“I’m learning at a rate that has essentially stayed the same or increased over time. There’s always something new to learn and grapple with.”
Gautam Agarwal, Neuroscience PhD Program alum (entering class of 2003)
Ever since he was a young child, Gautam Agarwal has been fascinated by his own experience, which has led him on a continually evolving quest to understand the brain and mind. After graduating with BS degrees in Molecular Biology and Computer Science from the University of Texas at Austin, he entered the Berkeley Neuroscience PhD Program and joined the Isacoff lab, where he studied variations in synaptic strength. After graduating, he decided he wanted to pursue more theoretical and mathematical approaches towards understanding the brain, so he did a postdoctoral fellowship with Berkeley Neuroscience member Friedrich (Fritz) Sommer at the Redwood Center for Theoretical Neuroscience.
Agarwal then went on to do a postdoc with Zachary Mainen at the Champalimaud Center for the Unknown in Lisbon, so that he could delve into human behavior. In the Mainen lab, he developed a video game designed to gain a better understanding of human intelligence and its diversity by gathering data about how players solve complex problems. The game will be released as a citizen science mobile app in early 2020.
Agarwal recently returned to the Redwood Center, where he is now a research scientist in the Sommer lab, studying brainwaves in different behavioral contexts. He also works on the video game as a side project, and takes classes in dance and other body movement techniques as a way to explore human experience from an entirely different perspective.
Read the following Q&A with Agarwal to learn about his path from synapses to human intelligence and how his experiences at Berkeley helped him explore the mind from a variety of angles. This Q&A has been edited for brevity and clarity.
Rachel Henderson: When did you first become interested in science?
Gautam Agarwal: I think that was probably around the age of five. One of my earlier memories was getting chicken pox and somehow inferring or positing that it was being caused by some microbiological agent — to my mom’s surprise. I guess I was always really fascinated by my own experience, just why things were the way they were.
RH: As an undergrad, you did a double major in Molecular Biology and Computer Science — why that combination?
GA: Biology was always very appealing to me. I think because it was a way of merging all of my first-person questions of ‘how am I built?’ with a third-person perspective. I started at the molecular level because it seems like the more you zoom in, the more you get closer to whatever source underlies our architecture, in a sense.
Computer science was partially motivated by my dad. At the time, computer science was seen as a useful thing, but fairly restricted to its own domain without any very clear application to other fields. But he encouraged me to pursue it in undergrad and I did. I found that it was much more about how to reason through things and how to think about things, as opposed to biology which tended to be more of a series of observations, with the exception of a few professors who really took it to that other level of trying to understand from a more systems level.
RH: What brought you to the Helen Wills Neuroscience Institute (HWNI) to do a PhD?
GA: Over time, my interests shifted from understanding myself as a body to understanding myself as a mind, as it seemed to be somehow closer to the essence of who I was. So during undergrad, I shifted more towards neuroscience courses. There was this professor named Gerhard Werner who taught a course in neurocomputing in the engineering department at the University of Texas. [It] went beyond a lot of the textbooks that I had seen in neuroscience and revealed the depth of scholarship in the field. Like people in neuroscience really questioning: what does it mean to model a system and when do we feel like we’ve reached a level that constitutes understanding? So that really excited me and motivated me to pursue more in-depth research on it. I pretty much applied to every school that was on the West Coast that had a good neuroscience program. Berkeley had its own kind of charm for me at the time because of its unique cultural take on exploring experience. So when I got in, I jumped at the opportunity of coming here.
RH: What did you work on for your PhD thesis?
GA: As kind of a general through line to what I do, I’m very interested in methods that allow you to visualize the organization of any system of interest. Udi [Ehud Isacoff] has approached that through imaging where you can literally see, at some level, the structure of the system just by observing it using a microscope. I think one thing that excited me about his lab is that he engaged with this technique but was very general in its application. So he applied it to flies, zebrafish, looking at the connection between the muscle and the nerve, ion channels, synapses, and circuits. It was a place where I could engage with many different projects that happened at different levels of nervous system organization, each which kind of seemed to have its own idiosyncratic, highly complex, but highly specialized organization.
