Can an Artificial Civilization Make Its Own History?

Why do strangers trust each other?

Why do people cooperate, even when acting alone might seem easier? Why do some communities flourish while others fall apart? And why can thousands of ordinary decisions create outcomes that nobody intended?

We encounter these questions every day. In markets, neighborhoods, workplaces, and even within our own families.

But what if we could build a small world and watch these processes unfold from the beginning?

That is the idea behind SHING CIVITAS.

SHING CIVITAS is an ongoing experiment in building an artificial civilization—not by writing its history in advance, but by giving its inhabitants the ability to make decisions, learn from experience, and live with the consequences.

Its residents are not all-knowing. They have limited information, resources, time, and cognitive abilities. They can make mistakes. They may discover useful tools, exchange resources, help others, or fail to cooperate.

And none of these outcomes should be guaranteed.

The project draws inspiration from decades of social science: Herbert Simon's bounded rationality, Thomas Schelling's work on how individual choices produce collective patterns, Robert Axelrod's studies of cooperation, Elinor Ostrom's research on institutions and shared resources, and theories of cultural evolution.

Using Python, agent-based modeling, computational decision-making, and reproducible experiments, I am exploring how these ideas behave inside a living digital world.

What happens if resources become scarce? What if residents learn differently? Could two identical societies develop entirely different histories?

These questions also inspire one of SHING CIVITAS's long-term goals: parallel universes, where different conditions can be explored from the same starting point.

SHING CIVITAS is not a prediction of humanity's future. Many of its mechanisms are still being developed, tested, and challenged.

It is an attempt to ask an old question in a new way:

If history is made by countless individual decisions, can we begin to understand how it emerges?

From Studying Society to Building One

Hi, I'm Martin Zhong, from Shanghai Jiao Tong University.

I've always been fascinated by human society and civilization. Why do people move? Why do institutions work differently across places? Why do some policies improve people's lives while others produce unexpected consequences?

Before the recent AI revolution, I was trained as a traditional social science student.

My research experience has covered internal migration, cross-border healthcare systems between Hong Kong and mainland China, the effects of education policies, and the factors influencing caregivers' well-being.

Along the way, I received systematic training in both exploratory and explanatory social science research, working with qualitative methods as well as quantitative approaches such as Difference-in-Differences (DID), Structural Equation Modeling (SEM), network analysis, and mediation analysis.

But one question has always bothered me:

How much can we really learn about a society from observing it?

Traditionally, we collect data from people through interviews, surveys, experiments, and administrative records. These methods have taught us an extraordinary amount about human behavior.

Yet they also face limitations.

What if we want to study two societies that begin under exactly the same conditions but develop differently? What if we want to observe how institutions emerge across hundreds of generations? What if the consequences of a decision only become visible centuries later?

We have experimental and quasi-experimental methods, of course. But controlling an entire society is practically impossible, and following one through centuries of change is extraordinarily difficult.

So, could artificial intelligence open up another way of asking social science questions?

Recent research on silicon samples has explored using large language models to simulate human responses to surveys and experiments. Many such applications offer something resembling a snapshot of behavior at a particular moment.

But I began wondering:

What if we could go beyond silicon samples and build a silicon society?

Not just AI agents answering questionnaires, but a living artificial world where individuals have memories, limited knowledge, resources, relationships, and histories of their own.

What if they could learn from experience, make mistakes, cooperate, disagree, and influence the world that future generations inherit?

And what if we could create parallel versions of that world, change one condition, and observe how their histories unfold differently?

That curiosity became SHING CIVITAS.

I don't believe artificial societies can simply replace research involving real humans. A simulated society is not the same as a human one, and its findings must be interpreted with care.

But perhaps such worlds could become another experimental instrument for social science—one that allows us to explore mechanisms, challenge assumptions, and ask questions that are difficult to investigate in reality.

I don't know how far this idea can go, or whether it will lead to genuinely new scientific discoveries.

But I find the possibility too fascinating not to try.

So I'll keep building, experimenting, learning, and questioning my own assumptions along the way.

And if you have technical questions, philosophical disagreements, research ideas, or simply share this curiosity, I'd love to hear from you.

After all, good research often begins with a question nobody knows how to answer yet.

What Can We Learn From a World That Doesn't Exist?

Why publish an experiment that failed?

Why share a technical problem that hasn't been solved?

And why should anyone care about what happens inside an artificial civilization?

Because sometimes, understanding why something didn't happen is just as interesting as understanding why it did.

SHING CIVITAS is an open-ended project, and this website is its public research journal.

Here, I document experiments, unexpected discoveries, technical challenges, and ideas that haven't worked as planned. I also share the methods and evidence behind the results, so others can understand, question, and, where possible, reproduce them.

Who is this for?

For curious people who want to understand how societies work. For developers exploring autonomous agents. For students and educators looking for more intuitive ways to understand complex systems. And for researchers interested in the possibilities—and limitations—of artificial social simulations.

You don't need to understand Python or read academic papers to begin exploring. I want to make complicated ideas accessible through clear explanations, visualizations, and real experimental examples.

The website is available in English and Chinese because curiosity should not be limited by language.

Will these experiments lead to scientific discoveries? Useful technologies? New ways of understanding social behavior?

I don't know yet.

But I believe the questions, the experiments, and even the failures are worth sharing.

Knowledge grows when we make our ideas open to examination, discussion, and improvement.