Increasing the Effectiveness of Educational Technologies with the Use of Machine Learning Methods

Mariheida Cordova Sanchez, Purdue University

Abstract

There is a vast amount of educational technologies that assist students and instructors in their academic experiences. Technologies have been developed in the recent past to help students carry out discussions inside and outside of the classroom, allowing students to voice their thoughts and get their questions answered. Other technologies known as intelligent tutoring systems adapt to students’ abilities and provide personalized learning experiences. These technologies, while highly convenient, most of the time require a great amount of domain expert work to set up and maintain in order to be effective. In order to be most effective, educational technologies should i ) prioritize relevant content in order to reduce distractions in classrooms; ii ) systematically assist in labeling important contextual data; and iii ) provide mechanisms to give feedback to the learners to improve their learning experiences and improve graduation rates. We propose modeling classroom content data, in the form of discussions being held by students, intelligent tutoring systems’ questions, as well as students’ institutional data, such as their demographics and academic records, in order to improve student success. In this dissertation, we discuss the use of machine learning techniques to model and analyze this data gathered from educational technologies to feed back to the students and instructors, with the ultimate goal of improving the students’ academic experience and learning process. This dissertation studies whether we can effectively use machine learning methods to increase the effectiveness of educational technologies. Specifically, we study how machine learning methods can be used to i ) promote the most relevant and diverse questions or discussions held during classes in order to avoid distractions caused by irrelevant content and reduce the amount of time required to maintain; ii ) assist in labeling content for adaptive learning or intelligent tutoring systems in order to reduce the time required to set up and allow for a more effective functioning of the tutoring algorithms; and iii ) identify students who are at risk of not graduating from their current academic program, so as to be able to provide early warnings to students and increase retention and student success.

Degree

Ph.D.

Advisors

Dunsmore, Purdue University.

Subject Area

Computer science

COinS