CROSSROADS—Promoting Cooperation: Insights and Challenges

We colead the Massachusetts Institute of Technology (MIT) Applied Cooperation Initiative, which collaborates with governments, nonprofits, and companies to apply the lessons from social sciences research on cooperation to real-world challenges such as increasing energy conservation, improving antibiotic adherence, reducing smoking in public places, and promoting philanthropy. Our aim has been to make social sciences research on cooperation more useful and to help people learn about it and use it. Our partners include the U.S. Department of Energy, the Colorado Energy Office, Pacific Gas and Electric Company, the American Bar Association’s Center for Pro Bono, Google, and Facebook as well as an ever-growing list of local nonprofits and scrappy start-ups. The research resulting from these collaborations has been published in academic journals such as Science, Nature, the Proceedings of the Academy of Science, and the New England Journal of Medicine and has been featured in popular press outlets such as The New York Times, The Wall Street Journal, Time magazine, Politico, and Huffington Post.

This piece summarizes how we’ve gone about helping social sciences research on cooperation radiate outward from the ivory tower and into the surrounding world and discusses some lessons from our work as well as challenges for the future.

A Checklist for Promoting Cooperation

Cooperation is defined as an individually costly behavior that benefits others or a group. There is approximately a half century of theoretical research in the biological and social sciences on how cooperation can arise and be sustained (Nowak 2006, McCullough 2020, Henrich and Muthukrishna 2021, Raihani 2021). This research has tremendously advanced our understanding of human cooperation, but to have a real-world impact, one needs to translate the theory into actionable insights and, ideally, ones that can be easily understood by a general audience. Our group has spent the last decade working on this problem.

Allow us to begin by introducing the three categories of interventions on which we typically focus for promoting cooperation: a three-part checklist, if you will (for reviews, see Rand et al. 2014, Hoffman and Yoeli 2022).

1. Increase observability by making the cooperative act more visible to others. For instance, Yoeli et al. (2013) tripled participation in a program that prevents blackouts using sign-up sheets, which made participation in the program observable to people’s neighbors.

2. Eliminating plausible deniability. Plausible deniability gives those who would have otherwise cooperated an opportunity to avoid doing so, which they often take up (e.g., Dana et al. 2006, Dana et al. 2007, Andreoni et al. 2017). Our guidance is, thus, to eliminate such plausible deniability to the extent possible. For example, in Yoeli et al. (2019), we chose a two-way short message service (SMS) intervention, wherein participants are required to respond to an SMS reminder and verify adherence to their medication regime, over a one-way SMS reminder that didn’t require a response. SMS reminders by themselves leave too many plausible excuses, such as “I didn’t notice the text” or “my phone ran out of battery.” (Two-way SMS can also be used to increase observability—a second benefit relative to one-way systems.)

3. Communicating expectations. This category includes some interventions that have received outsize attention in the literature, such as descriptive norms, wherein individuals are told that many others are engaged in the prosocial act (Goldstein and Cialdini 2011), and social comparisons, wherein people’s behavior is compared with that of others as in OPower’s now-famous electricity bill inserts (Allcott 2011). However, there are other effective means of communicating expectations, such as injunctive norms (Schultz et al. 2007), identity frames (Brewer and Kramer 1986), or public good frames (Cookson 2000), and these can sometimes be more practical or appropriate in a given context.

In our experience, this checklist provides a helpful scaffold for designing interventions to promote cooperation. It is consistent with empirical results in the behavioral sciences literature (Kraft-Todd et al. 2015, Rogers et al. 2018) as well as best practices at the highest performing firms (McAfee 2023). For instance, iterative, incremental software development methodologies, such as Agile (Beck et al. 2001) or Elon Musk’s “algorithm,” which, for example, prescribes that an individual’s name be explicitly associated with every nut, bolt, sensor, or screw included on SpaceX’s rockets.

We got to this checklist by combining two disparate literatures. The first was the literature on the evolution of cooperation, which uses simple game theory models to characterize socially coordinated enforcement of desirable behaviors, for example, models of indirect reciprocity or social norms (Panchanathan and Boyd 2004, Nowak 2006). An analysis of cooperative equilibria in these models provides the theoretical backbone for the checklist (Rand et al. 2014, Hoffman and Yoeli 2022). The second was the growing behavioral sciences literature by, for example, behavioral economists and social psychologists, and this documents a cornucopia of quirky aspects of people’s prosocial psychology as well as successful nudges for promoting prosociality. These populate the checklist with examples that make it easier for people to see how it can be applied.

