What is your current position and area of research?
I am a Chancellor’s Professor of Psychological and Brain Sciences at Indiana University, and my primary areas of research are learning, applications of cognitive science to education, and collective behavior. For all three areas, my laboratory engages in a two-pronged research strategy of obtaining empirical data on human cognition and developing computational models of the empirical data.
What led you to this field/area of research?
For me, building working computational models is a satisfying way of developing theories that explain how people learn and think. Building a computational model forces researchers to be precise in their theorizing because, for better or worse, computers only do what they are told to do, even though one of the most interesting things that they can be told to do is learn from their world on their own. As our psychological theories become more mature and complete, it will become increasingly important to build working computational and robotic models that implement psychological theories to make sure that all of the elements of theories work together in the intended manner. Plus, once constructed to explain human behavior, a computational model will sometimes end up be a useful product on its own, allowing machines to act in more flexible ways and better predict the wishes, limitations, and intellectual potential of the humans they are built to support.
What in your opinion has been the greatest achievement in your area of science?
One of the foundational ideas in cognitive science is that people create mental models to explain and predict their world. If we want to understand how people think, we need to understand their models. To help them think better, we need to give them better models. For example, two common but incompatible models for home heat control are the “valve model” and the “feedback model.” According to the valve model, the temperature that the thermostat is set at determines how hard the furnace works to produce heat – higher temperatures make the furnace run harder, much like depressing a gas pedal lower on a car makes the engine rev more. According to the feedback model, the thermostat sets the threshold below which the furnace turns on, but the furnace runs at a constant rate. These different models drive very different home heating behaviors. If two people come home from vacation to a 13 degree Celsius home and would like it to be 20 degrees, the valve theorist might set the thermostat to 26 degrees because they want the house to warm up quickly, whereas the feedback theorist would set it to exactly 20, realizing that setting the thermostat higher than 20 will not make the house warm up to 20 any faster. Across America, millions of dollars of energy are wasted every year because of inefficient temperature oscillations produced by the inaccurate valve model. It is estimated that 25-50% of Americans have this faulty model. We can dramatically improve learning and reasoning by giving people the right models at the right times, and even more fundamentally, by giving them the wherewithal to create and critique their own models.
Where do you see your field progressing over the next 10 years?
One exciting development in cognitive science has been the increased use of large, naturally occurring data sets for revealing how people think. Over the last hundred years, psychologists have become very sophisticated in their experimental designs, controls, counter-balancings, and novel laboratory methods, but this sophistication may have had the outcome of blinding us to the possibilities of discovering principles of behavior without conducting experiments. When creatively interrogated, a diverse range of large, real-world data sets provides powerful diagnostic tools for revealing principles of human judgment, perception, categorization, decision making, language use, inference, problem solving, and representation. Examples of these data sets include web site linkages, dictionaries, logs of group interactions, image statistics, large corpora of texts, history of financial transactions, photograph repository tags and contents, trends in twitter tag usage and propagation, patent use, consumer product sales, performance in high-stakes sporting events, dialect maps over time, and scientific citations.
Another area of rapid advance in cognitive science is the use of machine learning to understand and promote human learning. Progress in algorithms that allow machines to learn how to recognize patterns has had a truly transformative impact for building automatic and semi-automatic systems that can label the objects in a photograph, transcribe speech, play world-class games of Chess and Go, detect internet fraud, diagnose diseases, and make apt recommendations for books, movies, restaurants, articles, and personal contacts. There are strongly suggestive parallels in how humans and deep learning systems learn, both in terms of commonalities but also ways of improving learning in each by considering the strengths and weaknesses of the other. The next generation of deep learning systems will be even more effective than the current crop because they will emulate methods that people use to learn, and these more capable machines will return the favor by better predicting and facilitating human learning.
What is the best part of being on the board?
I have learned so much about the ways in which our understanding of behavior is being actively applied to better the human condition, including initiatives for improving: the decision making of citizens in health, education, retirement, and financial arenas; the effectiveness of individuals and teams; treatments for addiction; and the social, cognitive, and emotional well-being of people across the lifespan. After being on the board for several years now, it is hard not to be inspired by the myriad ways in which the behavioral, cognitive, and sensory sciences have made substantive contributions to the understanding and advancement of mental lives.
If you could meet anyone from history who would it be and why?
I would like to meet with Alhazen (in Arabic, Ibn al-Haytham), one of the progenitors of modern vision science. He passed away in 1040 A.D., 300 years before the Renaissance scientists of the 14th century, and yet he anticipated their scientific methods with his own focus on the power of rigorous, controlled experiments to resolve theoretical questions. Back when thinkers were arguing about whether vision worked by emission (the eye emitting rays of light at objects to be seen, championed by Euclid and Ptolemy) or intromission (rays of light going from objects to the eye, championed by Aristotle), Alhazen used empirical considerations and experiments to decisively settle the question in favor of intromission. In addition to seminal contributions to astronomy, geometry, number theory, and philosophy, Alhazen’s work on optics was a systematic exploration of vision, perception, and illusion. I would enjoy chatting about perception with him from very different historical and cultural perspectives. I think he would get a kick out of seeing how detailed is our modern understanding of the neural and mathematical underpinnings of vision.
What is the greatest book you ever read and why?
Like many active cognitive scientists of my generation, I was inspired to learn about the fledgling field by virtue of reading Gödel, Escher, Bach: An Eternal Golden Braid, by Douglas Hofstadter, now my colleague and friend at Indiana University. The book sparked the imagination of my high school self to think mechanistically about how a self could emerge from the activities of “clueless” neurons that are doing nothing more than activating and inhibiting other neurons. With humor, analogies, parables, visualizations, puzzles, and dramatic dialogs, the book opened up a world to me in which computer science, psychology, mathematics, philosophy, biology, and even the arts are combined to reveal how it is possible that minds exist.