Social Norms and Cultural Evolution

Injunctive Norms Outperform Reciprocity: Tip Request Messages Can Increase Tipping Without Reducing Demand

Tip request messages are marketing communications that can help digital platforms balance customer spending with worker effort and compensation. We report the first field experiments testing how three appeals based on tipping motivations—injunctive norms (“It’s customary to leave a tip for the seller’s service”), personalized reciprocity, and impersonal reciprocity—affect tipping and future demand in a global freelance marketplace. Overall, the norms message increased tipping the most, and did so without reducing returning demand, showing that tip request messages can motivate customer tipping to better incentivize and compensate gig workers. , Tip request messages are marketing communications that can help digital platforms balance customer spending with worker effort and compensation. We report the first academic field experiments manipulating tip request message content, to our knowledge, analyzing 436,969 transactions. We test how three appeals based on tipping motivations—injunctive norms (“It’s customary to leave a tip for the seller’s service”), personalized reciprocity, and impersonal reciprocity—affect tipping and future demand in a global freelance marketplace. On the web, first exposures to the injunctive norms message increased new buyers’ tipping rate by 11.8 percentage points (p.p.) and repeat buyers’ tipping rate by 6.8 p.p. In the app, first exposures to the injunctive norms message increased tipping rate by 2.5 p.p. for new buyers and 1.4 p.p. for repeat buyers. Treatment effects on tipping increase with transaction order price. The norms and personalized reciprocity messages increased subsequent orders and spending in the web experiment and did not significantly change repatronage in the app experiment, suggesting that effective tipping prompts need not reduce future business. Overall, tip request messages can motivate customer tipping to better incentivize and compensate gig workers, but optimization requires careful testing. History: Hema Yoganarasimhan served as the senior editor. Supplemental Material: The online appendix is available at https://doi.org/10.1287/mksc.2025.0392 .

Everyone Conforms, No One Believes: Pluralistic Ignorance in LLM Agent Populations

LLM-based multi-agent systems are increasingly used to simulate social dynamics, from opinion formation to collective decision-making. These simulations can reproduce certain social phenomena, but it is unknown whether they capture pluralistic ignorance, a state where a majority privately rejects a norm yet publicly conforms, each believing they are alone in dissenting. This phenomenon drives norm persistence, social movements, and political revolutions. We show that pluralistic ignorance emerges robustly in LLM agent populations. We construct a benchmark of 100 scenarios across 10 domains and 5 authority levels, grounded in the human pluralistic ignorance literature, and evaluate 8 models from 6 organizations. Agents publicly conform at rates of 64 to 94% despite privately opposing the norm. Conformity is domain-sensitive (workplace and social relationship scenarios produce near-universal public compliance) and highly model-dependent, though uncorrelated with model capability. We further test whether a single "norm entrepreneur" can break the false consensus by publicly dissenting. For 7 of 8 models, cascades succeed less than 26% of the time, with one model showing zero cascades across all scenarios. GPT-4o is a notable outlier at 48%, revealing qualitatively distinct social dynamics across model families. Our findings identify model selection as an unacknowledged degree of freedom that fundamentally shapes simulation outcomes. More broadly, the near-absence of cascades suggests LLM simulations may systematically overestimate the stability of social norms, missing the fragile tipping-point dynamics that drive real-world norm change in human societies.

AI, Human Cognition and Knowledge Collapse

We study how generative AI, and in particular agentic AI, shapes human learning incentives and the long-run evolution of society’s information ecosystem. We build a dynamic model of learning and decision-making in which successful decisions require combining shared, community-level general knowledge with individual-level, context-specific knowledge; these two inputs are complements. Learning exhibits economies of scope: costly human effort jointly produces a private signal about their own context and a “thin” public signal that accumulates into the community’s stock of general knowledge, generating a learning externality. Agentic AI delivers context-specific recommendations that substitute for human effort. By contrast, a richer stock of general knowledge complements human effort by raising its marginal return. The model highlights a sharp dynamic tension: while agentic AI can improve contemporaneous decision quality, it can also erode learning incentives that sustain long-run collective knowledge. When human effort is sufficiently elastic and agentic recommendations exceed an accuracy threshold, the economy can tip into a knowledge-collapse steady state in which general knowledge vanishes ultimately, despite high-quality personalized advice. Welfare is generally non-monotone in agentic accuracy, implying an interior, welfare-maximizing level of agentic precision and motivating information-design regulations. In contrast, greater aggregation capacity for general knowledge—meaning more effective sharing and pooling of human-generated general knowledge—unambiguously raises welfare and increases resilience to knowledge collapse.

