Let’s stop using computer models that reinforce racism within our communities
By Dr. Laura Schmitt Olabisi
My multi-racial family lives in East Lansing, Michigan. The neighborhood is peaceful, tree-lined, and conveniently close to Michigan State University, where we work. It’s a place where we bring our neighbors baked goods, and where our kids ride bikes together with the neighboring kids, up and down the streets.
In the 1960’s my family couldn’t have lived in our current neighborhood. At that time, property deeds and court orders explicitly prohibited non-white people from moving into the city. Local activism and civil rights legislation changed the policies, but the legacy of segregation remains in the racial makeup of East Lansing, which has a Black population of only 7.8%, compared to 22% in the neighboring city of Lansing.
In Lansing, there are many families that look like mine, while in East Lansing, there are probably more ‘Black Lives Matter’ signs on display, than actual Black people living here.
The whiteness of East Lansing and many other places in the United States was not merely a matter of families’ location preference; an outcome of the relatively innocent choices of homeowners, driven by personal preferences.
Color lines were codified into law. The United States, throughout its history, has ruthlessly enforced segregation in federal, state, and local law, in bank lending policies, in law enforcement practices, and in school zoning.
In order to model the causes of segregation, the economist Thomas Schelling simulated individual homeowners as living in a grid space, where they were given the choice to stay or move away from their neighborhood based on the race of their neighbors. The Schelling segregation model showed that even if homeowners had only a slight preference (requiring only 1/3 of their neighbors to be of the same race - white/Black), that given ‘enough’ time, complete segregation would occur.
The Schelling model reveals emergent behavior generated within the system (segregation), which cannot be predicted by looking at the individual components in isolation. Emergence is like tasting a cake. If you set out all the ingredients for baking a cake (the butter, sugar, flour, eggs, etc.) and tasted them one by one, you wouldn’t be tasting a cake. The emergent property (the taste) arises from the combination of the ingredients and their interaction with heat in the oven. Similarly, the emergent property of segregation, according to the Schelling model, arises from the interactions of thousands of individual household choices about where to live.
Two major problems exist within the Schelling model, reinforcing racism within it:
1. The assumption of personal preference – the desire to live among ‘alike’ people.
a. In most US communities—including the one in which I currently live—Segregation
was the law of the land.
2. The assumption that racial grouping is a ‘natural preference’.
a. There is nothing ‘natural’ about racist preferences in the United States – these
preferences are the result of hundreds of years of white supremacist ideas woven into
the very fabric of our culture.
Moreover, we should see segregation and racist preferences as two sides of a
reinforcing feedback loop, which is not a part of the Schelling model. Separating white
families from potential Black neighbors makes them more likely to believe negative
stereotypes about Black families, because they have no experiences to challenge the
views of such propaganda. Therefore, when white families believe these negative Black
stereotypes, they will then prefer not to live near them. This loop reinforces the distance
between white and Black families.
This is how segregation and racism feed on one another.
Putting these two assumptions together—that segregation is the result of personal preferences, and that this group preference is ‘natural’— leads to a story, perpetuated by the Schelling model, that lets white racism off the hook, because you can’t legislate personal preference!
Our nation is facing a long-overdue reckoning with racism, and as complex systems modelers, we have a responsibility to join this conversation. Segregation is one of the most enduring and effective tools of past and present white supremacy within the United States. When analyzing the simulations we use to model communities, we must embrace rigorous analyses and provide historical context to better understand how segregation influences racism and vice versa.
When analyzed, the Schelling model does not pass that test.
Let me be clear: I am not saying that everyone who used the Schelling model is racist, nor that Thomas Schelling was racist. After all, I have used the model myself, so I must be included in this category. Most of us accept that the original intent of the model was to figure out how to avoid segregation. But intent is not impact—and the impact of the model, now, is preventing us from dealing with the root causes of segregation and racism.
As Maya Angelou famously said, “when you know better, you do better”.
Systems modeling tools have a role to play in helping us address the most complex problems we face as a society today, including segregation, the COVID-19 pandemic, and combating climate change. In fact, models, like the Schelling model, are being used in policy discussions around these critical issues right now.
However, our models are only as good as the assumptions that go into them.
As I walk through my neighborhood, I think about my children, and all our children. No matter their race, our children deserve models whose assumptions critique and challenge the hard realities of the past and are improved upon in the hopes of building a better, and more equitable future. We can make different choices now. We can choose to be deliberately inclusive and build models that undo, rather than reinforce, systemic oppression, to develop policy solutions that benefit people from all walks of life.
Let’s begin to break the racist feedback loops, and truly build a better world.
DR. LAURA SCHMITT OLABISI is an Associate Professor in the Department of Community Sustainability and the Environmental Science and Policy Program at Michigan State University. She is an ecologist and a participatory systems modeler, working directly with stakeholders to build models that foster adaptive learning about the dynamics of coupled human-natural systems, and to integrate stakeholder knowledge with academic knowledge. She was an AAAS Leshner Leadership Institute Public Engagement Fellow in 2018, and is currently a board member of the Academy for Systems Change (formerly the Donella Meadows Institute), a non-profit organization dedicated to training systems leaders for sustainability transformations in the public, private, and nonprofit sectors.