Welcome to The Flux, where we hear stories from people who have asked what if questions to better understand the world and talk about how data can help tell stories that impact decisions and create an intentional impact on the future. This is your host, John Cordier, CEO at Epistemix. In a world where the flux capacitor from back to the future does not yet exist, people have to make difficult decisions without always knowing how the future will play out.

Our guests are people who've taken risks, made decisions when uncertainty was high, and who have assisted decision makers by using data and models. We hope you can turn lessons from our podcast into foresight, so you or your organization can make better decisions and create an intentional impact for others.

–
John: Hey there. Welcome to The Flux. We have Stefani Crabtree with us today. Stefani, thanks for joining.

Stefani: Nice to meet you and be here.

John: We’re at the CSSSA, the Complex Systems Society of the Americas Conference, which happens every year in Santa Fe, New Mexico. Stefani was a keynote this year. Why don’t you tell the group a little about what got you into agent-based modeling in the first place?

Stefani: My background is in archaeology, more or less. I always say I rewind the agent-based models in my head to different decisions. When I was an undergraduate, I actually had many different majors. I was a political science major, a music major, and a theater major. I went to study abroad in Paris and enrolled in a theater class and thought, oh no. But I also enrolled in an archaeology of Egypt and Islam class.

John: Okay.

Stefani: And I thought, I am changing my life. I am going to become an archaeologist. I got to do weekly tutorials with my amazing professor, and we went to the Louvre and all this stuff. I really drank the archaeology Kool-Aid. I finished my undergrad degree and had a Watson Fellowship. I traveled around the world and did anthropological research into traditional medicine.

I promise we will get to the agent-based modeling eventually.

John: It’s never a linear path.

Stefani: Never a linear path.

John: No.

Stefani: After the fellowship, I worked in a human rights nonprofit and then did contract archaeology for a while. If you are putting in a gas pipeline or a wind farm on public land, you have to bring an archaeologist to make sure you are not disturbing archaeological remains. That could be Native American remains or historical remains, anything older than 50 years.

I did this work for a while and thought I wanted to go back to graduate school. All my friends in Boulder were physicists, so I was aware of computational approaches, but not necessarily agent-based models. I saw my friends making models to understand distant stars and thought the past is distant. Could we use models to understand the past?

At the same time I was also a musician. My bass player was a PhD student at CU Boulder, and a book came out called A Model-Based Archaeology of Socionatural Systems. I asked him to rent it for me. I remember him showing up at my show at the Buffalo Rose with the book saying, I checked this out for you. Do not lose it or I will get in trouble.

John: Yeah.

Stefani: I read it and probably understood about 25 percent of it. It was written by Sander van der Leeuw and Tim Kohler. I wrote to Tim Kohler and ended up doing my PhD with him.

He was developing an agent-based model to understand the ancestral Pueblo Southwest. I did not know where I wanted to work archaeologically. I knew I wanted to understand human-environment interactions in deep time and be able to do that all over the world. I saw agent-based modeling as a path to do this.

I applied to multiple graduate programs and in every one I said I wanted to do agent-based modeling. Most of the professors said, what is that? This was 2009 when I started graduate school.

John: They did.

Stefani: But I got into a couple of programs and decided to work with Tim Kohler, who was the architect of the Village Ecodynamics Project agent-based model. I did my master’s looking at exchange networks in the American Southwest and then stayed with him for a PhD. I also got a second PhD in France using agent-based models to understand the beginning of the wine industry.

John: Now we’re talking. One thing that comes up is that agent-based modeling can be applied in a lot of different domains. My domain was public health, like epidemiology, where people somewhat accept agent-based models. What kinds of pushback show up in anthropology or archaeology related to agent-based models?

Stefani: I would characterize a couple different types of pushback. One comes from being a modeler. One of my biggest challenges working with people who do not build models but want to model human systems is that they want to put everything into the model. They think everything will be important, and that can be very difficult in an agent-based model.