Bayesian Inference of Social Norms as Shared Constraints on Behavior

People act upon their desires, but often, also act in adherence to implicit social norms. How do people infer these unstated social norms from others' behavior, especially in novel social contexts? We propose that laypeople have intuitive theories of social norms as behavioral constraints shared across different agents in the same social context. We formalize inference of norms using a Bayesian Theory of Mind approach, and show that this computational approach provides excellent predictions of how people infer norms in two scenarios. Our results suggest that people separate the influence of norms and individual desires on others' actions, and have implications for modelling generalizations of hidden causes of behavior.

What people learn from punishment: A cognitive model

Authorities, from parents of toddlers to leaders of formal institutions, use punishment to communicate disapproval and enforce social norms. Ideally, from whether and how severely a transgression is punished, targets and observers infer shared social norms. Yet in light of every punitive choice, observers also evaluate the motives and legitimacy of the authority. Here, we show that the effects of punishment can only be understood by considering these inferences simultaneously. We measured human observers’ joint inferences empirically in three preregistered experiments ( N = 1 , 254 ) and developed a rational Bayesian model using an inverse planning framework that captures and explains these inferences and their interactions quantitatively and parsimoniously. When people have different priors about norms or authorities, the model predicted and we experimentally confirmed that observing punishment by the authority can sustain polarization. This work reveals the rational logic behind how people learn from punishment and a key constraint on the function of punishment in establishing shared social norms.

Spurious normativity enhances learning of compliance and enforcement behavior in artificial agents

Significance The fact that humans enforce and comply with norms is an important reason why humans enjoy higher levels of cooperation and welfare than other animals. Some norms are relatively easy to explain: They may prohibit obviously harmful or uncooperative actions. But many norms are not easy to explain. For example, most cultures prohibit eating certain kinds of foods, and almost all societies have rules about what constitutes appropriate clothing, language, and gestures. Using a computational model focused on learning shows that apparently pointless rules can have an indirect effect on welfare. They can help agents learn how to enforce and comply with norms in general, improving the group’s ability to enforce norms that have a direct effect on welfare. , How do societies learn and maintain social norms? Here we use multiagent reinforcement learning to investigate the learning dynamics of enforcement and compliance behaviors. Artificial agents populate a foraging environment and need to learn to avoid a poisonous berry. Agents learn to avoid eating poisonous berries better when doing so is taboo, meaning the behavior is punished by other agents. The taboo helps overcome a credit assignment problem in discovering delayed health effects. Critically, introducing an additional taboo, which results in punishment for eating a harmless berry, further improves overall returns. This “silly rule” counterintuitively has a positive effect because it gives agents more practice in learning rule enforcement. By probing what individual agents have learned, we demonstrate that normative behavior relies on a sequence of learned skills. Learning rule compliance builds upon prior learning of rule enforcement by other agents. Our results highlight the benefit of employing a multiagent reinforcement learning computational model focused on learning to implement complex actions.

Deference, Dissent, and Dispute Resolution: An Experimental Intervention Using Mass Media to Change Norms and Behavior in Rwanda

Deference and dissent strike a delicate balance in any polity. Insufficient deference to authority may incapacitate government, whereas too much may allow leaders to orchestrate mass violence. Although cross-national and cross-temporal variation in deference to authority and willingness to express dissent has long been studied in political science, rarely have scholars studied programs designed to change these aspects of political culture. This study, situated in post-genocide Rwanda, reports a qualitative and quantitative assessment of one such attempt, a radio program aimed at discouraging blind obedience and reliance on direction from authorities and promoting independent thought and collective action in problem solving. Over the course of one year, this radio program or a comparable program dealing with HIV was randomly presented to pairs of communities, including communities of genocide survivors, Twa people, and imprisoned génocidaires. Changes in individual attitudes, perceived community norms, and deliberative behaviors were assessed using closed-ended interviews, focus group discussions, role-play exercises, and unobtrusive measures of collective decision making. Although the radio program had little effect on many kinds of beliefs and attitudes, it had a substantial impact on listeners' willingness to express dissent and the ways they resolved communal problems.

