities to rethink the role that population-level processes may play in generating resource-rational solutions to the computational problems that people face in moral judgement. The target article recognizes a rich literature on selection-like processes at the population level, and reasonably views these processes as out of scope for a primarily-psychological theory. However, cultural evolutionary theory distinguishes selection-like processes (that adjust the frequency of existing variants) from transformation-like processes (that modify the structure of existing variants); both selection and transformation are thought to have a substantial impact on the products of cultural evolution (Mesoudi, Reference Mesoudi2021).

This holistic view of cultural evolution offers a more direct link between resource-rational reasoning and the structure of moral norms: the psychological processes proposed in the target article play a transformation-like role in cultural evolution, leaving fingerprints of resource-rational computation in the norms we inherit. Such effects are particularly likely to arise in interactive forms of social learning that expose the traces of moral reasoning directly to participants and observers, supporting (cultural) program-induction (Kleiman-Weiner et al., Reference Kleiman-Weiner, Sosa, Thompson, Opheusden van, Griffiths, Gershman and Cushman2020; Piantadosi, Rule & Tenenbaum, Reference Piantadosi, Rule, Tenenbaum, Griffiths, Chater and Tenenbaum2024) in way that is analogous to the benefits of learning from process rather than product during cultural transmission of material artefacts (Caldwell & Millen, Reference Caldwell and Millen2009; Derex et al., Reference Derex, Beugin, Godelle and Raymond2013).

In summary, we see substantial value in the proposals put forward in the target article, and we anticipate the potential for an even more complete theory through greater integration of population-level phenomena. In particular, we have advocated for greater continuity between moral reasoning and culturally-inherited moral norms. Viewing norms as joint programs generated over the course of cultural evolution offers one way to recognize this continuity and integrate population-level processes into our developing understanding of resource-rational cognition.