Research in the Computational Cognitive Science lab focuses of development of computational models of learning combining structured knowledge representations with Bayesian inferential machinery. We draw upon inferential tools from Bayesian statistics, representational tools from Machine Learning and AI, and behavioral research methods from Cognitive Science.

Three specific areas of recent research in the lab, pedagogical learning, flexible learning and reasoning, reasoning in real-world contexts, are explained in more detail:

Pedagogical Learning

Much of human learning goes on in pedagogical settings -- settings where there is a teacher who is trying to communicate an idea by purposefully choosing data. Typical approaches to learning make relatively weak assumptions about how data are sampled (e.g. randomly). These assumptions clearly do not apply in pedagogical situations. Further, even young children understand that pedogogical situations warrant different kinds of inferences. The goal of this research is to develop and test a computational model of the assumptions that go into pedagogical reasoning. 

Flexible Learning & Reasoning

Flexibility is one of the most remarkable aspects of human reasoning. Subtle changes in context lead to qualitative shifts in reasoning. Knowledge is combined in novel ways to support predictions about never-before-seen scenarios. How is the flexibility achieved? This research attempts to identify when and why people learn multiple ways of thinking about domains, and how these different kinds of knowledge can be combined to support inductive inference. In addition to psychological applications, these models are published in the machine learning literature as novel algorithmic approaches to learning, and have potential applications in data mining.

Real-World Reasoning

Perhaps the most impressive aspect of human reasoning is that it works in the real-world. In one of the more ambitious strains of research, the lab is attempting to model aspects human reasoning in in real-world scenarios. One such project seeks to model reasoning about mental illness by including modeling the causal knowledge that people bring to the reasoning task.