Proof-of-Concept Computational Policy Model for TACT

Researchers in CMAC’s Tools against Child Sex Trafficking Project (TACT) are building computational policy models to virtually evaluate and experiment with anti-human trafficking policy options. This project also aims to demonstrate the effectiveness of computational modeling and simulation in analyzing policy proposals. Output data can be used as a measure to determine whether or not a policy is working well.

Using input data from King County, WA, CMAC Postdoctoral Fellow Khatera Alizada built an agent-based simulation model to explore the impact of two law-enforcement strategies a) arresting and prosecuting minor victims of sex trafficking for prostitution charges, and b) arresting and prosecuting buyers who purchase sex from minors. Alizada sought to learn how each tactic effects the rate of child sex trafficking in King County, WA.


The model simulates these two law-enforcement strategies in a virtual society and generates outputs in the form of the number of children trapped in commercial sex trafficking and the number of commercial sex transactions with children. The results show that targeting demand (arresting and prosecuting buyers) reduces child sex trafficking in King County, WA more effectively than targeting supply.

King County, Washington State

Switching law-enforcement focus from supply to demand reduces the number of sex trafficked children by more than 50 percent and the commercial sex purchase by about 1.5 percent per year.

The following graph depicts the percentage difference between purchases in 2009 and purchases in 2015 over the one-year period of the simulation, with 52 one-week time-steps. A positive percentage number indicates that the 2009 number is higher by that percentage than the 2015 number. At the beginning of the year there is little difference, but the difference grows, producing an end-of-year situation in which there are 1.5% fewer purchases in 2015 than in 2009, with a 3.5% decrease in the number of buyers in 2015. Since the simulation runs reflected the policy change from supply-side arrests to demand-side arrests, and since these models are causal and not merely correlational (as statistical models are), researchers know that the change occurs because of the policy shift.

Figure 1. Percentage Difference in Purchases 2009 Vs 2015

Researchers can also use the number of sex trafficked children over the course of one year to evaluate the policy change. The orange line displays the percentage decrease in the count of sex trafficked children in 2015 (with a demand-side focused policy). The positive percentage indicates that in 2009 the number of trafficked children is higher than in 2015. In 2015, the number of children trapped in sex trafficking reduces by more than half.

Figure 2. Percentage Decrease in the Count of Children Trapped in Sex Trafficking 2009 Vs 2015

The results therefore agree that targeting buyers and protecting victims has a more significant impact on overall reduction of commercial sexual exploitation of minors in King County, WA, matching the results of their policy experiment.


In October 2014, King County launched a policy program, Buyer Beware, designed to reduce the demand for purchase of commercial sex online. Its suggested policies shifted the prosecutorial focus from supply side (arresting prostituted people) to demand side (targeting buyers of commercial sex). The Buyer Beware program is committed to running reverse-sting operations on a monthly or bi-monthly basis. A reverse sting operation is when a police officer poses as a prostitute to arrest a buyer.

In 2009, before adopting the Buyer Beware program policies, 53 prostituted minors were charged for prostitution and 2 buyers were charged for buying sex from children. In 2015, 46 buyers were charged for buying sex from children and 0 children were charged for prostitution (Richey 2018). Historically, over 90 percent of those arrested for prostitution in the United States are sellers and fewer than 10 percent are buyers. When comparing the effectiveness of supply-focused versus demand-focused intervention strategies, the empirical evidence suggests that demand-focused law enforcement is more effective in reducing commercial sexual exploitation. For example, in Sweden after prohibiting purchase of sex, prostitution reduced more than 50 percent. There is very little evidence that targeting supply yields more than a temporary suppression or displacement of prostitution.

Modeling Religion Project Research

When most people think of the study of religion, they think of ancient languages, sacred texts, congregational surveys, or anthropologists setting up camp in a tribal society. What if we could use computer simulations to recreate historical events, learn about religious violence and terrorism, and understand religion’s role in transitions from pre-agricultural to industrial societies?

Computational social scientist Justin Lane thinks computer modeling can help us understand complex social phenomena like religion. Once we unlock this powerful method, he says, “We can no longer be content with rudimentary answers to the hard questions that religion poses.”