When I joined [the lab], one of the projects that I was involved with was to try to find any sort of rules that govern how strong a synapse would be. [The lab] had developed this imaging technique where you could look in the neuromuscular junction. The neuromuscular junction is cool because you have one pair of cells where there’s many contacts between the two cells. They developed a technique which allowed you to look at each of those contacts. We found that the contacts grew stronger as you went down the length of the axon. This was kind of strange, because why would one cell talking to another cell really care about having any sort of gradient of strength along its set of connections? Because the canonical way we think about synapses is basically ‘presynaptic excites postsynaptic’, without any sense of space or geometry associated with it. So we were kind of surprised when we found this conserved gradient of synaptic strength in this junction, and it made us think about why that existed.
I would say we never really figured out why the system chooses that kind of configuration. But it inspired me to look at the geometry of synaptic coupling and how strong synapses are in the central nervous system where space is much more at a premium, in a sense. You have all of these very complex arbors that are intertwining, and the geometry of those arbors influences how a cell is excited. [Also], the insect brain is remarkable because these insects can do seemingly complex things, but their brain is much smaller. So somehow all of the computational complexity of these systems is really compressed into these 100 micron-wide brain structures. I was interested in trying to extend what we’d found in the neuromuscular junction in[to] the central nervous system to try to image thousands of synapses in these tiny bundles using calcium imaging, and try to unpack to what degree there’s heterogeneity in how different cells talk to each other. Long story short, the calcium sensors at the time weren’t really good enough to ultimately get at what I was super interested in. Now I think, chances are, we would get much further with that question than we [did] at the time.
RH: What was your experience like as a student in HWNI?
GA: One of the things that I really loved about the program was our class. There were about three or four of us out of a class of 10 that I really connected with. We would spend a lot of time discussing the implications of what we learned and our prior knowledge and history, and it was a very tight community. And at the same time, being on a campus like Berkeley, I had the chance to sit in on philosophy of mind courses and psychology courses and physics [courses]. It was a time where I could really get a more bird’s-eye view of investigating the brain. Udi made it clear to me that really the purpose of my being there was to become a very good scientist, so more of an educational perspective, as opposed to a production perspective. That really stuck with me.
RH: What did you do after you graduated?
GA: When I graduated, I realized that the systems that we were trying to study and visualize, I was doing it, in a way, kind of intuitively. I didn’t really have a clear sense of what is the so-called ‘right way’ of approaching or modeling the complexity of a system. So I turned towards mathematics for more of a theoretical perspective. I eventually found the Redwood Center, which happened to be on campus. I started going to talks there and decided that would allow me to get a way of systematically approaching the modeling of systems. I eventually joined Fritz Sommer over there who was very patient with me, since my background wasn’t really suited, at the time, to work with him. He worked very closely with me to write a grant, which we were lucky enough to get, and so I continued working there.
During that time we figured out ways of deciphering what brain circuits are doing using, let’s say, coarser-grained measures of activity — in this case, local field potentials — to try to recover what is going on at the single neuron level. It turns out that these more coarse-grained measures, they seem to somehow preserve a lot of the informational detail of the fine-grained measures, so it has almost like a holographic quality to it. That got me really excited to move towards more noninvasive ways of measuring the complexity of human behavior.
Along this whole time, I was interested in using the precision of neuroscience techniques, but moving away from animal models more towards humans, and being able to use that precision in noninvasive ways. So after the Redwood [Center], I started looking at studying human behavior in precise ways. Not just human behavior, but human behavior in more complex everyday circumstances, which brought me into the video game realm.
RH: Yes, so you worked on a citizen science video game during your postdoc in Portugal — tell me about that.
GA: I guess, in my mind, there’s a huge gap between how we as scientists create environments where we can understand what our subjects are doing, and the environments that we experience on a daily basis [that] really bring up the challenges that the brain has to solve, which are well beyond anything that we test in a typical lab setting. I’m part of a subculture of scientists that’s very interested in pushing experimental protocols towards greater levels of complexity. Over time it kind of dawned on me that this is exactly the thing that games and video games achieve. Video games are experimentally appealing because they’re all done through a computer where you register every action that a player does, and you generate the stimulus that the player perceives. In a way, there’s a complete sensory-motor loop that you have complete control over.
I got really excited about that as being a medium where you could actually look at how our behavior is embedded in a stream. When you look at behavior in a stream, it brings up all these levels of complexity that aren’t present in momentary choices, for example, the so-called two-alternative forced choice task. Namely, when you’re in a sequence, you face the computational problem of the curse of dimensionality, where in order to figure out what to do now, you need to think of every step in the future. And that kind of unfolds into an exponentially growing tree of possibilities. So my vision at the time, and still to some degree, was to create an environment where people experienced that curse of dimensionality that really challenges their intelligence in this sequential setting, while being scientifically and analytically tractable.