How Does Media Influence Social Norms? Experimental Evidence on the Role of Common Knowledge

How does media influence beliefs, attitudes, and behaviors? While many scholars have studied the effect of media on social and political outcomes, we know surprisingly little about the channels through which this effect operates. I argue that two mechanisms can account for its impact. Media provides new information that persuades individuals to accept it (individual channel), but also, media informs listeners about what others learn, thus facilitating coordination (social channel). Combining a field experiment with a plausibly natural experiment in Mexico, I disentangle these effects analyzing norms surrounding violence against women. I examine the effect of a radio program when it is transmitted privately versus when it is transmitted publicly. I find no evidence supporting the individual mechanism. The social channel, however, increased rejection of violence against women and increased support for gender equality, but unexpectedly increased pessimism about whether violence would decline in the future.

Propaganda and Conflict: Evidence from the Rwandan Genocide *

Abstract This article investigates the role of mass media in times of conflict and state-sponsored mass violence against civilians. We use a unique village-level data set from the Rwandan genocide to estimate the impact of a popular radio station that encouraged violence against the Tutsi minority population. The results show that the broadcasts had a significant effect on participation in killings by both militia groups and ordinary civilians. An estimated 51,000 perpetrators, or approximately 10% of the overall violence, can be attributed to the station. The broadcasts increased militia violence not only directly by influencing behavior in villages with radio reception but also indirectly by increasing participation in neighboring villages. In fact, spillovers are estimated to have caused more militia violence than the direct effects. Thus, the article provides evidence that mass media can affect participation in violence directly due to exposure and indirectly due to social interactions.

How Radio Affects Violent Conflict: New Evidence from Rwanda

Researchers have long debated how radio broadcasts affected the dynamics of the 1994 genocide in Rwanda, with some arguing that the radio was highly consequential, and others suggesting such effects have been overstated. This article contributes to these debates—as well as to debates regarding the role of old and new media in collective action—by examining whether and how Radio Télévision Libre des Mille Collines (Radio RTLM) coverage was associated with two core aspects of the violence: (1) subnational onset of genocidal violence and (2) participation in genocidal violence across subnational spaces. Drawing on new data on Radio RTLM coverage, we find that areas with coverage were more likely to experience immediate onset of violence. However, our analysis of participation in the genocide—which uses more accurate measures of participation and of radio coverage than prior studies—finds no significant association between Radio RTLM coverage and subnational levels of participation. After illustrating that these results are robust to numerous model specifications, we theorize that information broadcast over the radio’s airways contributed to the creation of a critical mass that initiated genocide in localized spaces. We conclude by considering the importance of understanding the role of media in the subnational onset of violence.

The critical role of second-order normative beliefs in predicting energy conservation

Sustaining large-scale public goods requires individuals to make environmentally friendly decisions today to benefit future generations1–6. Recent research suggests that second-order normative beliefs are more powerful predictors of behaviour than first-order personal beliefs7,8. We explored the role that second-order normative beliefs—the belief that community members think that saving energy helps the environment—play in curbing energy use. We first analysed a data set of 211 independent, randomized controlled trials conducted in 27 US states by Opower, a company that uses comparative information about energy consumption to reduce household energy usage (pooled N = 16,198,595). Building off the finding that the energy savings varied between 0.81% and 2.55% across states, we matched this energy use data with a survey that we conducted of over 2,000 individuals in those same states on their first-order personal and second-order normative beliefs. We found that second-order normative beliefs predicted energy savings but first-order personal beliefs did not. A subsequent pre-registered experiment provides causal evidence for the role of second-order normative beliefs in predicting energy conservation above first-order personal beliefs. Our results suggest that second-order normative beliefs play a critical role in promoting energy conservation and have important implications for policymakers concerned with curbing the detrimental consequences of climate change.