The Modeling Religion Project (MRP) is an ambitious attempt to connect the sciences of modeling and simulation with the scientific study of religion. It has unfolded over the last three years, assembling a group of computer simulation experts and religion scholars to collaborate on computer simulations that reveal insights into how religion affects people and societies.

Together, team members studied the cultural, cognitive and ritual processes that undergird religion and culture. They produced models of terror management and mutually escalating religious violence, and made progress on models exploring social identity, ritual competence and credibility enhancing displays.

Team Member Carlos Lemos

This project demanded that religion scholars and computer specialists work far outside their comfort zone and rise above disciplinary constraints in order to understand each other. The religion scholars worked to clearly describe the theories in their field to computer scientists, and they pushed themselves to assemble precise computational expressions of these theories.

This kind of radical interdisciplinarity “forces all involved to spend inordinate amounts of time digging into domain specific literature, reworking and retesting code and explaining concepts and ideas that have always been accepted as fact,” said computer scientist and team member Dr. Ross Gore. “Frequently, a model smaller in scale and more limited than originally intended is produced but it’s amazing how much insight can be gained from such a tiny, limited model.”

Team Members Ross Gore (L) & LeRon Shults (R)

Another team member, philosopher and theologian Dr. LeRon Shults, was so inspired by the complexity and clarity harnessed by computer simulation that he joked, “this is the methodology from which I was separated at birth!”

Working together, these experts developed a fruitful balance between the simplicity needed to understand the models and the complexity needed for the models to accurately reflect human behavior and religious processes. The team built virtual computer agents with complex cognitions, capable of testing many theories on beliefs and experiences. What they forged was the first comprehensive effort to systematically model the predictions of many different cognitive and social theories of religion in various simulated settings.


A brief overview of some of the models produced will illustrate the rich, broad range of this project. One model, MERV, was an agent-based model exploring xenophobic anxiety in an artificial society of two different religious groups where agents experience hazards. The model was developed based on several empirically validated theories about the role of religion in intergroup conflict. The model identifies some of the conditions and mechanisms that engender mutually escalating xenophobic anxiety between religious groups.

Another model, the Shamanism and Possession Model, is an agent-based model that simulates the social and psychological dynamics that influence whether men or women will fulfill a shamanic or trance-healer role in a society. Our results showed that in different types of societies—depending on how hierarchical the society is—women and men are differently afflicted by psychosocial trauma, which leads to gender variations in who is more susceptible to the dissociative trances of the shaman healer.

In the model we call Forest, we ask, why aren’t there more atheists? In this systems dynamics model we identify the conditions under which a population with widespread supernatural beliefs can change into one in which most individuals reject supernatural explanations. 

An Example of a Model Generated

While the conditions for producing widespread rejection of supernatural worldviews are highly specific, historically rare and difficult to produce, when those conditions combine, there emerges a stable social equilibrium that makes atheism widespread. However, this equilibrium is easier to disrupt than those whose cohesion is stabilized by supernatural religion.


One major goal of the MRP was to produce a web-based simulation platform that will allow scholars and students to create complex simulations with no knowledge of programming. Currently, the simulation platform can be used to test a range of questions about social behaviors, and it is still being refined to accommodate more specific questions about religion.

When it is complete, it will allow modelers to specify the cognitive, emotional, and social characteristics of agents; the causal architecture governing how those characteristics interact; the processes by which agents learn from one another; and the types of groups that agents can form…all without any coding. This product is part of our strategy to stabilize and grow the use of these methods within different disciplines.

The Modeling Religion Project, as a research methodology, pushed the study of religion into the computational and scientific space in a way that’s never been done before. There have been a few agent-based simulations of social and cognitive processes involving religion; they use relatively simple agents that clarify causal architecture but make the capturing of any complex cognitive cultural phenomena, including religion, extremely difficult. This project has created a new standard for theoretical consistency and empirical accountability in the study of religion, enhancing understanding of human life and its complex, adaptive systems.