That creates pushback from the broader scientific community because people think you can put anything you want in a model. A lot of early models tried to include as much as possible without understanding how the parameters interacted.

As a modeler, you have to be aware of when people want to put too many things in.

When you are doing ecological models, like wolves and sheep predation using Lotka-Volterra dynamics, wolves cannot talk back. They cannot say you need a better model of reproduction or that they had chipmunks for breakfast instead of sheep. But people can talk back.

One challenge of being both an anthropologist and a modeler is stopping myself from putting too many things in and managing collaborators who want to add too much. Early models sometimes had too many components, which made it difficult to say whether they were telling us something meaningful about reality.

John: Do you think part of that is that people get excited about having data and feel they need to use it?

Stefani: Absolutely. People get excited about the thing they care most about in their system. One thing that has improved in agent-based modeling with more training, more papers, and more replication is that people are getting better at building models from simplicity to complexity. Instead of throwing everything in at once, they start simple and gradually add components.

Another pushback I have received is about flattening society. People say humans are so complex, can we really model them? Is it a good thing to model people at all? My work shows there are questions where models enrich our understanding rather than oversimplify it. They allow us to address aspects of the past that are very difficult to study without computational models.

John: That makes sense. Was some of your inspiration to understand the past also about how it informs decisions today?

Stefani: Absolutely.

John: Have you been able to bridge people into that perspective?

Stefani: Yes. A lot of my work is not agent-based modeling. One of my papers uses network models to examine what humans do in ecosystems and uses simulations to study natural fluctuations in ecosystems.

This paper was published in 2019 and came from work I did while living with an Aboriginal community in the western desert of Australia. It has been picked up by the Australian government for policy documents, which surprised me because I originally pursued it simply out of curiosity about human-ecosystem interactions and through anthropological work with Aboriginal communities.

It was rewarding to see something that felt esoteric helping to guide policy in Australia.

Another paper I published in 2017 used agent-based modeling to study the growth of hierarchy in the American Southwest. The inspiration partly came from being a middle child who likes to make sure everyone gets along.

There was a long-standing debate in the 1980s. Archaeologists studying Chaco Canyon saw burials of individuals with copper bells, turquoise, macaw feathers, and cylindrical jars containing residues of theobromine from chocolate traded from Mexico. They argued this indicated a hierarchical society.

Anthropologists working with descendant communities such as the Hopi, Zuni, and San Ildefonso observed strong leveling mechanisms that prevented individuals from accumulating more status than others, suggesting an egalitarian society.

These two perspectives clashed. Some scholars argued ancestral Pueblo society was hierarchical, while others insisted it was not.

I built an agent-based model to create a digital archaeological record that could be compared to real data. The conclusion was that both sides were right. The society began relatively egalitarian. Hierarchy emerged around the 1100s. Then during periods of poor climate conditions, that hierarchy broke down into regional hierarchies that eventually collapsed as people abandoned the area.

This paper was my first major use of agent-based modeling to resolve a long-standing debate, and it has helped advance Southwestern archaeology.

John: When both sides could see how the system played out, it allowed them to have a discussion rather than an argument.

Stefani: Exactly.

John: We see that in health systems too. For example, when introducing a new medical product, companies want to sell it while insurers want to control costs. Modeling can help find solutions where both parties benefit.

Stefani: It can be very helpful. Like any scientific tool, agent-based models can be used poorly. But if they are used carefully and with clear questions, they can advance science. Because you know what you are putting into the model, they can also serve as a mediating tool that helps people understand complex systems more deeply.

John: One question we ask everyone on the podcast is what would make agent-based modeling more accessible to more people?

Stefani: In 2021, with two co-authors, Za Romanowska and Colin Wren, I published an open-access textbook called Agent-Based Modeling for Archaeology.

You can buy a physical copy on Amazon for about ten dollars, or download it for free. In the book we recreated about 76 models and walked readers through building them. We included examples like an SIR model, models from political science, anthropology, archaeology, and the tragedy of the commons.

