Despite remarkable recent progress, current machine learning and A.I. systems are limited and inefficient compared to the intelligence of even a young child. Notably, most current machine learning systems are trained with extensive pre-selected, curated data and corrective feedback. In contrast, children are remarkable in their ability to bootstrap knowledge from their independent, self-guided interactions with the world. They ask questions, curiously explore new toys, and push at the boundaries of what they currently know on their own, often without a teacher or parent. What are children born knowing about how to learn that our current engineering fields haven’t yet grasped?
For the last decade or so, I have focused the main thread of my research program around understanding the nature of this human ability in computational terms. How is it possible to learn without a teacher (unsupervised learning)? How are we able to figure out how something works by tinkering with it? How do we formulate and ask creative questions that reveal useful information about abstract concepts? How do we judge what actions we can take to obtain useful information? How do our choices to gather information affect our learning and memory? How do these abilities develop across the lifespan? I believe such questions strike at the heart of what makes us such an adaptable and intelligent species.
My research approach primarily aligns with the field of Computational Cognitive Science, which aims to understand human intelligence in computational terms (i.e., precisely describing the algorithms the mind uses to perform acts of intelligence). As a result, I try to build intellectual bridges to aligned fields such as data science, computer science, education, and neuroscience. I am a devoted and purposeful computational ployglot – research in my lab frequently combines methods such as neural networks, Bayesian methods, causal learning, reinforcement learning, program induction, and symbolic systems.
In recent years I have branched out from this core theme of active/self-directed learning to consider research on decision making, the cognitive neuroscience of memory, virtual reality, and how we reason and interact with the physical and social world. I am committed to help develop open source software to enable new types of high-quality cognitive research. I am also interested in the idea of using “very very big experiments” to explore and test theories of cognition.
I spent my childhood in the dusty North Texas oil fields playing too much soccer. My parents offered me a used car in high school but I upsold them on buying a computer instead, which opened the door to a career for me. I didn’t want to go to college because I was working for a small internet company (which eventually became part of Rackspace). However, my boss at the time said he’d fire me if I didn’t (which required me to quit anyway). I applied to one college, the University of Texas at Austin, in no small part because of the music scene, and stayed for my PhD because my advisor and I liked similar music (and he had the patience to teach me a lot of basics). After grad school I applied to one postdoc, at Indiana University, because there was no other place with a comparable concentration of mathematical psychologists interested in higher cognition. Ever since I’ve been at New York University carying the torch of computational cognitive science, running, surfing, and rocking climbing over everything. Here is a more formal bio.