Beyond research innovations, the MRP team members will continue to press forward in developing social simulation technologies for addressing humanitarian and societal challenges. The next step is to effectively share these simulations with policymakers and opinion leaders, in order to help them make fact-based arguments for robust solutions to pressing social problems such as religiously-rationalized violence, radicalization, mass migrations, and child trafficking.

Dr. Ross Gore noted that computer simulation “enables decision makers to enact policies that are not only more likely to succeed but in many cases simpler to implement. This approach is ruthless because it highlights bad decisions and policies quickly and thoroughly in a way that informed discussion cannot.” The applications of the Modeling Religion Project’s pioneering research are on the horizon.


MRP team members produced a series of simulations, sharing results at many conferences, producing a number of articles and an edited volume, a documentary film series, and greatly enlarging the conversation around this cutting-edge adventure in the academic study of religion. The project also helped spearhead a new grant from the Research Council of Norway that led to the Modeling Religion in Norway (MODRN) Project, and soon after, the Center for Modeling Social Systems (CMSS).

Christine Legare Presents at a Team Working Conference

Selection of Conference Presentations
2017 meeting of the International Association for the Psychology of Religion in Hamar, Norway. Paper title: “A New Model of Religion, Ritual, and Cybernetic Self-Regulation.” Authors: Wood & Diallo.

2016 meeting of the Complex Systems Society in Amsterdam. Paper title: “Why Hierarchy? Coupled Oscillator Dynamics and the Emergence of Social Complexity.” Authors: Wood & Lane.

2016 meeting of the American Academy of Religion in San Antonio, Texas. Paper title: “Simulating Shamanism: How Modeling and Simulation Can Help to Formalize Theories from Religious Studies.” Authors: Wood & Diallo.

2016 meeting of the International Association for the Cognitive Science of Religion in Vancouver, BC. Paper title: “The Logic of Ritual and Self-Regulation: A System Dynamics Model of Costly Signaling in Religious Groups.” Authors: Wood & Diallo.

For a full list of publications associated with the MRP, click here

Other Tools for Discussion
Video Narratives
Project Blog

Charisma, Ecology and Social Collapse: A Simulation Experiment

From the Roman Empire to the USSR, even the most sophisticated societies eventually collapse, leaving behind ruins, a displaced populace, or (at best) a whole lot of unpleasant paperwork. But some societies last a lot longer than others. Why? One answer might come from examining an unexpected topic: religion.

Religious communities are intensely cooperative, require lots of commitment from their members, and center on unfalsifiable claims. In this way, they’re nearly prototypical human societies. Most importantly, many religious communities fail – in fact, the vast majority fizzle out after only a few years, and truly enduring religious institutions are rare but notable.

To learn about the difference between communities that fall apart and those that stick around, researchers at the Center for Mind and Culture (CMAC) built and tested a computer model of a religious community as a complex adaptive system. Experiments showed that religious groups that were higher in charismatic leadership – but lower in structural authority – survived longer before outstripping their resource base.

As part of the Modeling Religion Project, CMAC Postdoctoral Fellow Connor Wood collaborated with University of Connecticut anthropologist and CMAC Senior Research Associate Richard Sosis to build a system dynamics model (SDM) of a complex adaptive religious system, or a dynamic system that uses feedback and constant adjustments to adapt to changes in its environment. The simulated religious community existed in an environment with limited resources, which it could harvest to use for economic, reproductive, or ritual purposes. The more the community invested in ritual, the more entrenched its religious beliefs and worldview became. But when things weren’t going well – that is, if the community’s overall death rate surpassed its birth rate, or if it couldn’t harvest enough energy – the religious system began to lose legitimacy.

Figure 1. Schematic of the System Dynamics Model of Religion as a Complex Adaptive System

For some communities, these crises of legitimacy triggered a switch to “charismatic” forms of authority, centered on high levels of supernatural belief and lower levels of formal authority. For experiments, the simulation engine produced 16 different combinations of model parameters that maximized the community’s population at a given time. But when allowed to run to completion, many of these combinations led to social collapses shortly after the end of the original timeframe. Wood and Sosis then re-ran all 16 simulations with the value for a parameter called “Charismatic Potential” maximized. In 14 out of 16 cases, the simulated communities survived longer – in some cases, much longer.

