Understanding algorithms through shared representations

Learning and teaching algorithms are notoriously difficult for undergraduate computer-science students and instructors. Research on using algorithm animation for teaching algorithms to these students has a long history. My research began by investigating this literature on student learning from algorithm animations and conducting five experimental studies of an existing hypermedia algorithm visualization system. The first two experiments compared a paper-based version of the hypermedia system to the computer-based system. The surprising result of the second experiment was that students using the paper-based version learned more than the students using the computer-based version. The third experiment was done to investigate whether giving students exercises designed to have them mentally simulate the algorithm, while working with an algorithm visualization system or pseudocode, encourages the students to be more active interpreters of the representations shown. The result was that the simulation exercises did not result in significantly improved learning, but using the visualization system did. The fourth experiment compared different versions of the visualization system and found that students who saw interactive animated analogies as part of the visualization learned more than students who saw a text description instead of the analogy. The fifth experiment was a usability study that uncovered difficulties students had in interpreting animations.; After these studies with the hypermedia algorithm animation system, I conducted a qualitative study to find the strategies, resources and practices students normally use to learn algorithms. This study found that students are normally collaborative, working in small groups to solve problems and understand difficult concepts. The study also found that classroom authority played a key role in which explanatory materials students chose to focus on.; Results of these efforts led me to develop a system, CAROUSEL (Collaborative Algorithm Representations Of Undergraduates for Self-Enhanced Learning), to support students in collaborative algorithm learning. Students created representations of the algorithms they were studying in various media, and then they used the system to share, evaluate and discuss these representations. The system and the activities of representation creation, sharing, evaluation and discussion that it supports were then studied in three experiments. Results indicated that these activities help students better understand algorithms. These studies also provided quantitative evidence that constructing representations significantly improved student learning.; Two more experimental studies were done to explore what characteristics of representations are beneficial for learning an algorithm from a representation created by students using CAROUSEL. Students evaluated representations by rating how well they fit with six characteristics; usefulness, understandability, salience, familiarity, pleasure and originality. I found that salience and pleasure have a positive impact on learning in both experiments. Familiarity was shown to have a positive impact on learning and originality a negative impact.; Thus, this research has shown that computer-supported constructive and collaborative activities enhance algorithm learning. Furthermore, characteristics of representations that impact learning have also been identified.