Learning and revising task-specific rules in ACT-R
Niels Anne Taatgen, U Schmid, J Krems, F Wysotzky · 1996
This paper discusses the problem of how to learn and revise rules for a new task. It will use the framework of the ACT-R theory, which uses analogy to learn new productions. As an example, two general rules to learn new productions will be discussed that can learn two different tasks, a beam-task and a card-classification task. 1 Introduction If, in an experimental situation, some new task is presented to a subject, he or she is, after brief instructions, almost always capable of doing the task. This means that in a short period of time the subject has acquired the necessary task-specific knowledge to be able to get started on the task. Cognitive architectures based on production rules often ignore this phase of the experiment and just assume that subjects somehow produce this knowledge while reading instructions or studying an example. When the task involves complex problem-solving, the set of rules the subject initially comes up with is often insufficient to reach the goal. In that ...