|John Robert Anderson|
|Born||August 27, 1947|
University of British Columbia (B.A.) |
Stanford University (Ph.D.)
Intelligent tutoring systems|
Cognitive psychology (mathematics education)
|Institutions||Carnegie Mellon University|
|Thesis||A Stochastic model of sentence memory (1972)|
|Doctoral advisor||Gordon Bower|
Anderson obtained a B.A. from the University of British Columbia in 1968, and a Ph.D. in Psychology from Stanford in 1972. He became an assistant professor at Yale in 1972. He moved to the University of Michigan in 1973 as a Junior Fellow (and married Lynne Reder who was a graduate student there) and returned to Yale in 1976 with tenure. He was promoted to full professor at Yale in 1977 but moved to Carnegie Mellon University in 1978. From 1988 to 1989, he served as president of the Cognitive Science Society. He has elected to the American Academy of Arts and Sciences and the National Academy of Sciences and has received a series of awards:
In cognitive psychology, John Anderson is widely known for his cognitive architecture ACT-R and rational analysis. He has published many papers on cognitive psychology, including recent criticism of unjustified claims in mathematics education that lack experimental warrant and sometimes (in extreme cases) contradict known findings in cognitive psychology.
He was also an early leader in research on intelligent tutoring systems, such as cognitive tutors, and many of Anderson's former students, such as Kenneth Koedinger and Neil Heffernan, have become leaders in that area.
Anderson's research has used fMRI brain imaging to study how students learn with intelligent tutoring systems. Most of his studies have looked at neural processes of students while they are solving algebraic equations or proofs.
Anderson and colleagues generated a cognitive model that predicted that while students were learning an algebra proof, neuroimages showed decreased activation in a lateral inferior prefrontal region and a predefined fusiform region. This decrease in activity showed an increased fluency in retrieving declarative information, as students required less activity in these regions to solve the problems.
In a 2012 study, Anderson and Fincham (a Carnegie Mellon University colleague) conducted a study that looked at the cognitive stages participants engaged in when solving mathematical problems. These stages included encoding, planning, solving, and response. The study determined how much time participants spent in each problem solving stage when presented with a mathematical problem. Multi-voxel pattern recognition techniques and Hidden Markov models were used to determine participants' problem solving stages.
Results of the study showed that the time spent in the planning stage was dependent on the novelty of the problem. The time spent in the solving stage was dependent on the amount of computation required for the particular problem. Lastly, the time spent in the response stage was dependent on the complexity of the response required by the problem.
In another study, Anderson and colleagues used a video game task to test the Decomposition Hypothesis, or the idea that a complex cognitive task can be broken down into a set of information processing components. The combination of these components remains the same across different tasks. The study used a cognitive model that predicted behavioral and activation patterns for specific regions in the brain.
The predictions involved both tonic activation, which remained stable across brain regions during game play, and phasic activation, which was present only when there was resource competition. The study's results supported the Decomposition Hypothesis. Individual differences were also found in participants' learning gains, which indicated that learning a complex skill is dependent on cognitive capacity limits.'