Comparing Video Feedback Provided by Multiple Students

In this experiment we sourced hint feedback from two students on the same problem. These students were not directed in their explanation, they were simply asked to provide the method they used to solve the problem. As you can tell by watching the videos below, these explanations were rather different. The goal of this study is to observe any potential differences in learning gains across multiple video solutions for the same problem, thereby learning how to select the best crowdsourced content. Eventually, this idea can be scaled to all users, sourcing videos from students helping their peers, and automated using k-armed bandits to evolve system content.

This problem set begins by asking students whether or not they are able to see video content. If they are not, they are removed from the study and given a traditional problem set with text feedback. For students who are able to access video content, they are randomly assigned to receive either video feedback (experimental condition) or text feedback (control condition). Within the experimental condition, students received one of two versions of the same problem, with feedback provided by two different students (the videos shown below). Students only receive this feedback if they answer the problem incorrectly.

Click here to try this problem set. You will be randomly assigned to a condition each time you refresh your browser.

Heffernan, N., Heffernan, C., Ostrow, K., Minh, D., Nguyen, L, Heffernan, C. & Heffernan T. (In Preparation) How to Find the Best Explanations: Crowdsourcing Student Video Feedback.

A current page under construction is here.