Generalizability of Predictive Performance Optimizer Predictions across Learning Task Type

Publication Year:
2016
Usage 42
Downloads 25
Abstract Views 17
Repository URL:
https://corescholar.libraries.wright.edu/etd_all/1549
Author(s):
Wilson, Haley Pace
Tags:
Department of Psychology; Training Optimization; Cognitive Model; Predictive Performance Optimizer; Task Type; Learning; Forgetting; Parameter Generalizability; Training Optimization Scheduling; Industrial and Organizational Psychology; Psychology; Social and Behavioral Sciences; Department of Psychology; Training Optimization; Cognitive Model; Predictive Performance Optimizer; Task Type; Learning; Forgetting; Parameter Generalizability; Training Optimization Scheduling
thesis / dissertation description
The purpose of my study is to understand the relationship of learning and forgetting rates estimated by a cognitive model at the level of the individual and overall task performance across similar learning tasks. Cognitive computational models are formal representations of theories that enable better understanding and prediction of dynamic human behavior in complex environments (Adner, Polos, Ryall, & Sorenson, 2009). The Predictive Performance Optimizer (PPO) is a cognitive model and training aid based in learning theory that tracks quantitative performance data and also makes predictions for future performance. It does so by estimating learning and decay rates for specific tasks and trainees. In this study, I used three learning tasks to assess individual performance and the model's potential to generalize parameters and retention interval predictions at the level of the individual and across similar-type tasks. The similar-type tasks were memory recall tasks and the different-type task was a spatial learning task. I hypothesized that the raw performance scores, PPO optimized parameter estimates, and PPO predictions for each individual would be similar for two learning tasks within the same type and different for the different type learning task. Fifty-eight participants completed four training sessions, each consisting of the three tasks. I used the PPO to assess performance on task, knowledge acquisition, learning, forgetting, and retention over time. Additionally, I tested PPO generalizability by assessing fit when PPO optimized parameters for one task were applied to another. Results showed similarities in performance, PPO optimization trends, and predicted performance trends across similar task types, and differences for the different type task. As hypothesized, the results for PPO parameter generalizability and overall performance predictions were less distinct. Outcomes of this study suggest potential differences in learning and retention based on task-type designation and potential generalizability of PPO by accounting for these differences. This decreases the requirements for individual performance data on a specific task to determine training optimization scheduling.