Using Different Matching Protocols to Assess the Effectiveness of Personalized Feedback in Web-Based Learning

• In Event: 77.015 - Canadian Educational Researchers' Association

Abstract

The history of feedback (FB) research is one of inconsistent findings and contradictory results - due perhaps to the interference of unaccounted for individual learner characteristics with FB. Despite this, FB has long been considered an essential component of the teaching process, with a further emphasis on personalizing it for individual students. With increasingly large post-secondary class sizes, there is a need for efficacious computer-generated personalized FB. Thus, this study aimed to see if computer generated personalized FB is superior to computer-generated generic FB in terms of academic achievement?
Students from multiple sections of an undergraduate biology course were provided either one of two forms of FB. Generic FB was common across all members, while personalized FB was tailored to individual students in accordance with 40 distinct attributes. Ethical limitations permitted students to self select their FB condition following random assignment. To ensure the data approximate a fully-blocked experimental design, subjects were matched on preidentified covariates of final grade with different matching protocols.
Results showed that students receiving the personalized FB significantly outperformed those receiving generic FB. This study also found that different matching protocols yield different results. Implications of different matching protocols for assessing effectiveness of educational interventions are discussed.

Authors

• Matthew Todd Schmidt, University of Saskatchewan

• Amin Mousavi, University of Saskatchewan

• Vicki Squires, University of Saskatchewan

• Kenneth Wilson, University of Saskatchewan