In the simulation, the Charismatic Potential parameter governed how a crisis of legitimacy affected the religious community. If its value was high, the religious community would respond to a legitimacy crisis by suppressing normal, structural authority (the equivalent of a formal, priestly hierarchy) and cranking up levels of supernatural belief. Many anthropologists have found that charismatic, emotion-driven religious practices result from significant changes in the sociocultural environment. Wood and Sosis posited that this response may be an adaptive attempt to adjust to these changes.

Figure 2. Difference between Time at Collapse between Low-Charisma and High-Charisma Settings for 16 Simulated Societies (T-Test P = 0.006)

Unexpectedly, lower levels of authority led to changes in the configuration of the rest of the system that kept it from outstripping its resource base. During this time societies looked more like cults than like normal religious communities – their core populations were very low, but they attracted a constant flow of interested converts who then quickly lost interest and departed.

This result makes intuitive sense, since cults often have high levels of charisma and low retention rates. But now we know that they might have an unanticipated effect on how the community lives within its ecological context.

The model was based on theoretical work by Sosis, whose previous research has examined the longevity of religious communes in 19th-century America. In that work, Sosis had discovered that religious communes that imposed higher costs on their members survived longer than those that were cheaper to join. But there’s more to community longevity than just investment barriers – religious communities also face challenges when their leadership loses credibility, or when their members can’t meet basic biological needs.

When religious communities lose legitimacy in this way, they can fall apart…or they adjust themselves by adapting to the new crisis. This realization led Sosis to conceive of religious communities as complex adaptive systems, or multilevel networks of social connections, practices, and cognitive factors that take in energy from their surroundings, reproduce themselves, and attempt to stay “alive” in the face of perturbations and challenges in their environments. This new simulation model attempted to explore the consequences of looking at religion from this angle – and ended up offering new insights into how societies stay alive.

Women, Ritual and Domestic Violence: A Cross-Cultural Analysis

Many American girls have been taught to consider their wedding day the “happiest day of their life.” But how do we reconcile that sense of hope with the fact that entering into an intimate relationship with a man is one of the leading risk factors for gendered violence? Women are more likely to be raped, injured, or killed by their current or former intimate partner than by any other person.

Balinese women in the Usaba Dangsil Ceremony; Balinese show high levels of female solidarity and low levels of domestic violence

Researchers in CMAC’s Sex Differences Project (SDP) are working to understand some of the sociological factors that lead to both increased risk and potential protection from domestic violence. Through an extensive cross-cultural study, they found that the prevalence of specific types of religious rituals—especially rituals that fostered solidarity among women and community, female initiation ceremonies, and the presense of female shamans—seems to actually help protect women against domestic violence.

Before jumping into why or how rituals generate protection against violence, let’s first take a look at the fraught relationship between marriage and domestic violence. Anthropologist Judith K. Brown notes that the excitement that we typically associate with weddings is far from universal. She explains in some cultures “the tears mothers shed at the weddings of their daughters are tears of true grief because their daughters will be separated from them and will embark on a life of toil, possible abuse, and the dangers of child-bearing under traditional conditions.” Similarly, among the Yanomamo of the Amazon rain forest, women’s fear of marriage appears to be somewhat proportional to the distance away from her family’s village—the further they are away, the less likely their brothers will be able to protect them from a cruel and violent husband.

Misogyny and antagonism toward women are incredibly complex phenomena that vary and morph depending on things like economics, social norms, and gender roles. But cross-cultural comparisons can help us make sense of some of those broader cultural factors.

For example, SDP researchers found that societies in which women’s bodies and menstrual cycles are viewed with shame or disgust have higher levels of spousal violence. Likewise, when women are excluded from elaborate male-only rituals it appears that domestic violence increases.

Such societies tend to have similar norms around where and how a couple lives after they get married. Namely, they tend to have what anthropologists call patrilocal marriage residence patterns. This means that when a couple gets married, the woman must leave her family, friends, and social network to go live with the man’s family; about 67% of cultures historically and worldwide are patrilocal.