I took many years to try to find what I would consider a suitable balance between complexity and simplicity. Complexity from the standpoint of experience and the potential of problem solving, but simplicity in terms of deconstructing the behaviors into something that can be modeled. I could be wrong, but I feel like we’re finally at a point where we’ve managed to balance the two. Kind of early on, about halfway through this five-year experience, we came up with a system, but it turned out to be too complex to model. Basically, every person had their own idiosyncratic way of solving the problem, at some level. Then at another level, they all found these similar shortcuts that really subverted the challenge that we’d posed to them. You might’ve heard the term ‘satisficing’ where people figure out ways of solving problems in manners that are far simpler, almost disappointingly simple, compared to the problem that’s posed to them because they can get away with it. You can often, in life, come up with solutions that save you a lot of time and energy and are much simpler than the full complexity of the problem.
So over time we pushed the game towards something that’s closer to chess. Games like chess are really interesting because they’re extremely complex, they’re very structured, but they have this kind of binary, win-it-all or lose-it-all quality that really pushes people to their limit. By creating a game where essentially the level of skill that you can attain has a very high ceiling, and how quickly you attain it is basically unbounded, we hope to present people with an immersive virtual experience that really motivates them to perform at their limits, and get distributions across different groups — like people who live in different places, who are different ages, [and] who have different personality types. Another interesting thing about cranking up the complexity knob is that it allows for many more ways of expressing yourself to solve a problem. That introduction of greater degrees of freedom is also something I’m really excited about to try to map the diversity of intelligent behavior, but still in a mathematically grounded way.
RH: What are you working on now that you are back at Berkeley?
GA: The video game project is still going, but at the moment, it’s kind of a side project. The reality of the kind of position that I’m in — which is now a research scientist — is that you’re essentially at the whims of whatever grants are there to support you. As I was finishing up my time [as a postdoc in Portugal], one opportunity that emerged was to come back here. Fritz [Sommer], my old mentor, got a grant to continue the work we’d been doing several years ago, so now I’m actually back working on a project trying to decipher brainwaves. But also, in parallel to my inspiration in Lisbon, looking at brainwaves in environments that are more complex to try to see how much can you decipher brain activity in these more free, unstructured environments.
RH: What is it like to be a research scientist, compared to being a postdoc or a faculty member?
GA: It’s kind of similar to my postdoc experience. Much of my work involves craning my neck over a computer for eight hours a day, [and that] mechanical aspect is probably [my] least favorite aspect of it. The thing that I really love about it is that it’s intellectually very stimulating. I would say I’m learning at a rate that has essentially stayed the same or increased over time. There’s always something new to learn and grapple with.
The thing that I’m kind of missing, is that as a postdoc and as a research scientist, much of my day-to-day activity is very detail oriented. As time goes on, I get more and more of a sense of how many different parts go into the brain bucket, in a way. There’s definitely a part of me that wishes that I could grapple with a lot of the questions of neuroscience at a higher level of abstraction. Often the reality is, as a postdoc or a research scientist, a lot of your time is spent debugging or really zooming into the minutiae of code. So that makes it somewhat of an interim position for me. I’m still looking for a medium where I can interact more with different groups of people, and at a higher level of neuroscience. That’s essentially this tradeoff between depth and breadth that I think many of us have to negotiate.
RH: What do you do outside of work for fun?
GA: I’ve been getting a lot into dance and both taking classes in improvisational dance and various kinds of body-centered techniques — one is called Feldenkrais. What I find really interesting about these is that they remind me about the beauty and the mystery of experience. Because as a neuroscientist, I’ve been trained to try to dissect these systems that I observe, but often when it comes to my own experience, my own emotional state or physical state, I am as irrational and myopic as anyone else. What’s really cool about these techniques for me is that people have developed them often just following their intuition as ways of systematically accessing and kind of modifying different thought patterns or different physiological patterns in the body. As well as from an artistic perspective, like dance practices. My wife is a dancer and after just sitting in a lot of dance performances — these peak experiences of being in a room where someone uses their body to present an experience that words cannot convey — it really kind of shakes you at this deep, felt level. Those are the kinds of things I like doing now.
- Gautam Agarwal, Redwood Center for Theoretical Neuroscience
- A podcast interview with Agarwal
- Humans of science profile of Agarwal
- View all alumni profiles