Computer Models of How Ideas Evolve and How We Became Creative

This post is Part 1 of a series.

Creative Culture as an Evolutionary Process

When I was a graduate student studying "artificial life," I developed what she thought was the first computational model of biological evolution through natural selection. You really understand something when you can build it from scratch. My enthusiasm for this project dimmed when I found out that someone else, John Holland, had created a very similar model 15 years earlier, and his had even already been applied to all kinds of optimization problems, such as scheduling algorithms and pipeline configurations.

I realized that there was another process that appears to evolve—and that is culture. I resolved to be the first person to develop a computational model of how culture evolves—and I was. (There’d been an earlier model of cultural transmission, in which there is only one possible cultural output, with only two versions of that cultural output, and an artificial society that converged on one version or the other. However, for cultural evolution, there must be enough possibilities to exhibit cumulative change over time, with new innovations building on previous ones, and her model was the first to do that.)

The computer model of cultural evolution used what is called an agent-based model. An “agent” can be thought of as a character in a simulation that represents a human being. It learns ideas for actions that it can implement with its "body," at which point the action becomes visible to other agents. In each iteration (step) of a run of the program, each agent does one of two things: It either imitates what a neighbour is doing or invents a new action, building on an action in its current repertoire. Through iterations of inventing and imitating, the mean fitness and diversity of the actions across the artificial society of agents increases over time. It’s possible to experiment with the effect of varying certain parameters, such as the ratio of imitating to inventing, or the capacity to mentally simulate the result of implementing an action before actually committing to it, and observe how that affects the pattern of cultural change.

The Origins of Human Creativity

While that branch of my research program focuses on how ideas evolve, another branch focuses on how humans evolved the capacity to be creative. This branch involves figuring out what kind of cognitive transitions were behind transitions in the archaeological record.

Our distant ancestors started developing simple tools more than three million years ago by smashing one rock against another to sharpen it so it could be used for digging or cutting. Then, some 1.76 million years ago, Homo erectus developed a three-step procedure that resulted in more effective, easier-to-grasp tools. While the result may appear primitive by our modern standards, in a world where everything people encountered was shaped by forces of nature, something that began as a spark of insight in someone’s mind and culminated in a product that actually made people’s lives easier and helped them survive would have seemed like a miracle. Moreover, humans today actually find these tools quite tricky to make!

Synthesizing this archaeological evidence with evidence from anthropology, cognitive science, and neuroscience, my colleagues and I hypothesized that this transition 1.76 million years ago occurred because larger brain size enabled memories to be encoded in more detail, which made them overlap more, such that they could evoke each other, allowing one thought to lead to another in a "stream of thought." This, in turn, was the beginnings of cultural evolution. When this hypothesis was tested using an agent-based model by giving agents the capacity to chain thoughts together, they did indeed start to slowly evolve cumulative culture.

There wasn’t a lot of improvement in stone tools for more than a million years; there was "stasis" in the archaeological record. Then, about 100,000 years ago, we see innovation: art (cave paintings, stone statues), evidence of ritualized religious practices (burial of the dead with objects), and task-specific stone tools. This period has (albeit controversially) been referred to as the “birth of art, science, and religion.” Artifacts were no longer solely functional; they could now help us express ourselves creatively and develop distinctive cultural traditions.

Again synthesizing literature from multiple fields, we hypothesized that this second cultural transition was due to the onset of the capacity to shift between different modes of thought according to the task at hand. Divergent thought involves thinking about things in unconventional ways and making new connections. It is conducive to imagination, metaphors, analogies, and dreaming up wild new ideas. If you’re stuck in a rut, it can be helpful to think more divergently.

On the other hand, in the convergent mode of thought, less energy is devoted to finding obscure connections, so there’s more energy left for logic and reasoning. Think of an engineer building a bridge and determining what the best structure will be for durability, effectiveness, and ease of use.

The ability to shift between these modes of thought is called contextual focus because it’s carried out by focusing or defocusing attention. We tested the theory that the onset of contextual focus brought about cultural transition 100,000 years ago using the agent-based model. With the contextual focus parameter turned on, agents could adjust their mode of thought to the situation they were in. If they weren’t having much problem-solving success, they generated riskier, more unconventional possibilities, not at random, but using trends they’d picked up over time about what makes for a good solution. Alternatively, if they were on a productive roll, they generated minor refinements on what they were already doing. We found that the mean fitness and diversity of ideas increased, as hypothesized.

Thank you to Eric Bollman of the editorial staff at the Canadian Psychological Association for helping to shape this post.