Looking Inside of Learning

Michael Connell’s fascination with “neural networks”–computer programs that simulate the activity of brain cells or neurons and actually learn over time–stems in no small part from a “crystallizing moment” he experienced in ninth-grade trigonometry.

His teacher brought in a mathematics journal to show students a picture of a spiraling flower made up of numbers–the result of mapping a sophisticated trigonometric equation. Curious about this phenomenon, Connell copied the equation, took it home, and tried to plot it on his Atari computer. “It took me eight hours to figure it out,” he recalls with a rueful laugh. “But when I did, it was one of those thrilling experiences, one of those eureka moments that changed my life.”

Connell went on to study computer science and electrical engineering at MIT, as well as psychology and cognition. He even enrolled in MIT’s doctoral program in computer science, until a course on learning environments provided him with another eureka experience.

“We talked about problems in understanding and problems in knowledge. We spent weeks at a time creating computer simulations of the ways that people come to understand what causes traffic jams or why the moon goes through phases,” says Connell. “I didn’t know anyone could get paid to do this.”

Inspired to study human learning processes further, Connell made his way to GSE–after a stint at Microsoft–and began designing mathematical computer models to explore human cognition. Now a third-year doctoral student, he uses software to explore how the human brain thinks and learns.

“Take learning physics, for example,” he says. “What if there are a small number of conceptual holes into which people fall or get stuck? Can we map those holes and thereby help people to avoid falling into them? That’s what I’m trying to find out.”

His long-term vision is simple, says Connell. If neural networks can be used to understand how the brain processes information, “educators can design curriculum more effectively.”

Connell has long been fascinated by artificial intelligence–the field that studies how machines can be designed to replicate intelligent human behavior–as the potential means by which everything from education to life in the workplace can be improved. Now he is using the tools of artificial intelligence to maximize human potential rather than the capacities of machines.

Instead of approaching learning problems from a hypothetical scientific premise in search of solutions, as he did at MIT and Micro-soft, Connell uses the needs of the child or the adult as a starting point. “Many engineers never get to see the real applications of what they do,” he muses.

“While I don’t expect to solve the problems of cognition in my lifetime, I might come up with an adaptable software system for a child with a learning disability or help other educators better understand under- and overachievement in kids. And that’s pretty exciting.”

Education, adds Connell, serves as an important nexus between the sciences and human values. “Science can tell us how best we should use the brain, or how kids differ from one another, but it can’t tell us what we should teach.”