Classroom response systems (“clickers”) are popular among instructors to promote student engagement and to check students' understanding of class material. Since free-text responses can give a richer picture than multiple-choice answers, we have built functionality into ISLE (Integrated Statistics Learning Environment) that allows one to analyze student answers on-the-fly during lectures or computer labs.
The ISLE response visualizer displays graphs and statistics for the collected data from a question or a chosen subset. For free-text responses, the response visualizer clusters answers and displays a representative answer from each cluster. Because documents arrive in real-time, we rely on an an incremental k-means clustering approach (Lloyd, 1982) using cosine distance and employ the “hashing trick” (Weinberger et al., 2009), whereby words are mapped to indices via a hash function instead of building a dictionary of encountered words.
We discuss insights we have gained on how students from different backgrounds learn introductory statistics and data analysis, and how the system can be used by instructors in the classroom to clarify points in case of student misconceptions.