But does it really do that? Using formal analysis to ensure desirable ACT-R model behaviour.
- Vincent Langenfeld, Department of Computer Science, Albert-Ludwigs-Universitat Freiburg, Freiburg, Germany
- Bernd Westphal, Department of Computer Science, Albert-Ludwigs-Universitat Freiburg, Freiburg, Germany
- Rebecca Albrecht, Department of Psychology, University of Basel, Basel, Switzerland
- Andreas Podelski, Department of Computer Science, Albert-Ludwigs-Universitat Freiburg, Freiburg, Germany
AbstractCognitive modelling uses computer models to investigate psychological theories. To conclude from executions of a cognitive model to properties of the theory, the model needs to be a correct implementation of the theory since a defective cognitive model may yield wrong statistical figures. We present three common reasons for a model to be incorrect wrt. a theory: situations which unintentionally do not enable any production rule, rules which erroneously construct undesired declarative knowledge, or wrongly chosen architecture parameters. Unfortunately, defects of these kinds are hard to spot; repeated execution and observation of the model does not guarantee to uncover such defects. In this work, we give formal definitions of these defects, in terms of an existing abstract formal semantics of the hybrid architecture ACT-R. We demonstrate the application of formal analysis techniques to ACT-R models to ensure that the model behaves according to the psychological theory.
Return to previous page