Project-based learning (PBL) is a growing component of engineering education in the United States. Its perceived educational value is exemplified by its explicit mention in ABET’s Criterion 5, which requires engineering programs to provide a culminating design experience that incorporates engineering standards and multiple constraints. Capstone courses and design-build-test projects allow students to synthesize and apply engineering knowledge, skills, and tools to open-ended design problems. Students work and communicate in teams to complete tasks like generating requirements, and testing and integrating equipment. There appears to be widespread consensus that project-based learning is valuable, but, how well do these projects prepare students for engineering challenges in professional practice?
We consider one aspect of professional practice—failure. Despite many improvement efforts by organizations, systems engineering failures continue to occur. Previous research identified a set of common causes for these project failures. Does PBL provide students with opportunities to fail safely, and thereby learn to avoid failure in professional practice? We present here an approach to compare the rates of occurrence of failure causes in student team projects with industry projects. By comparing the occurrence rates, we achieve our first research goal to evaluate whether PBL offers sufficient opportunities of failure to students. Out of the ten failure causes we examined, we found that four are underrepresented in PBL, two of which are fundamentally related to systems engineering practice.
Failure causes may be hard to identify without the benefit of hindsight, so, we developed a set of crowd signals that may point to their presence. We observed 18 different student design projects across two semesters. Each week, the students answered a set of questions we developed to measure these crowd signals, while the instructors directly pointed out any instances of the failure causes they observed during the students’ efforts. With the availability of such data we built logistic regression models to find correlations between specific crowd signals and the occurrence of a particular failure cause. By interpreting the regression coefficients, we achieve our second research goal to suggest specific improvements that instructors can use to give their students more failure opportunities during PBL.
Are you a researcher? Would you like to cite this paper?
Visit the ASEE document repository at
for more tools and easy citations.