Trading in high-stakes tests for project-based assessment – Josh Edge, Mathematics

My concern

With the challenges of the hybrid model this semester, I was concerned about how to assess my students. Typical timed tests would be a challenge to deliver in a remote setting. Not only would such tests add considerable stress to my students lives (Is my connection stable? Do I have enough time? How do I ask questions?), but such stress can often place students

at greater risk of an Academic Integrity Board infraction. Moreover, it is not clear to me that an exam-based assessment would be the best way to evaluate student performance in the class.

My solution

To avoid high-stakes testing, I opted for a project-based assessment in my Introduction to Statistics course. This semester-long project simulates a research study where students create a series of questions, conduct a survey, then use statistical methods to analyze the survey results and present their findings. To make the project manageable, there are a number of small assignments that scaffold to the final project at the end of semester – a written report and recorded presentation to the class.

To start, students were assigned to groups of three to six people based on a common-interest survey I created. Groups then settled on a topic they would survey Denison students: exercise habits, political engagement, study habits, etc. Next, the groups created a survey to gather their information. Since I knew what type of statistical tests they would ultimately use (basic hypothesis testing – t-test and ANOVA and multiple linear regression), I helped refine the questions so that the data would be useful for such testing.

Once each group had their refined questions on one topic, I used Qualtrics to compile one large survey from all the groups, roughly 100 questions. Students then asked peers to complete the survey, which took about 15 minutes. I incentivized participation and got over 250 responses!

Once the data was collected from the master survey, I gave each group their relative data. As the semester progresses, they will perform (1) exploratory data analysis – graphs, visuals, (2) basic hypothesis testing – t-test, ANOVA, etc. and (3) multiple linear regression to build a model to predict likely outcomes. This scaffolded work will culminate in a recorded presentation to the class coupled with a live question and answer session, as well as a final written report. The written report will just be a summary of the work they have been creating all semester.

My reflections

So, how much work did this take? It took roughly two hours to create the Qualtrics master survey and code the questions to ensure useful data. I had another two hours working with students to create useful questions. But this initial time investment is not much more than I would have done creating one test, and clearly less time than creating three or four tests over the semester.

Another benefit of this approach is student anxiety has decreased considerably. I no longer have the lines out my (virtual) door as students cram for a high-stakes test. Student morale—even in a pandemic—seems quite high, and I get the sense that students are actually able to learn without the “Will this be on the test?” blinders. Moreover, because we were able to eliminate days for review and exams, we are actually able to cover more material in the class than usual.

With the scaffolded approach, students are staying on task, getting things done in small incremental steps. Plus, these smaller assignments make it is easy for me to step in to nudge a group along. So far, though, I have found that students have not needed motivation to complete these assignments. Students are turning in their assignments early and are very proactive in coming to me with questions to this point. My hypothesis is that students feel more ownership of their project because they got to pick the topic.

While the global pandemic forced me to rethink my assessment for this semester, I will continue to use this project-based approach even when we return to “normal times.” The project is more in line with what statisticians actually do, so my students get a better appreciation of the course, and the low-stakes semester-long approach creates a more positive mental-health environment for my students. In the future, I might also consider partnering with another introductory course (in psychology, biology, etc.) to expand the possibility of project topics.

Joshua Edge
Visiting Assistant Professor
Computer Science