Every year in mid- to late summer, cognitive scientists from around the world gather expectantly in a hotel foyer or a university courtyard, eager to learn that year's winner of the David E. Rumelhart Prize. Established in 2001, the yearly award honors "an individual or collaborative team making a significant contemporary contribution to the theoretical foundations of human cognition." The award includes $100,000 and a custom bronze medal. It's the closest thing you'll find to a Nobel Prize in cognitive science, the interdisciplinary study of the mind that arose after the "cognitive revolution" of the 1950s and 60s.
Michael I. Jordan is the Pehong Chen Distinguished Professor of Statistics and of Electrical Engineering and Computer Science at the University of California at Berkeley, where his research has focused on learning and inference in both humans and machines. Jordan is a leading figure in machine learning and Bayesian nonparametrics — a statistical approach that supports flexible models that can "grow" as more data becomes available. The computational models he's developed have been applied to learning, memory, natural language processing, semantics and vision, among other facets of natural and artificial intelligence.
Fittingly enough, Jordan was a student of David Rumelhart (1942 - 2011), the pioneering cognitive scientist in honor of whom the Rumelhart Prize is named. Rumelhart made seminal contributions to our understanding of the human mind, developing mathematical and computational approaches that still shape the field today. He's probably best known for his work with Jay McClelland (the 2010 Rumelhart Prize winner) on neurally-inspired, parallel distributed processing systems.
In an email conversation with me, Jordan praised Rumelhart's broad, interdisciplinary approach:
"Dave had an expansive vision of cognitive science — ideas from psychology, linguistics, AI, statistics and philosophy infused his thinking. Where lesser minds tended to develop cartoon versions of ideas from other fields — the better to dismiss those ideas and continue to focus on one's narrow perspective — Dave always found something useful in other traditions, and was able to creatively shape ideas from those traditions to tackle challenging problems in cognition."
At various points in his career, Jordan has asked himself, "What would Dave have thought about this?"
The Rumelhart Prize owes its existence to another student of David Rumelhart, Robert Glushko, whose foundation funds the award. Glushko was also kind enough to correspond with me by email, explaining that he established the yearly prize because:
" ... [Rumelhart was] a brilliant scientist who made fundamental contributions to the foundations of cognitive science, but also to honor a person who was personally and professional generous – a great model for aspiring scientists."
Other recipients of the Rumelhart Prize have included Ray Jackendoff (2014), Linda Smith (2013), and my own PhD advisor, Susan Carey (2009).
I asked Glushko what Rumelhart might make of the field today:
"Rumelhart liked mathematical rigor but he also liked elegant models that gave unexpected insights, and that perspective characterizes the best scientists of any field at any time. I think that Rumelhart might be surprised at how computational cognitive science has become."
Glushko noted that students don't just drift in to cognitive science from psychology anymore — they often come from mathematics, computer science and related fields.
Jordan speculated that Rumelhart would have embraced the computational tools available to today's cognitive scientists:
"I don't recall Dave ever uttering the word 'Bayesian' ... [but] as I've explored Bayesian ideas over the years I often thought that Dave would have enjoyed this exploration as well, with its natural connections to psychology, computer science and stochastic processes."
"I think that he would have been a particularly enthusiastic participant in the rise of Bayesian nonparametrics, appreciating its freedom to allow new entities and structures to emerge as data accrue. Anyone who is pondering the use of statistical models as models of thought must eventually wonder 'how does the model itself evolve?' and 'what happens when data arrive that appear to go beyond the scope of the current model?' The ability of Bayesian nonparametrics to face these questions should be appealing to anyone in cognitive science."
Of course, cognitive science is itself evolving, arguably becoming more relevant than ever. With the rise of intelligent technology comes greater need to understand both the nature of intelligence and of ourselves. In concluding our interview, Glushko noted:
"Our machines are getting increasingly smart, especially when they can learn. Cognitive science is essential in making them learn the right things and in interacting with each other and with people in ways that make human existence better."
So, about this time next year, you can expect to find another cohort of eager cognitive scientists congregating in some hotel lobby or reception room, waiting to learn the next winner of the Rumelhart Prize and hoping to play their own small part in the next cognitive revolution.
You can keep up with more of what Tania Lombrozo is thinking on Twitter: @TaniaLombrozo