How We Use This Checklist

Over the years, the model that has developed—it’s hard to claim we developed it; it arose organically—is for partners to come to us with a problem requiring real-world cooperation, codevelop an intervention, and test it in a randomized trial. We publish the results, write about them in a popular press outlet, and new partners—learning about our work—reach out to work with us on new projects. We also sometimes also cold call organizations to suggest a collaboration, but it is this cycle of collaboration–publication–press attention that primarily feeds our research pipeline.

Practically speaking, projects typically begin with both sides signing a data sharing agreement and/or a memorandum of understanding (particularly when fundraising but sometimes regardless). We typically handle institutional review board (IRB) review at MIT. Some of our projects qualify for “exempt” status because the research involves minimal risks and activities such as data collection and A/B testing that our partners are anyhow engaged in as part of their regular course of business.

Projects are collaborative in two senses. First, internally, our team meets regularly with the team member who developed a partnership presenting the opportunity to the team, seeking ideas, and sometimes help (e.g., by checking if anyone has an available student or research assistant). Sometimes, this means members of the team coauthor the resulting publication; sometimes, we just consult with one another but publish separately. Second, most projects involve a fairly intensive collaboration with our nonacademic partners, who typically codesign the intervention and the data collection plan and are sometimes also involved in disseminating the results of the research.

Speaking of disseminating results, perhaps more so than the typical academic group, our team has emphasized communicating about research—and, especially, the checklist itself—to the general public, for example, via participation in conferences; by proactively writing op-eds in popular press outlets; or via a TedX talk that, today, has more than 2.5 million views. This helps us generate new partnerships. We also view it as a test of the checklist: if it’s useful, people want to know about it.

One of the challenges our team faces is how to publish our research because it often runs up against publication norms. For instance, in some fields, there is an emphasis on using experiments to identify the mechanism by which an intervention works. Whereas we are occasionally able to do this in a field experiment, it can be difficult because we are often practically constrained to having relatively few experimental treatments and, therefore, fold multiple categories of interventions into a single experimental arm. Or, in some fields, there is an emphasis on novelty, wherein the second paper to use a particular intervention is typically more difficult to publish even if it uses very high-quality methods or applies the intervention to a particularly important problem and population. Yet, if we were to eschew powerful interventions such as observability or descriptive norms just because they’ve previously been written about in an academic journal, this would be doing no favors to our partners or to the general public, which would be less likely to learn about them. A third problem we’ve encountered is that some fields in the social sciences gatekeep or limit review papers, making it harder to publish, for example, a review on how to design an intervention, which can be of particularly high value to practitioners.

There are three solutions we’ve come to (which aren’t mutually exclusive). One is to position our field experiments as demonstration projects, which illustrate the usefulness of using a particular theoretical lens to generate an intervention design even if we can’t always prove why an intervention is working, or which parts of the intervention are crucial. A second is to pair a field experiment with laboratory experiments that get at the mechanism in the field experiment. A third is to publish in general science journals. We’ve found that these are amenable to field experiments or reviews of the kind we wish to publish. These journals also tend to be widely read, which helps generate the necessary interest in our work and perpetuate our cycle of collaborations.

Opportunities for Further Research

We wish to highlight two limitations of our checklist, and these present opportunities for further research in organization science.

First, as readers of this outlet likely have noticed, the focus of this checklist is on motivating individual behavior change outside a particular organization. Certainly, cooperation is critical for the performance of an organization, for example, in driving knowledge acquisition (Edmondson 1999), innovation (Hülsheger et al. 2009), and resilience (Bigley and Roberts 2001). Theoretically, to the extent that the psychology of cooperation and social norms that undergirds our guidance is relevant within the organizational context—and we believe it is—then, the three categories of interventions in our checklist can be effective within an organizational context as well. However, there are reasons to believe that interventions may need to be designed differently to be effective within an organization. One reason is that norms tend to vary in how tightly they are enforced, what is enforced, and how (Gelfand et al. 2011, Hoffman and Yoeli 2022), and such variation in norms can lead to interventions to work in unexpected ways as when Ariely et al. (2009) used an observability manipulation on a college campus to reduce donations to a conservative charity or Rand and Yoeli (2024) used descriptive norms of nonmasking Trump supporters to increase mask-wearing intentions among Biden supporters. A second reason is that monetary incentives, which are the expected means of motivating behavior in many organizations, can lead to perverse effects (Gneezy and Rustichini 2000, 2004). A third is that the social structure of the organization is very different from those employed in most theoretical models of cooperation, and this might mean, for example, that cooperation arises more easily because of mechanisms that are distinct from those on which we’ve focused. For these various reasons, we believe that more work is needed to understand the best ways to promote cooperation within the organization.

