A NOVEL TRAINING PARADIGM FOR KNOWLEDGE AND SKILLS ACQUISITION: HYBRID SCHEDULES LEAD TO BETTER LEARNING FOR SOME BUT NOT ALL TASKS

Open Access
Author:
Paik, Jaehyon
Graduate Program:
Industrial Engineering
Degree:
Doctor of Philosophy
Document Type:
Dissertation
Date of Defense:
April 29, 2011
Committee Members:
  • Frank Edward Ritter, Dissertation Advisor
  • Frank Edward Ritter, Committee Chair
  • David Arthur Nembhard, Committee Chair
  • Andris Freivalds, Committee Member
  • Ling Rothrock, Committee Member
  • Darrell Velegol, Committee Member
Keywords:
  • Cognitive Modeling
  • Effective Training Schedules
Abstract:
Studies on effective practice schedules have been generally investigated by comparing the performance of two relative extreme practice schedules, distributed and massed, at a retention test. The results of most of these studies have consistently shown that distributed practice schedules result in higher retention rates than massed practice schedules because of the spacing effect associated with human memory. These studies, however, failed to show either the optimal interval between learning sessions, or to consider the knowledge types used to perform tasks. Furthermore, these studies did not provide any theoretical supports except the spacing effect for predicting performance at the specific time of the schedules. To address these problems, I explored both theoretically and empirically. First, I investigated ACT-R’s learning and forgetting theories to help identify an optimal practice schedule. Second, psychological experiments were conducted to validate these theories. In the experiments, four kinds of tasks were tested using four practice schedules (distributed, massed, Hybrid1, and Hybrid2). Finally, models were developed using ACT-R, and then compared with empirical data. ACT-R’s learning theories suggest that hybrid practice schedules (schedules consisting of distributed and massed practice) could produce better performance than an exclusively distributed practice schedule. The results of empirical data indicated a more complex picture. Like previous studies, the results of experiments showed a higher correlation between retention and distributed practice schedules than retention and massed practice schedules. There were, however, no significant differences between distributed and hybrid practice schedules when testing a declarative memory task. Nevertheless, the results also suggested that some hybrid schedules might produce better performance than distributed practice schedules for perceptual-motor skills. Specifically, my results show that the Hybrid1 practice schedule produced greater skill retention than the distributed practice schedule for the perceptual-motor task, indicating that the spacing of distributed and massed practiced with respect to each other also influences performance. When comparing the ACT-R models’ performance with that of the participants’, the models could predict the learning and forgetting trends of the participants in each group for declarative memory tasks; however, there were differences in the correct responses between the models’ prediction and the human data. These results indicated that ACT-R could be used to predict the learning and forgetting trends of practice schedules, however, revisions might be necessary to fully map the models’ predictions to the participants’ specific responses.