Jen's ACT-R Page |
|
Relevant Links |
PrefaceI began studying Anderson and Lebiere's ACT-R 4.0 model of human central cognition in order to gain a better understanding of the reaction time data that specific ACT-R task models were generating. I was skeptical of modelers' claims about what this data proved. Computers are very good at producing numbers, I thought, but what these numbers signify is often another matter.In order to decide whether or not the modelers were justified in their claims about their models, I decided to investigate how ACT-R models were generating their data. In the process of pursuing this goal I have learned a great deal more about ACT-R. I have also had a number of thoughts about what computer models are and what they should be. As a result, this site contains a project in two parts. The first part is an essay about ACT-R. It asks the question- Can ACT-R be succesfully evaluated as a scientific model? In the process of trying to answer this question, I outline some modeling theory, discusses what is involved in computer modeling and put forth some criteria for making computer models evaluable. For the second part of my project, I have written a library of lisp functions that allow the working memory content of an ACT-R model (specifically, the state of global variables, procedural and declarative memory) to be recorded and saved at desired points during the running of the model. This effectively allows modelers to take 'snapshots' of ACT-R's memory as the model is running. These memory snapshots can then be later examined for particular changes or events that are of interest to the modeler or person trying to evaluate the model. I hope that this will be a step towards making ACT-R a more transparent and accesible model. |
|
Jen ACT-R Home Link Index |
|