Impact of Flipped Classroom Model on High-workload and Low-income Students in Upper-division Computer Science

The flipped classroom (FC) improves student experiences, attitudes, interaction, and performance. With FC, collaborative problem-solving activities replace the lecture (peer-instruction). Outside of class, students watch self-paced videos rather than complete homework. The instructor checks for understanding with a quiz before each lecture, Just in Time Teaching (JiTT). This work concludes a two year study on FC at a medium sized Hispanic Serving Institution involving 70 consenting participants, three sections of Computer Architecture and three sections of Artificial Intelligence (AI). Literature establishes the benefits of FC, yet there is still much to be investigated. Much has changed in the two decades since the FC model was proposed. Modern classrooms are highly online with video content (recorded lectures). We are motivated to reassess student preferences, resource use, performance, and attitudes in comparison to the modern class. Most FC studies focus on CS-0, lower division, and non-majors leaving upper-division core understudied, and we uniquely present our work relative to a control population (CON) with a fixed curriculum and set of instructors. Surprisingly, students favor our CON model, where lecture is spent modeling example problems. The result is statistically significant (p < .1), whereas FC attitudes do not change with participation. When responding to questions about the quality of lecture, 70.59% of comments support our CON model. Literature in secondary education suggests minority students do not prefer dialogic instruction, and it is possible this phenomenon continues with postsecondary education. However, CON has no measurable impact on academic performance while FC does. For Artificial Intelligence midterm exam performance is 79.97% ± 7.75% and 73.22% ± 7.73% for FC and CON respectively, which is significant (p < .1). FC may improve performance in algorithm/theory-based classes despite no significant difference in attitude. FC instructors found prep times to increase by hundreds of hours, concurring with other works, which is considerable in the context of our other findings. It is possible for a traditional classroom to reproduce the attitudes of a FC if well taught and using specific teaching strategies, such as recording lectures. We find that students will enroll in MOOCs or online courses concurrently with the class to obtain quality videos if they are not provided. Some resources are not free, and this behavior is concerning for programs that serve low-income or high-workload/low-availability students. Instructors should make their own videos or provide a curated list, regardless of classroom model. Future work will study effective ways to better engage of female students.