Computational
  Cognitive Science Group

Josh Tenenbaum

I study the computational basis of human learning and inference. Through a combination of mathematical modeling, computer simulation, and behavioral experiments, I try to uncover the logic behind our everyday inductive leaps: constructing perceptual representations, separating "style" and "content" in perception, learning concepts and words, judging similarity or representativeness, inferring causal connections, noticing coincidences, predicting the future. I approach these topics with a range of empirical methods -- primarily, behavioral testing of adults, children, and machines -- and formal tools -- drawn chiefly from Bayesian statistics and probability theory, but also from geometry, graph theory, and linear algebra. My work is driven by the complementary goals of trying to achieve a better understanding of human learning in computational terms and trying to build computational systems that come closer to the capacities of human learners.

Noah Goodman

i approach the study of mind with a combination of formal (mathematical) analysis, philosophical orientation, and empirical grounding. my research focusses on concepts and causality: what is the nature of causal and conceptual knowledge? how do we acquire this knowledge, and how do we use it?

David Wingate

My research interests lie at the intersection of perception, control and cognition, and how all three have synergistic effects on learning. Specific interests include reinforcement learning, unsupervised learning of useful knowledge representations (including predictive representations of state and structured nonparametric Bayesian distributions), information theory, manifold learning, kernel methods, massively parallel processing, visual perception, and optimal control.

Virginia Savova

My research centers on language as a symbolic communication system that makes infinite use of finite means. I believe that the question how such a system is represented and implemented in the brain is fundamental to cognitive science. In the past, I have employed different methods for studying this question -- from structural descriptions of syntactic phenomena, to Bayesian models and reaction-time experiments.

Amy Perfors

I'm interested in applying Bayesian models to aspects of cognitive development, in particular to issues of learnability. What biases must children have in order to acquire knowledge in different domains (syntax, word and feature learning, understanding of kinds)? To what extent are these biases domain-general?

Lauren Schmidt

I'm interested in how people learn the meanings of words and infer relationships between words or between concepts. I'm also interested in how learning language can influence conceptual structure and development. One of my recent projects looks at how people can understand what is sensible (but possible, extremely rare, or not occurring in nature, like a blue banana) and what is nonsense (like an hour-long banana) based on the limited evidence of what actually occurs in the world.

Chris Baker

I'm broadly interested in social cognition and more specifically in theory of mind. Social cognition describes how people reason about interpersonal situations and interact with other people; theory of mind describes people's reasoning about the mental states of others, such as beliefs, desires and emotions. My research uses insights from psychology and philosophy to guide the development of computational models of people's intuitive theories of other people. These models are implemented with technology from artificial intelligence and machine learning, and tested empirically using behavioral experiments.

Liz Bonawitz

I started as Josh Tenenbaum's lab coordinator in July of 2002. Now I'm a graduate student at MIT working jointly with Professor Laura Schulz in the Early Childhood Cognition Lab and with Professor Josh Tenenbaum. I'm interested in the development of human causal reasoning from infancy to adulthood and in the computational models that may shed insight on that process.

Vikash Mansinghka

My research lies at the junction of induction, computation and cognitive modeling. I try to make computers smarter by framing computation in terms of inference under uncertainty and leverage this perspective to understand human cognition and behavior. My research draws on tools from Bayesian statistics, computational statistical mechanics and formal approaches to knowledge representation (both procedural - e.g. Scheme - and declarative - e.g. graphs, grammars and logics). I work on theoretical and applied questions involving languages for inductive computing, compilers that generate Monte Carlo algorithms for inference, machines that natively execute these algorithms, and models built using these tools that find patterns and make actionable predictions. I am also involved in an ongoing effort to commercially deploy this research, focusing on problems of statistical inference and nonlinear optimization.

I am interested in the computational properties of systems which represent and process natural language. In particular, linguistic research has revealed a large number of empirical phenomena which seem to be "structural" in nature -- that is they reflect (perhaps idiosyncratic) underlying properties of natural language computation rather than following from other design features (such as function). I believe that the Bayesian perspective offers important new insights into these types of phenomena by allowing us to ask whether such properties really are arbitrary and/or specific to the computational system of natural language or whether they follow from the principles of rational inference combined with rich and abstract, but perhaps domain-general representations.

Yarden Katz

I'm interested in the computational principles that underlie high-level cognition, and how these might be implemented in the brain. I'd like to explore how ideas from Bayesian statistics and machine learning could help us interpret and systematically analyze neural data (such as fMRI), and how such analysis could in turn inform and constrain the computational models we build.

Brian Milch

I'm interested in understanding how anything made of non-intelligent parts could behave as intelligently as a human being. More specifically, I develop models and inference algorithms that combine the ability of probability theory to quantify uncertainty, with the ability of first-order logic to efficiently describe large sets of related objects. I'm a post-doc in Leslie Kaelbling's group at CSAIL, but I also collaborate with Josh Tenenbaum's group.

Dan Roy

I work on statistical learning algorithms inspired by cognitive theories of categorization (usually expressed as non-parametric Bayesian models), the design of scalable, approximate inference algorithms for these (intractable) models, and the application of these models to both machine learning problems and cognitive data.

