Theses and Dissertations

An Algorithm for Identifying Episodes of Learned Helplessness in a Computer Assisted Learning Tool: A Structured Query Language Task

Date of Award

12-2021

Document Type

Dissertation

Degree Name

Ph.D.

Department

Computing

Committee Chair

J. Harold Pardue, Ph.D.

Abstract

In a world interrupted by a pandemic, the stresses created by rapid development of remote learning processes have reinforced the need for an effective proxy for direct observation of the learning process. Computer Assisted Learning (CAL) tools provide a comprehensive, remote, asynchronous learning environment. A limitation of these systems is the relative lack of real-time insight into the learner’s cognitive and emotional processes. What is needed is the ability to identify that a learner is struggling, frustrated, and in need of an intervention by the instructor. In this study, Learned Helplessness (LH) is used as a metric for frustration and as an indicator that a learner is struggling. LH is a phenomenon that consists of three parts: contingency, cognition, and behavior. The goal of this study is to develop an algorithm to identify, in real time, learner behaviors associated with LH. LH behavior is defined as the learner submitting a series of identical or nearly identical Structured Query Language (SQL) statements in response to receiving an error message from the CAL tool. Three types of errors are identified: syntax, SQL, and logical. The focus of this study is on logical errors because these types of errors benefit most from an intervention by an instructor. The algorithm was developed through quantitative and qualitative analysis of historical exam data from the CAL tool. The ix algorithm identifies an episode of LH as the point at which a learner submits 5 consecutive incorrect SQL statements where the average change in the SQL statements over a series of 5 submissions is less than or equal to 5 characters, or the 5/5/5 rule. Accuracy of the algorithm is defined as the degree to which the algorithm identifies learners who exhibit LH behaviors during an exam and subsequently self-report a tendency to LH as measured by two well established survey instruments. Initial results indicate an accuracy of 82%, that is, 82% of the learners identified by the algorithm as exhibiting LH also self-reported a tendency to LH. However, a qualitative review of the episodes of LH identified by the algorithm revealed that some of the episodes were the result of syntax or SQL errors, and not logical errors. When these episodes were removed, the accuracy of the algorithm increased to 92%. These results support the conclusion that an algorithm can be developed that alerts educators that a learner is struggling and in need of an intervention. Future directions for research include studying the impact of real-time instructor interventions and the impact of real-time tuning of the parameters of the 5/5/5 rule on learner success.

This document is currently not available here.

Share

COinS