A second limitation is that we know relatively more about how to harness norm enforcement to promote cooperation but less about how to establish new norms or fix counterproductive ones. Norms can sometimes lead to behavior that is antisocial as when workers subject to relative incentives slowed down relative to when they were subject to a piece rate so as not to put pressure on one another to work harder (Bandiera et al. 2005). They can also lead to behaviors that are counter to their seeming aims as when particularly environmentally conscious individuals were found to be more likely to engage in “wishcycling” that contaminates the recycling stream (Kramer et al. 2025). We believe this is particularly relevant in the organizational context, in which developing and maintaining norms can mean the difference between a productive and unproductive organization (Sidorenkov and Borokhovski 2023) as starkly illustrated by recent changes in corporate culture that are leading tech firms to dominate incumbents even in industries not traditionally viewed as tech (McAfee 2023). Perhaps because of these recent changes, research that demonstrates effective ways to change norms within an organization is currently of particularly keen interest to nonacademics.

Lessons for Promoting Impact

We also wish to highlight opportunities for structural change that can enhance the impact of research.

First, we believe that some of the value of our research derives from the fact that it translates the findings of others’ research into actionable insights. However, as discussed previously, work that focuses on translation can be difficult to publish and, consequently, to consider during hiring and promotion. This makes it less likely that researchers engage in this kind of work. A solution to these institutional problems involves changes to both publishing and promotion. Journal editors can create a space for articles focusing on translation, which can take multiple forms, such as field experiments and reviews. They can also lift restrictions on review articles when these involve translation. They may need to highlight to reviewers what they are looking for in such articles and advertise to the community the existence of these new article types and/or policies. Academic departments can change hiring and evaluation policies to lend more weight to work published outside the researcher’s field and in general science journals. They can also explicitly recognize impact in hiring and promotion evaluations—a solution that has been adopted in some locales (e.g., England and Australia).

Second, certain institutional practices at our home institution of MIT have substantially greased the wheels when it comes to collaborating with nonacademic entities. Our IRB offers frequent office hours, and this helps us quickly resolve questions related to IRB proposals, automated and immediate evaluation of exempt status for proposals, and very fast evaluation of full proposals. Our legal team maintains templates of data sharing agreements and memoranda of understanding, responds to our requests very quickly, and is happy to hop on the phone with our partners. On the rare occasion when we’ve requested something from our IRB or legal team that has raised a red flag, they have always responded constructively with a proposed alternative. The efficiency and speed at which these bureaucracies operate help to build trust and sustain momentum with our partners and are worthy of emulation by other institutions seeking to enhance their researchers’ policy impact.

Third, our team crosses institutional boundaries with the core members of our team being located at MIT, Arizona State, Stanford, Swarthmore, and soon Cornell. This has the benefit that our team brings in a much more diverse set of partnerships than would be possible if we were working out of a single institution. Although it is possible to arrange cross-institution teams such as ours with little to no funding, we believe that more such partnerships would arise if funders explicitly looked for them when evaluating funding proposals and if institutions offered small internal grants to help spark such collaborations.

Fourth, our work has benefitted immensely from an interdisciplinary perspective, having emerged from a fortuitous marriage of the theoretical literature on the evolution of cooperation emerging from departments such as mathematics, biology, and anthropology with the empirical literature on nudging and prosocial behaviors emerging from social psychology and behavioral economics. Unfortunately, as is well recognized, the academy is poorly structured to motivate interdisciplinary work. We recognize that the fix is difficult, requiring coordinated changes to the way research is funded, published, and evaluated during hiring and promotion and that much ink has already been spilled on the topic. We end by simply adding our voice to the chorus already supporting these changes.