Therefore, when SDP researchers began exploring the societal factors affecting the prevalence of domestic violence, marriage residence patterns took center stage. Using the Sex Differences and Religion Database (SDRD), a large cross-cultural dataset constructed by the SDP and consisting of 268 variables for 215 cultures, the researchers verified that, indeed, in societies that are patrilocal, domestic violence levels were significantly higher than when men were the ones to leave their families of origin to join their in-laws (matrilocal societies).

This particular finding is, unfortunately, not too surprising. When women leave their social support systems behind they are reliant upon relative strangers for protection, solidarity, and security. Anyone who has moved to a new town knows that it can be a significant challenge to make new friends and to find a place in a new social structure. For many people, one of their first tasks will be to find and join a religious community. Why?

Because, as the SDP researchers and many prominent anthropologists argue, religious communities are one way that people create fictive kin networks—essentially, networks of people who are not biological family, but who behave and feel as though they have a familial relationship.

With this in mind, the researchers next looked into the relationship between female-specific ritual participation and levels of domestic violence. The idea is that the female-centered rituals promote bonding and solidarity among women, which in turn, may have a protective effect against the high levels of domestic violence women in patrilocal cultures face.

And indeed, this is exactly what they found: cultures that present more opportunities for certain types of women-centered rituals have lower levels of domestic violence. This finding is strong support for the idea that rituals are really important bonding strategies for women who must create new support structures or fictive kin networks. The payoff is dramatic and clear: a potential decrease in the risk of domestic violence. More research is certainly warranted for such an important finding.  

Graphs generated by SDP Researchers depicting the likelihood of domestic violence in cultures with various rituals.

The 3P Directory: Categorizing Efforts to Combat Human Trafficking in MA

The Tools against Child Trafficking (TACT) project equips anti-trafficking efforts with the necessary resources to accelerate their solutions. The 3P Directory is one such tool.

The Directory categorizes anti-trafficking organizations based on their key activities in three policy dimensions – protection, prosecution and prevention – to shed light on the existing efforts.

The Directory demonstrates where resources are mainly focused and provides intel to better allocate the limited funding in the future. The 3P Directory can foster the sharing of knowledge, the development of best practices and the initiation of partnerships among stakeholders. Those assisting the survivors of human trafficking will be better informed and thus more effective.

The 3P Directory will also help strengthen the 4P policy dimension – partnership.

Design by Bernd Durrwachter

The 3P Directory includes organizations from Allies Database, developed by MA Coalition to End Human Trafficking (MCEHT), mainly focusing on Massachusetts.

Users have the flexibility to filter the directory by the 3P policy dimensions –protection, prosecution and prevention, location, types of trafficking, population served and the key terms by clicking on one of the filter buttons or using the search button. The search button allows for filtering one or more words using a space between the words. For example, if a user wants to see all the organizations located in Boston that focus on children and protection, they can type “Boston children protection” in the search box. For a list of organizations with their main activities focused on prevention, the user can click on the filter button “prevention” or use the search box.

The 3P policy dimensions – protection, prosecution and prevention – are reflected in the Protocol to Prevent, Suppress and Punish Trafficking in Persons, Especially Women and Children (Palermo Protocol or Anti-Trafficking Protocol), the United Nations Convention Against Transnational Organized Crime (UNTOC) and in the United States’ Trafficking Victims Protection Act (TVPA). They are used as the fundamental framework around the world to combat human trafficking. The 3P Directory measures and evaluates the three main policy dimensions based on the requirements of Anti Trafficking Protocol.

Our categorization is an adoption of the 3P Policy Index Cho, Dreher & Neumayer (2014) developed using the US Department of State’s Annual Reports of Trafficking in Persons and UN Office on Drugs and Crime Reports on Trafficking in Persons: Global Patterns. The index measures the three main dimensions of anti-trafficking policies of the governments of up to 180 countries over 2000 – 2010 period based on the requirements of Anti Trafficking Protocol.

If the organization you represent is missing and you would like your organization to be added to the 3P Directory or if you have any other inquiries, please send an email to