Levine et al. propose that finding mutual benefit among individuals faced with interdependent choice is the core computational problem that moral action and judgment seek to solve, with other well-studied parts of moral cognition (i.e., deontological and utilitarian reasoning) applying as resource-rational approximations to this ideal. If true, this has implications for how policymakers, governments, and firms—choice architects, more generally—could design social systems to facilitate implicit, explicit, and simulated agreement-finding given cognitive constraints. In this commentary, we explore how systems could be designed to help effectuate resource-rational agreements at the different levels of abstraction and anticipation proposed by Levine et al. in ways that better support individuals’ ability to converge on mutual benefit.

While Levine et al.’s contractualist framework is a descriptive account of morality, there are reasons to believe that aligning social systems with its principles may foster positive individual and societal outcomes, even without assuming it as purely normative. A growing body of research finds that belief in a fair and just society is positively associated with improved psychological well-being (Lima, de Souza, & Jost, Reference Lima, de Souza and Jost2025), greater interpersonal trust (You, Reference You2012), and even national-level economic growth (Sztaudynger, Ambroziak, & Starosta, Reference Sztaudynger, Ambroziak and Starosta2022). On a smaller scale, research shows that companies can benefit from gaining “social license” for their operations – broad community support and acceptance beyond merely complying with the law (Gunningham, Kagan, & Thornton, Reference Gunningham, Kagan and Thornton2004; White et al., Reference White, Dennis, Tomko, Bell and Winter2021).

To ground these ideas, we begin with the most concrete level of agreement-finding in Levine et al.’s taxonomy: “actual agreements” over “specific instances.” One familiar domain where such agreements are made and considered is contract law, where resource-rational contractualist principles have long been relevant. A famous example is Lord Denning’s “red hand rule,” which states that the more onerous a contract clause is, the greater the effort must be to ensure the other party is aware of it. As he indelibly put it, “Some clauses which I have seen would need to be printed in red ink on the face of the document with a red hand pointing to it before the notice could be held to be sufficient” (J Spurling Ltd v Bradshaw [1956] EWCA Civ 3, 1956). A similar notion, though more constrained, is reflected in the U.S. Uniform Commercial Code § 2–316(2).

While the justification of such laws and principles is seldom made by explicit reference to resource rationality, this perspective fits well with their function. Legally, the rationale often rests on the doctrine of “the meeting of the minds”—the idea that both parties must have a mutual understanding and intent to be bound by the terms of a contract. If a key clause is hidden or obscured, it casts doubt on whether the other party showed their intention to be bound by it. Viewed through a resource-rational lens, however, such rules not only ensure the objective manifestation of the meeting of the minds, but they also reduce the cognitive effort required to recognize and evaluate key terms, thereby enabling agents to reach agreements that genuinely achieve mutual benefit. In this way, such rules help realize the core computational goal of contractualist morality: supporting mutually advantageous coordination under resource constraints.

Reframing legal principles like the red hand rule through a resource-rational lens highlights how public systems more broadly can be designed to reduce cognitive burdens and facilitate mutual benefit under bounded rationality (Schulz & Bhui, Reference Schulz and Bhui2024). For example, consumer contract laws could be strengthened to require that key terms be presented clearly, upfront, and in plain language (see, e.g., Pennsylvania’s Plain Language Consumer Contract Act). Martínez, Mollica, and Gibson (Reference Martínez, Mollica and Gibson2022) show that it is not legal concepts themselves that make such texts difficult to process, but the syntactic complexity and working memory demands of “legalese”. Thus, mandating or incentivizing plain language in consumer contracts could free individuals to better evaluate terms and reach agreements that genuinely reflect mutual benefit.

However, we encourage choice architects to consider resource rationality beyond satisfying minimum legal requirements, but toward truly facilitating the process of real agreement-finding. For example, companies could design end-user license agreements in line with principles of resource rationality to allow users to truly assent to their terms rather than merely accepting them unread (which almost everybody does; Bakos, Marotta-Wurgler, & Trossen, Reference Bakos, Marotta-Wurgler and Trossen2014). Similarly, governments and political organizations could leverage Large-Language-Model- mediated conversations as a means of simplifying and clarifying complex policy issues, allowing individuals to find broad agreement on them (see, e.g., Tessler et al., Reference Tessler, Bakker, Jarrett, Sheahan, Chadwick, Koster, Evans, Campbell-Gillingham, Collins, Parkes, Botvinick and Summerfield2024).

Choice architects could also support agreement-finding at the greater levels of abstraction outlined in Levine et al.’s framework. At the virtual bargaining level—simulated bargaining “in the head” over “specific instances”—policy can aid by clarifying others’ preferences. For example, Shaddy and Shah (Reference Shaddy and Shah2018) show that effort-based resources such as time and labor more clearly signal people’s preferences than money or social influence. An employer could harness this by implementing an effort-credit system: employees who take on less desirable tasks earn credits that they can redeem for high-demand opportunities. Structuring allocation this way (rather than, perhaps, a seniority- or lottery-based system) makes preferences more transparent to others, enabling more accurate simulated agreements as colleagues can better infer who truly values what. At the model-free level, institutions can scaffold moral behavior by reinforcing cached action standards that approximate mutually beneficial norms. For instance, when people are immersed in environments characterized by cooperative equilibria, the intuitive propensity to cooperate spills over into new environments where cooperation would be mutually beneficial but is not an equilibrium (Peysakhovich & Rand, Reference Peysakhovich and Rand2016).

By designing policies that support agreement-finding across the taxonomy outlined by Levine et al.—from explicit contracts to simulated bargaining—choice architects can support individuals in achieving mutual benefit despite cognitive constraints. This calls for a shift in policy from outcome facilitation, which aims to steer people toward predefined endpoints, to process facilitation, which focuses on structuring environments to support individuals’ own capacity to reason and coordinate (McKenzie et al., Reference McKenzie, Sher, Leong and Müller-Trede2018). Rather than nudging people toward predefined moral choices, the goal is to scaffold the reasoning processes by which genuine mutual benefit becomes possible.