Carnegie Mellon University

Building Better Brains


A Q&A with Michael Tarr and Nathan Urban

Urban and Tarr
From left are Nathan Urban, head of the Department of Biological Sciences,
Michael J. Tarr, co-director of the Center for the Neural Basis of Cognition (CNBC).

Michael J. Tarr, co-director of the Center for the Neural Basis of Cognition (CNBC), and Nathan Urban, head of the Department of Biological Sciences, will be honored on Jan. 20 as part of a celebration and launch of a new strategic research initiative for CMU — brain, mind and learning sciences.

Tarr has received the George A. and Helen Dunham Cowan Professorship in Cognitive Psychology, and Urban has received the Dr. Frederick A. Schwertz Distinguished Professorship in Life Sciences. The two will participate in a panel discussion titled "How To Build a Better Brain." Moderated by CMU Executive Vice President and Provost Mark Kamlet, the panel also will include Justine Cassell, director of CMU's Human-Computer Interaction Institute, and Marcel Just, the D.O. Hebb Professor of Psychology who is an expert on the architecture of human thought.

The hiring of Tarr as co-director of CNBC and the promotion of Urban to head of biological sciences has provided major new momentum to this research thrust, and is but one sign that Carnegie Mellon is committed to enhancing its leadership position in this field.

Tarr and Urban talk here about why brain, mind and learning research is different at Carnegie Mellon, and about where they see the university headed.

On Jan. 20, the university will host a panel discussion called "How To Build a Better Brain." What can you tell us about the event?

The event will highlight and celebrate some of the great things going on in brain, mind and learning sciences here at CMU. Building Better Brains will bring people together to talk and learn about the exciting future of this research.

TARR: Carnegie Mellon has a unique approach to these topics that leverages its focus on world-class computational programs and real-world applications to develop new technologies that bear on fundamental questions about the brain, the mind and learning.

What does Carnegie Mellon have to offer on brain science research that distinguishes itself from other universities?

Brain science at CMU is extremely interdisciplinary with a strong focus on computation and real-world applications. It's very common at CMU for experimentalists coming from biology or psychology to work closely with statisticians or computer scientists or engineers. This can result in computational scientists becoming interested in knowing more about how the brain solves many different kinds of problems — for example, in object or speech recognition — and also in neuroscientists and cognitive neuroscientists becoming interested in knowing more about the computational principles that underlie solutions to such real-world problems.

TARR: Consider that inventors have always used the natural world for inspiration. DaVinci and others tried to make bird-like wings to fly. Although such simple mimicry does not always lead directly to a solution, it often does point to components of the eventual solution. Studying how the brain performs tasks, such as visually recognizing faces and objects, decoding the auditory stream into different speakers and sounds, and recalling past events may all provide insights into how to build devices to solve these problems.

URBAN: At the same time, these insights may provide enormous benefits both in helping repair or treat brain dysfunctions, including autism, dyslexia and Alzheimer's, and in facilitating more effective educational models.

TARR: For example, CMU researchers are applying cognitive theory and cognitive modeling to identify the critical instructional conditions that facilitate effective student learning.

How do the Psychology and Biology departments work together in brain science?

In my opinion, Biology and Psychology are extremely complementary, although people in these fields are often satisfied with rather different kinds of answers to the same questions. Biologists tend to be strongly reductionist in that they like to explain things in terms of the behavior of cells or even molecules. That being said, the ability to image the detailed structure and function of the human brain — for example, using functional Magnetic Resonance Imaging (fMRI) — is providing more and more of a link between the biological and psychological approaches.

Because CMU is blessed with a highly interactive and multi-disciplinary faculty — particularly as exemplified and facilitated by the Center for the Neural Basis of Cognition (CNBC) — these links are both meaningful and productive. Researchers anchored in specialties as diverse as machine learning to cellular neuroscience (and everything in between) work together at CMU, forging new ways to look at old problems on the brain, mind and learning.

What are some of the other departments on campus that are involved in similar research areas?

URBAN: Work in the brain, mind and learning sciences spans an incredible number of departments and centers across CMU. These include departments in H&SS (psychology, social and decision sciences, statistics and philosophy), in the Mellon College of Science (biological sciences), in the School of Computer Science (machine learning, computer science, robotics, human-computer interaction) and in the College of Engineering (biomedical engineering).

Cross-college centers and programs include the CNBC, the new Scientific Imaging and Brain Research Center, the Pittsburgh Science of Learning Center and the Open Learning Initiative.

If we understand how the brain is wiring itself, how do we apply that knowledge to benefit people's health and well-being?

