by John DeNero, Electrical Engineering and Computer Science
Teaching Effectiveness Award Essay, 2008
Artificial intelligence is an exciting field and students arrive on the first day of Berkeley’s introductory class full of enthusiasm. The course surveys a broad and diverse set of techniques, moving on to a new type of problem every few weeks. To the seasoned researcher, these problems are all related. To many undergraduates, however, our course seems to jump haphazardly among a grab-bag of unrelated topics. The course’s original syllabus began with a conceptual roadmap of how various problems related to each other. But since our students didn’t understand the individual problems yet, they didn’t understand the purpose of the framework. The roadmap was soon forgotten and the students’ sense of the course’s direction was forgotten along with it. Every time the topic changed, more students were left wondering how the current chapter followed from the last and more students disengaged.
To infuse continuity into the course, I designed a series of projects around Pacman, a classic video game with lots of retro charm. By varying the rules and goals of the game, each project required a different artificial intelligence technique, from finding paths through mazes to outmaneuvering smart opponents to anticipating the position of moving objects. I made the projects distinct in technical content, but as consistent as possible in all other respects.
The projects have been a big hit with students: the theme taps into the students’ enthusiasm for games, the graphics let students quickly visualize the output of their programming, the projects include both clear directions and opportunities for creativity, and the scaffolding I built allows students to focus exclusively on the interesting parts of the program. But the largest benefit is that the project’s content lends continuity to the students’ entire course experience that keeps them engaged throughout the semester. The familiarity of Pacman makes unfamiliar content more approachable. In office hours, Pacman serves as a running example for all types of problems. I also built Pacman-based demonstrations to complement new lecture topics. Pacman even shows up in the course’s exam questions. This thread of consistency carries the students’ excitement into each new topic, which helps to mitigate their confusion and disorientation.
Pacman also prompts students to find connections between topics. The course strives not only to teach students a set of techniques, but also to teach them how to integrate and synthesize those techniques to solve new problems. To this end, I designed an optional course contest. This new twist on Pacman included all of the challenging aspects of the previous projects and pressed students to combine techniques in novel and complex ways. To our delight, nearly half the class participated last semester. The completely optional “play-off” event spawned four hours of shouting, cheering, and frantic code editing as student teams pitted their programs against one another to determine a contest winner. Several students even asked me to organize a rematch event after the final exam so that they could continue tuning their programs after the semester ended.
Responses to the new projects have been overwhelmingly positive. In response to the evaluation survey question, “What worked best for you in this course?” 65% of students listed the course projects. Students’ numerical responses to the question, “How worthwhile was the course?” have increased by 15%, as have teaching effectiveness ratings, all under the same instructor. Enrollment in the course increased a dramatic 69% the semester after Pacman was first introduced. A similar course at the University of Pennsylvania has already adopted the projects and faculty from other universities are considering using them as well.
By building projects that unified a set of disparate technical topics, we found a compelling way to keep students engaged in our course. Other courses could benefit from a similar approach: when the subject matter of a course might seem incoherent to students, in-depth assignments like this one can serve to connect themes in ways that even the clearest of explanations cannot.
I would like to thank Professor Dan Klein for his excellent guidance, inspired ideas and constant input that helped to make these projects a success.