Za, Colin, and I had been teaching workshops at conferences that filled up quickly. We realized we should turn those tutorials into a textbook to reach more people.

Many of us who went to graduate school in the 2000s had to teach ourselves through trial and error. I did have one course in agent-based modeling at Washington State University.

John: Was that NetLogo or something else?

Stefani: It was Repast.

John: Right into the fire.

Stefani: Exactly. Many things we had to learn on our own. Having affordable and accessible resources helps bring agent-based modeling to more people. We also ran a blog called Simulating Complexity for a while.

Beyond building models, I think model literacy is important. Thinking like a modeler changes how you approach problems. You break systems down step by step. If you want to model wolf-sheep dynamics, you think about reproduction, interactions, and processes. If you want to model hierarchy, you decide whether agents represent individuals or families and what mechanisms drive inequality.

That step-by-step reasoning is valuable for any scientist.

John: Since you study the past, we ask everyone a “What the Flux” question. A What the Flux moment is a point in time when a different decision might have changed how the world unfolded. What moment would you go back to?

Stefani: I will give you two, one for the world and one for my life.

Since 2012 I have worked in Mongolia. Mongolia is a beautiful country, and everyone there loves Chinggis Khan. Outside Mongolia he is often seen as a destructive force because the Mongol expansion spread across Eurasia and caused massive upheaval.

The expansion moved west toward Europe. They reached as far as Poland. But the expansion stopped when Chinggis Khan died. The generals and his sons had to return to Mongolia to choose a successor.

My What the Flux moment is what would have happened if that had not occurred. What if the Mongol armies had continued west into Europe? Medieval Europe might have been very different.

During a conference in Kraków I saw an exhibit showing houses burned by the Mongols. It was a powerful experience. In Mongolia he is a national hero, but elsewhere he is seen very differently. I sometimes wonder what Europe would look like if the Mongol expansion had continued all the way to France.

John: That is incredible. What about your personal one?

Stefani: I applied to graduate school several times and was rejected. I now have two PhDs and I am an associate professor, but at the time I was not a competitive applicant and I did not clearly know what I wanted to study.

I decided to apply one last time after discovering agent-based modeling, and that is when I finally got accepted.

At the same time I was also a musician. I released two CDs, one live and one studio album. I played a show at the Buffalo Rose in Golden, Colorado with a few hundred people in the audience. At that show was an agent named Lance Penksy who had helped launch the band The Fray.

He came to my show, invited me to his studio, and wanted to produce an EP and launch my career. He told me I was getting a little old for the music industry, I was 26, so we would have to move quickly. He thought I could be the next Jewel.

Right around that time I received my graduate school acceptance letter. I had to decide between pursuing music or going to graduate school.

I thought about it and realized that life on tour was not for me. Contract archaeology had already shown me what constant travel felt like. My What the Flux moment is what would have happened if I had chosen the music career instead.

I imagine I might have spent eight years trying very hard, gotten exhausted, and eventually gone to graduate school anyway.

John: That makes sense. I think the lesson there is that you have a lot of grit.

Stefani: One thing I tell my students is that there are many paths to happiness. Failing at something can be hard, but sometimes it opens another door.

My yoga instructor once said that when you say yes to something, you are simultaneously saying no to something else. That lesson applies to academia as well.

There are many ways to be happy and many careers you could pursue. I am 43 and still have not decided what I want to be when I grow up. I love what I am doing right now. I love my department, my students, and writing. Archaeology is exciting.

But I also love cooking, painting, and many other things. We only get one life, and there are so many ways to spend it.

The cool thing about agent-based models is that they let you ask what if. You can model alternative possibilities and imagine how different decisions might play out.

John: That is a beautiful way to tie it together. Agent-based models let you ask what if and explore it.

Stefani: Exactly.

John: Stefani, thank you so much for being on. This has been great.

Stefani: Thank you so much.