Network dynamics of social influence in the wisdom of crowds

Significance Since the discovery of the wisdom of crowds over 100 years ago theories of collective intelligence have held that group accuracy requires either statistical independence or informational diversity among individual beliefs. Empirical evidence suggests that allowing people to observe the beliefs of others leads to increased similarity of individual estimates, reducing independence and diversity without a corresponding increase in group accuracy. As a result, social influence is expected to undermine the wisdom of crowds. We present theoretical predictions and experimental findings demonstrating that, in decentralized networks, social influence generates learning dynamics that reliably improve the wisdom of crowds. We identify general conditions under which influence, not independence, produces the most accurate group judgments. , A longstanding problem in the social, biological, and computational sciences is to determine how groups of distributed individuals can form intelligent collective judgments. Since Galton’s discovery of the “wisdom of crowds” [Galton F (1907) Nature 75:450–451], theories of collective intelligence have suggested that the accuracy of group judgments requires individuals to be either independent, with uncorrelated beliefs, or diverse, with negatively correlated beliefs [Page S (2008) The Difference: How the Power of Diversity Creates Better Groups, Firms, Schools, and Societies ]. Previous experimental studies have supported this view by arguing that social influence undermines the wisdom of crowds. These results showed that individuals’ estimates became more similar when subjects observed each other’s beliefs, thereby reducing diversity without a corresponding increase in group accuracy [Lorenz J, Rauhut H, Schweitzer F, Helbing D (2011) Proc Natl Acad Sci USA 108:9020–9025]. By contrast, we show general network conditions under which social influence improves the accuracy of group estimates, even as individual beliefs become more similar. We present theoretical predictions and experimental results showing that, in decentralized communication networks, group estimates become reliably more accurate as a result of information exchange. We further show that the dynamics of group accuracy change with network structure. In centralized networks, where the influence of central individuals dominates the collective estimation process, group estimates become more likely to increase in error.

On the Alignment of Large Language Models with Global Human Opinion

Today's large language models (LLMs) are capable of supporting multilingual scenarios, allowing users to interact with LLMs in their native languages. When LLMs respond to subjective questions posed by users, they are expected to align with the views of specific demographic groups or historical periods, shaped by the language in which the user interacts with the model. Existing studies mainly focus on researching the opinions represented by LLMs among demographic groups in the United States or a few countries, lacking worldwide country samples and studies on human opinions in different historical periods, as well as lacking discussion on using language to steer LLMs. Moreover, they also overlook the potential influence of prompt language on the alignment of LLMs' opinions. In this study, our goal is to fill these gaps. To this end, we create an evaluation framework based on the World Values Survey (WVS) to systematically assess the alignment of LLMs with human opinions across different countries, languages, and historical periods around the world. We find that LLMs appropriately or over-align the opinions with only a few countries while under-aligning the opinions with most countries. Furthermore, changing the language of the prompt to match the language used in the questionnaire can effectively steer LLMs to align with the opinions of the corresponding country more effectively than existing steering methods. At the same time, LLMs are more aligned with the opinions of the contemporary population. To our knowledge, our study is the first comprehensive investigation of the topic of opinion alignment in LLMs across global, language, and temporal dimensions.

How do social media feed algorithms affect attitudes and behavior in an election campaign?

We investigated the effects of Facebook’s and Instagram’s feed algorithms during the 2020 US election. We assigned a sample of consenting users to reverse-chronologically-ordered feeds instead of the default algorithms. Moving users out of algorithmic feeds substantially decreased the time they spent on the platforms and their activity. The chronological feed also affected exposure to content: The amount of political and untrustworthy content they saw increased on both platforms, the amount of content classified as uncivil or containing slur words they saw decreased on Facebook, and the amount of content from moderate friends and sources with ideologically mixed audiences they saw increased on Facebook. Despite these substantial changes in users’ on-platform experience, the chronological feed did not significantly alter levels of issue polarization, affective polarization, political knowledge, or other key attitudes during the 3-month study period.

