Understanding student behaviors using immediate feedback features in a blended learning environment

Xin Chen, Purdue University

Abstract

Feedback serves to close the gap between learners’ current understanding and the desired understanding. Informative feedback can keep students from holding onto misconceptions, actively engage learners in knowledge acquisition, and increase confidence and motivation to learn. Yet, in the context of higher education, it is usually not possible for instructors to provide timely feedback to every individual student. This is especially difficult in first-year foundational courses due to the large number of students. Online learning platforms offer a solution by providing students immediate feedback during the course of their interactions with formative assessment tools (e.g., online homework, quizzes, embedded questions in lecture videos). However, how students choose to interact with these features and how these features influence students’ learning experiences have not been well understood. Even less is known about student behaviors with these immediate feedback features in a blended learning class. Fortunately, there is now a mechanism to address these questions since the ever-increasing usage of online learning platforms such as edX (www.edx.org), Coursera (www.coursera.org), and FutureLearn (www.futurelearn.com) enables the detailed recordings of student activities (e.g., usage logs, message streams, mobile device data). These large-scale data allow researchers to understand student behaviors in ways that were not possible before. This study addresses the need to understand student behaviors using immediate feedback features in a blended learning environment utilizing rich server tracking log data. A mix of inductive and deductive statistical and data mining techniques including correlation analysis, agglomerative hierarchical clustering, K-means clustering, and multiple linear regression with LASSO variable reduction are used to recognize the students’ common behavioral patterns and understand their correlations, if any, with the students’ course performance. The results were supplemented with survey results and semi-structured interview data. The implications are two-fold: (1) increased understanding of student behaviors with immediate feedback can provide instructors an anchor to give timely interventions and recommendations for students’ study strategies, and (2) although the subject studied here is physics, it is a required course for all first-year students who are likely to be majoring in a number of STEM fields afterwards, so the results could inform the pedagogical design of other first-year STEM classes that utilize a blended learning format.

Degree

Ph.D.

Advisors

DeBoer, Purdue University.

Subject Area

Statistics|Educational technology|Computer science

COinS