Reducing procrastination on introductory physics online homework for college students using a planning prompt intervention

We examine the effectiveness of a planning prompt intervention to reduce procrastination on online homework for college students. The intervention asked students to indicate their intention to earn small amounts of extra credit for completing assignments earlier and form a plan to realize their intentions. Students’ learning behavior is measured by five data metrics collected from gradebook and student interaction logs from four sections of the same college level physics course, including three metrics that capture how students space their work on assignments and two that measure the level of student engagement when completing assignments. To separate the impact of extraneous factors from the treatment effect, we employ a “difference in differences” method—initially developed in economics—and construct multilinear models for each of the five metrics. Our models show that by simply asking students to form a plan prior to the assignment, students on average earned 5% more extra credit, completed homework significantly earlier, and spread out their work significantly more. However, the intervention did not significantly change students’ level of engagement with the learning materials, nor did it change students’ work distribution on the next assignment.

Figure 1

Observed and modeled work distribution metrics. Blue lines indicate observed averages, with error bars for 1 standard error of the mean. Orange dotted lines show the modeled values from Eq. (1). Purple crosses show the model’s prediction of each metric if the planning-prompt intervention had not been implemented $({\beta }_{T}T\equiv 0)$. The gap between the orange dotted line and the purple cross thus indicates the estimated effect size. The purple marks appear only for sequences that received the intervention, which sequence this is differs by class section.

Figure 2

Observed and modeled engagement metrics. Blue lines indicate observed averages, with error bars for 1 standard error of the mean. Orange dotted lines show the modeled values from Eq. (1). Purple crosses show the model’s prediction of each metric if the planning-prompt intervention had not been implemented (${\beta }_{T}T\equiv 0$). The gap between the orange dotted line and the purple cross thus indicates the estimated effect size. The purple marks appear only for sequences that received the intervention, which sequence this is differs by class section.