Understanding the fundamental principles of human brain function and behavior will yield both unexpected and high impact gains in how we treat developmental diseases, mental health disorders, and brain dysfunction and injuries. Some of the effects of this knowledge will manifest as preventions or treatments that reverse disorders or dysfunction, but the largest impact will be felt in both neuropharmocological and behavioral interventions.

TARR: We will also see significant advances in how we exploit our understanding of brain function in the realm of education — there seems little question that a better model of how children's brains change over development will give rise to improved tools for educators at all levels.

Where is the field of brain science and learning going in the future?

As our understanding grows and our technologies advance, models and datasets are both becoming more and more complex. The only way to address these challenges is through computational thinking. As the future brings us more powerful and better integrated models of brain and behavior, computation will become intrinsic to the field. At CMU we are trying to train the next generation of brain scientists who will see such an integrative approach as the natural way to do things.

What are your primary areas of research?

Work in my lab is focused on questions about how our visual system processes, perceives and recognizes faces and objects. To do this we use a variety of human neuroimaging methods, the most prominent being fMRI. We are trying to unravel how we are able to effortlessly identify visual objects across widely variable conditions that can radically alter the appearance of an object. One of my key interests within this area has been the changes that occur in our brain as we learn a new object category, both upon initial learning and after a great deal of experience. This question is also central for understanding how we acquire our incredible proficiency at recognizing individual faces. As I move into my third decade of research, I have come to realize that only a truly interdisciplinary approach involving computation, psychology, and neuroscience — as exemplified by the CMU community — has any hope of providing meaningful answers to these questions.

URBAN: Work in my lab is focused on questions about how brains use biological hardware to process information. To do this we record from neurons, most often in vitro, and try to understand how the properties of neurons and their connections allow the brain to process information. Most of our work is in the olfactory system. Working in a sensory system allows us to make inferences about how the properties that we study in vitro are important for processing real sensory stimuli. Often we find it helpful to build computer models of neurons and circuits in order to make the connection between our experimental data and the kinds of processing that we believe goes on during behavior. 

Why did you choose those topics?

: I grew up and hung around CMU and was strongly influenced by its prominence in both cognitive science and computer science. I went off to college figuring I would study artificial intelligence (AI), build the first "thinking" computer and the rest would be history. Somehow the rest of the country wasn't where CMU was in studying AI, so I drifted into cognitive science, where at least the questions were similar to those with which I was fascinated. In graduate school I tried to bridge the gap between what I had learned as a cognitive scientist and what was going on in computer science, vision being one topic where there was some hope of meaningful connections. Although building such connections was and is difficult, I have continued to keep this goal in sight throughout my research career: coming back to CMU offers a natural way for me to pursue these links in a serious way.

URBAN: I started off doing my Ph.D. in an experimental lab thinking that I wanted to learn how experiments were done so that when I went and built models they would not be completely unrealistic. My primary goal when I went to graduate school was to do computational neuroscience. I discovered, much to my surprise, that I was not terrible at all kinds of experiments, only some kinds. So my Ph.D. and postdoctoral work were in neurophysiology, trying to understand how single neurons and single synapses (connections between neurons) work. For me this work was always motivated by thinking about neurons and circuits as computational devices. Questions like, "What's the most complicated mathematical operation that a single neuron can perform?" provided motivation for me.  

Anything else we should know about?

The study of the mind and brain is the last frontier of science. Sometimes I envy my physics or chemistry colleagues in the depth to which they truly understand the systems they study. Then I remember it is (arguably) much more exciting to work in an area where most of the interesting stuff remains undiscovered.

Panel To Focus on
Interdisciplinary Brain
Research at CMU Jan. 20

What: How To Build a Better Brain
When: 5 p.m., Thursday, Jan. 20
Where: Rashid Auditorium

The panel discussion will be moderated by Mark Kamlet, provost and executive voice president of the university. Additional members of the panel include:

Justine Cassell, director of the Human-Computer Interaction Institute

Cassell's research focuses on deconstructing the linguistic elements of conversation and storytelling in such a way as to embody machines with the conversational, social and narrative intelligence they need to interact with humans in human-like ways. It also addresses the impact and benefits of technologies such as these on learning and communication.

Marcel Just, director of the Scientific Imaging & Brain Research Center and D.O. Hebb Professor of Psychology

Just's research uses brain imaging (fMRI) in language and perception tasks to study the neural basis of human thought. The projects examine normal cognitive functioning in college students and in adolescents, as well as in special populations, such as patients with autism and children with dyslexia

This article first ran in the December 2010 issue of the Piper.

Jocelyn Duffy and Shilo Raube