The goal of my research is to understand how people organize knowledge that they acquire from their diverse experiences to produce intelligent behavior. The concern is very much with how it is all put together and this has led to the focus on what are called "unified theories of cognition." A unified theory is a cognitive architecture that can perform in detail a full range of cognitive tasks. Our theory is called ACT-R (Anderson & Lebiere, 1998) and takes the form of a computer simulation which is capable of performing and learning from the same tasks that subjects in our laboratories work at.
ACT-R is also an instance of a hybrid cognitive architecture in that it represents knowledge symbolically as rules and facts but also has a neurally-based activation process that determines which facts and rules get deployed in which situations. We have engaged in extensive analyses of the situations which people have to deal with in order to understand how each of these components should work together to yield adaptive behavior.
Our research has two major branches. First, in the laboratory we are looking at how people learn and solve problems in very well-defined situations. Here we are interested in things like how strategies for problem-solving evolve, how people discover things about a new domain, how they deal with the working memory load imposed by the tasks, and how they get faster at accessing information relevant to task performance. Our subjects all interact with experiment-running computer programs and we try to develop ACT-R simulations that can interact with the same programs, take the same actions, make the same eye movements, and display the same latencies. The emphasis in this research is very much in getting the detail of the simulation to match up with the detail of the behavior.
The other branch of our research involves a much broader focus. We have taken on modeling the cognitive competences that are taught in the domains of mathematics, computer programming, and cognitive psychology. Much of the motivation for this research is to be able to tap into real situations where people learn and solve problems and understand the implications of these domains for the cognitive architecture. We have built larger-grain ACT-R simulations that are capable of solving problems in these domains and have developed computer-based instruction around these cognitive models. Many of these computer-based instructional systems have the cognitive models as a component and attempt to understand student behavior by actually simulating what the student is doing in real time. These are called cognitive tutors and are currently being used to help teach courses in schools around the country. Much of this research has gone beyond the original goals of understanding human cognition and now is part of a major effort to produce a significant improvement in American mathematics education.
Return John Anderson's Home Page
Return to the ACT-R Home Page