View Site Map

Lessons Learned

Carnegie Mellon University

Pedagogical Improvement Techniques

What techniques contributed most to improving the quality of student learning?

StatTutor. StatTutor is an automated, intelligent tutoring system that monitors students' work as they go through lab exercises, provides them feedback when they pursue an unproductive path, and closely tracks and assesses individual student's acquisition of skills in statistical inference—in effect, providing an individual tutor for each student. StatTutor

  • increased the immediacy and consistency of feedback that students receive during problem solving;
  • made labs and homework more open-ended, exploratory, and active;
  • supported a dynamic model of problem-solving in lab exercises by asking students to choose and categorize relevant variables, and select the appropriate statistical package tools; and,
  • facilitated student learning of problem-solving skills that had previously appeared to be most difficult, namely, (1) planning and selecting appropriate statistical analyses, (2) evaluating the validity of statistical inferences, and (3) transferring these and other skills to new contexts (e.g., from homework to exam, from current class to downstream courses).

Cost Reduction Techniques

What techniques contributed most to reducing costs?

StatTutor. StatTutor made the course less labor intensive, while serving the same number of students. StatTutor took on the roles of supervising and tutoring students and grading their lab and homework exercises. The remaining TAs responded to students' in-depth questions.

Implementation Issues

What implementation issues were most important?

Software development. The expectation for this project was that the development time would be greater than desired because developing a sophisticated tool that improves pedagogy necessarily involves a design-test-redesign cycle. In fact, the development of StatTutor and its full implementation in the course have taken longer than other less programming-intensive strategies might take. There are still a number of lab assignments and a large number of homework assignments for which StatTutor lessons have not been constructed. The deployment of the tool, however, was relatively easy. StatTutor was always under refinement, yet each time the 250 students came to lab to use it, there were no big surprises (read: crashes of the technology).

Professional preparation of TAs. Reassigning TAs to research projects rather than to labs became a concern to the department as the project progressed, which in turn caused the department to move more conservatively than originally expected. Faculty discovered the value of TA involvement in the introductory course because it gave them good professional preparation for future teaching. In addition, the faculty valued having TAs in labs because it allowed them to develop better calibrations for grading assignments. As an example, a TA working in a lab using StatTutor developed a solution set for a homework problem (one of the TA's other assignments is to develop such solution sets) using the StatTutor "structure." The professor reported that the solution set was far better ("more organized and clear") than those usually developed by graduate students. He viewed this as a by-product of inculcating the TAs into the StatTutor idea in the lab setting.

Student input into design decisions. Not only did StatTutor inspire the teaching in the course to change, but also the teaching inspired changes and improvements in StatTutor. This symbiotic relationship helped to address many design decisions in how to implement the tool (to fit best with the course) and how to teach the course (to fit best with the tool). For example, feedback from students indicated that one component of the data-analysis procedure seemed redundant (specifically, students considered reporting their results to be redundant with drawing conclusions). Students were quite vocal about redundancies when it involved their work time. Upon further reflection, the instructor and other members of the design team realized that the students' view of these pieces as redundant indicated that they had not understood a distinction in the course content (i.e., reporting results really is different from drawing conclusions from them). This resulted in a change both to the wording in StatTutor, including clarification in the hints and feedback associated with these pieces, as well as more careful exposition and examples of the relevant concepts in the lecture.

Collecting students' reports on how the tool, the TA allocation, etc. were working from a student point-of-view was an invaluable tool in this project. Students were asked specific, open-ended questions regarding design decisions for which the team saw multiple approaches, soliciting comments on what students liked, didn't like, and would suggest adding. Just as human-computer interaction specialists intone to designers that they remember, "You are not the user", so curriculum and learning tool designers should remember, "You are not the student."

Laboratory availability. StatTutor was originally developed to be used in computer lab clusters. It has been difficult to reserve adequate computer cluster time for the course's lab sessions. A real concern in fall 2002 was that these clusters would be unavailable and StatTutor could not be used in the course. The team is now creating an online version of the StatTutor tool so that students can use it in their dorms or in their local clusters for homework, thus making it widely available.

Back

 

Program in Course Redesign Quick Links:

Program In Course Redesign Main Page...

Lessons Learned:
Round 1...
Round II...
Round III...

Savings:
Round I...
Round II...
Round III...

Project Descriptions:
Sorted by Discipline...
Sorted by Model...
Sorted by Success...
Sorted by Grant Rounds...