Michael Frank

In order to communicate successfully, children acquiring a language have to learn to segment words from continuous speech, learn the meanings of those words, and figure out how to put them together to make coherent sentences. I'm interested in all three of these problems, and I study them using artificial language learning experiments with adults and infants as well as probabilistic computational models. I work jointly with Josh Tenenbaum and Ted Gibson.

Ed Vul

Without constraints and assumptions, it is impossible to figure out what sorts of stuff in the physical world caused our retinal input. I am primarily interested in the priors and structures our visual system uses to to solve this problem given limited resources. To this end, I study adaptation, attention, and other visual processes with psychophysics, computational methods, and fMRI.

Kobi Gal

How can computers learn to make good decisions in groups comprising both people and other computer agents? I study this question by developing representations and algorithms for learning the social factors that affect people's decision-making in a variety of domains, such as negotiation, intelligent tutors and game playing.

Charles Kemp

At some level, semantic representations are mathematical objects. I'm interested in finding a small set of structures and operations that can be composed to build these objects. I also enjoy thinking about formal models of social systems.

I am interested in the kinds of things that make people seem clever -- especially the ability to make robust, flexible, & reliable inferences from limited data. Particular areas of recent interest are learning multiple ways of organizing knowledge in a domain, flexible use of background knowledge to support inferences in different contexts, and learning in pedagogical & communicative settings. I try to understand these abilities through formal mathematical analyses and behavioral experiments. The goal of my research is to try to develop an understanding of people's "real-world" reasoning.

I am interested in using cognitive models of human belief and causality reasoning to understand psychological phenomena such as base rate neglect. The ultimate goal of my research would be not just a computational model of human cognition, but also a method for developing AI systems that think like people. I got my PhD in BCS in 2006.

I did computational and cognitive neuroscience in the group of Josh Tenenbaum, BCS, MIT. (previously with Daniel Wolpert and Peter Konig). I specialize in modelling and movement psychophysics. I expect my theories to make experimental predictions, provide a compact description of data and lead to computationally strong algorithms. My experiments should falsify theories.

My research interests are developing computational models of higher level cognition. In particular, I'm interested in developing rational accounts of cognition using probabilistic generative models and Bayesian statistics. My current areas of interest are understanding people's everyday inductive leaps - difficult inductive problems we solve every day, like predicting the future, learning causal relationships, and noticing coincidences - and the interface between psychology and machine learning in developing statistical models of language.

My research interests span a diverse set of topics in cognitive science such as episodic and semantic memory, dynamic decision making, and causal reasoning. In each of these areas, I combine mathematical and computational modeling with behavioral experiments. The models and experiments are tightly coupled: I try to formulate empirical questions with the goals of constraining, developing, or testing between alternative computational models of how people learn, process, and represent information. My research interests also include some computer science topics in the domain of statistical machine learning and information retrieval. The adoption of recent machine learning methodology is useful in advancing cognitive science research, especially in the area of semantic memory.

I'm interested in some of the conceptual groundwork that might someday make psychology more coherent -- how we (or anything) can weave a web of concepts to explain experience. More technically, I'm interested in how to extend the range of probabilistic models of induction to richer descriptions of experience than those available to traditional probabilistic models. I'm also interested in the form of the associations of words with conceptual formulae, how we learn those associations, and how we use them to invoke thoughts in each other and ourselves.

I study the neural and psychological basis of social cognition. Do we have dedicated mechanisms for recognising and/or reasoning about other minds? How and why does the human brain succeed so easily where computers and logicians fail? Addressing these questions, my work spans the disciplines of cognitive neuroscience, developmental psychology, social psychology, computational modelling and philosophy.

My research in the Tenenbaum lab is focused on exploring computational models of how humans learn and generalize through inductive inference. My undergraduate honors thesis analyzed how a computer might construct a hypothesis space (i.e. candidate guesses about the concept to be learned) that could match human generalization performance. In that work, I examined several different unsupervised learning techniques to build a hypothesis space for a Bayesian concept learner. Recently, I have been looking at how humans seem to be using, on an abstract level, taxonomic, tree-based models of similarity to guide their generalization. In my other lab, I am working on understanding the dynamics underlying neural computations in small networks of neurons.

I graduated from the Brain and Cognitive Science Department at MIT doing a UROP with Professor Tenenbaum. My personal research interests for the future include modeling human social cognition.

I graduated from the Brain and Cognitive Science Department at MIT doing a UROP with Professor Tenenbaum. I am primarily interested in concept learning and conceptual change during learning and development.

I graduated from Brain and Cog Sci, only to stick around MIT in the Speech and Hearing Bioscience and Technology program. (I'm a cognitive scientist at heart, of course.) I'm interested in language -- both its evolution as a communication system and its purpose as an internal evoker of mental representations -- and how speech sounds might act as the elementary units of those representations.

I graduated MIT (majoring in BCS & minoring in music.) My research interest lies in searching for the fundamental ways humans generalize and understanding how they interpret novel ideas.