Modes of learning in problem solving—Are they transferable to tutorial systems?

摘要

Complex systems and their features are described. Demands of complex systems are compared to well-defined problems. Deficient acting is discussed concerning the knowledge base and data processing during system control. There is evidence that goal-related knowledge is a precondition for adequate control strategies. Verbal feedback from one time interval to the next is assumed to hinder data integration during system control with respect to given goals. To test this assumption standardized, stable and unstable as well as individual graphical learning conditions are designed. Subjects have to generate and test hypotheses by means of these graphs. To test the effects of specific vs. global knowledge, naive and experienced subjects are provided with the standardized graphs. These learning conditions are compared in their effects on structural knowledge, performance, and strategies with a learning by doing condition. Structural knowledge is improved by all learning conditions. Unstable graphs alone cause a deliberate change in strategies, whereas unstable graphs combined with experience also cause an improvement in performance. Stable graphs support conservative behavior without effects on performance. Neither graphical feedback nor learning by doing are effective in improving strategies and performance. It can be shown that simple generating and testing of hypotheses with unstable graphs effectively reduces deficits in system control.