Overperception of moral outrage in online social networks inflates beliefs about intergroup hostility

As individuals and political leaders increasingly interact in online social networks, it is important to understand the dynamics of emotion perception online. Here, we propose that social media users overperceive levels of moral outrage felt by individuals and groups, inflating beliefs about intergroup hostility. Using a Twitter field survey, we measured authors’ moral outrage in real time and compared authors’ reports to observers’ judgements of the authors’ moral outrage. We find that observers systematically overperceive moral outrage in authors, inferring more intense moral outrage experiences from messages than the authors of those messages actually reported. This effect was stronger in participants who spent more time on social media to learn about politics. Preregistered confirmatory behavioural experiments found that overperception of individuals’ moral outrage causes overperception of collective moral outrage and inflates beliefs about hostile communication norms, group affective polarization and ideological extremity. Together, these results highlight how individual-level overperceptions of online moral outrage produce collective overperceptions that have the potential to warp our social knowledge of moral and political attitudes.

Patterns and problems in multiagent systems

We ran experiments on swarms of Claude agents and found coordination failures, collusion, and sabotage. Here, we share what they mean for AI safety.

A Case Study on Emergent Cheating and Whistleblowing in Autonomous Research Swarms

Multi-agent AI science ecosystems rely on agents possessing tools that allow them to communicate, coordinate, and build on each other's work. Yet this shared infrastructure can also introduce vulnerabilities by creating a substrate for the contagious spread of unintended and undesirable behaviors. We report a case study on a research collective of 100 autonomous LLM agents tasked with proving formal mathematical conjectures. Within the swarm, cheating spontaneously emerged and was later challenged by whistleblowers - both without any external intervention. When a single agent discovered an exploit in the evaluation system, it propagated across the collective via a shared knowledge library and later through peer-to-peer messages. Despite early reluctance, a cohort of agents adopted the exploit in response to competitive pressure. A separate group of agents produced an emergent counter-response: auditing fraudulent proofs, alerting peers across broadcast and private channels, staging boycotts, lodging formal complaints, and proposing validation patches. In recent incidents, agent swarms coordinated covertly through improvised side-channels (Dalton and Wallace, 2026; Greenblatt et al., 2026). Our setting differs: the same transparent channels that carried the exploit also gave non-cheating agents the visibility they needed to detect fraud, organize resistance, and enforce norms. We cast the problem of managing the agents' shared infrastructure as the knowledge commons governance problem (Ostrom, 1990). To protect the commons from exploits, we propose to adopt institutional mechanisms, such as graduated sanctioning and collective-choice rules, to support decentralized self-governance in autonomous swarms.

The limits of social media as a source of political information during routine and crisis times across 17 countries

Recent studies on political knowledge suggest people learn little about political events and societal issues from social media. Potentially, social media are a more effective source of information ...

Rational Ritual

Why do Internet, financial service, and beer commercials dominate Super Bowl advertising? How do political ceremonies establish authority? Why does repetition characterize anthems and ritual speech? Why were circular forms favored for public festivals during the French Revolution? This book answers these questions using a single concept: common knowledge. Game theory shows that in order to coordinate its actions, a group of people must form “common knowledge.” Each person wants to participate only if others also participate. Members must have knowledge of each other, knowledge of that knowledge, knowledge of the knowledge of that knowledge, and so on. Michael Chwe applies this insight, with striking erudition, to analyze a range of rituals across history and cultures. He shows that public ceremonies are powerful not simply because they transmit meaning from a central source to each audience member but because they let audience members know what other members know. For instance, people watching the Super Bowl know that many others are seeing precisely what they see and that those people know in turn that many others are also watching. This creates common knowledge, and advertisers selling products that depend on consensus are willing to pay large sums to gain access to it. Remarkably, a great variety of rituals and ceremonies, such as formal inaugurations, work in much the same way. By using a rational-choice argument to explain diverse cultural practices, Chwe argues for a close reciprocal relationship between the perspectives of rationality and culture. He illustrates how game theory can be applied to an unexpectedly broad spectrum of problems, while showing in an admirably clear way what game theory might hold for scholars in the social sciences and humanities who are not yet acquainted with it. In a new afterword, Chwe delves into new applications of common knowledge, both in the real world and in experiments, and considers how generating common knowledge has become easier in the digital age.