Are You an Analyzer or a Synthesizer? Or, What Robots Can Tell Us About Creativity

What we can’t program into AI tells us a lot about ourselves.

I started my PhD on an interdisciplinary fellowship that required me, a social scientist, to be housed in a STEM lab with other fellowship students who were in fields such as ecological engineering and biochemistry. I’ll never forget what our faculty mentor said on our first day.

“We don’t need more analyzers,” he said. “We don’t need more people who study ever-smaller parts of the problem. We need synthesizers. People who see the bigger picture and cross disciplinary boundaries.”

The fellowship was meant to train a new generation of scholars who were not constrained by the traditional academic project of specialized analysis in disciplinary silos. Unfortunately, I realized early on that while academic institutions may pay lip service to the value of interdisciplinary scholarship, the way they and their peripheral institutions (i.e. funders, publishers) are structured makes it extremely difficult to do good interdisciplinary work.

The problem is that analytical work is just easier to value within our system. We know how to fit it into existing structures of knowledge, and it’s easy to calculate its value based on known parameters. Analytical methodologies are more predictable and easily systematized. A prime example is the scientific method. It looks the same regardless of your project.

We have a system built on analysis. So what about synthesis? What is it about the work of synthesis that makes it kind of like that guest you regret inviting to the party because they’re so perplexing and disconcerting? Here’s where robots come in.

Artificial intelligence (AI) has seen amazing advances in the last fifty years, and is currently a burgeoning and exciting field of inquiry. One of the challenges is figuring out how to program self-directed learning. AI robots are skilled at analytical tasks like deduction and modeling – both important in terms of learning based on expected outcomes. But they have made few strides in synthesis – which is a learning process that results in unexpected outcomes. Analysis is inward-looking and makes use of a set of constraints that guide the inquiry and outcome. Synthesis is outward-looking and requires both an open field of inquiry and open-ended outcomes. Analysis can deliver understanding of current conditions. Synthesis can deliver new solutions to current problems.*

What are AI robots missing? According to roboticist Hod Lipson, two essential types of intelligence: creativity and curiosity. We have not yet figured out how to program and operationalize these traits in AI. Partly this is because we don’t understand how they work, particularly creativity. Creativity is fundamentally an experimental and evolutionary process. Evolution proceeds through trial and error, and without a specified or predictable end goal. It is experimental in nature, and very much dependent on a complex interplay of constantly changing inputs and incremental outputs.

Curiosity is perhaps the greatest driver of creativity, besides the problem structure itself. Curiosity is an active drive that pushes us to pursue knowledge not only about the world right there in front of us, but about abstract unknowns. It is what enables us to not only learn what we need to know about our immediate environments in order to survive, but to imagine what lies beyond the horizon. It is at the core of human adaptability and our success as a species. And on an individual level, curiosity functions in a similar way. Curious people often fare better because they’re better at handling uncertainty, ambiguity, and novelty. They’re creative problem solvers.

The challenge of programming creativity and curiosity into AI underlines just how distinctive and exceptional these traits are in human beings. They are perhaps among our most valuable characteristics, and should be fostered at every turn. While it remains to be seen if we’ll figure out how to design AI robots that match humans in these capabilities, we can intentionally direct and grow our own creativity and curiosity. And may I suggest we combine those with another trait that thus far eludes AI: kindness.

*This and the following sections on AI and robots were inspired and informed by this paper by roboticist Hod Lipson and this interview with scientist Lex Fridman.