Labor Saving Techniques
Carnegie Mellon University
Given that a major goal of the course re-design project is to substitute technology "capital" for faculty teaching "labor," in what particular aspects of the course and its delivery are you finding that you are able to do this?
Two areas of capital-for-labor substitutions that have taken place by using StatTutor are (1) encouraging and monitoring student interaction and (2) creating and grading assignments. When students work on data-analysis problems using StatTutor, they receive individualized feedback on their solution steps. This level of step-by-step help could not be achieved, even with the relatively large number of graduate and undergraduate TAs proctoring the statistics lab sessions. The human component is still an important part of the lab sessions, as we encourage students to discuss intermittent “thought questions” with the TAs as they work through each lab assignment. The StatTutor system gives the TAs an easy way to document that students have completed these thought questions and summarizes lab attendance automatically. These characteristics will allow us to cut the number of TAs in the statistics labs in half in the second year of implementation. We chose not to reduce the number of TAs during the prototyping and testing period.
These were all anticipated benefits of our project. An unanticipated benefit was finding that the creation and grading of assignments was even easier than expected. With the StatTutor system we can collect more fine-grained assessments of students’ work without adding any TA labor. Also, assignment creation with StatTutor was fairly quick for the faculty teaching our course pilot.
December 2001 Update: Our goals and intentions for achieving capital-for-labor substitution have not changed. We expect to be able to reduce the number of TAs in the labs and TA labor in grading homework assignments. We intend to conduct a specific experiment in spring 2002 in reducing the number of TAs in labs with the expectation of achieving the reduction originally planned by the fall of 2002. We had hoped to do such experiments in fall 2001, but for technical reasons we were only able to deploy three of the StatTutor versions of lab assignments this fall, so the reduction was not feasible. Thus, our timeline on achieving the full level of TA reductions has slipped by one semester. It currently appears that we will be deploying 6 to 7 StatTutor versions of assignments in labs in the spring of 2002 as well as using StatTutor versions of a number of homework assignments. This level of implementation will allow us to conduct experiments to verify that StatTutor will change the nature of TA-student interactions in labs in ways that will allow the full reduction of TAs in labs (from 2 TAs to 1 TA in each lab) originally proposed.
Specifically, our plan for the spring of 2002 is to reduce the number of TAs in half of lab sections to 1 TA, while leaving the number at 2 TAs in the others. The StatTutor versions of lab assignments will be implemented in all of the sections. We will conduct a field study of TA-student interactions in the two different settings. A graduate student in psychology will be an observer in the labs to conduct this field study. (Some statistics graduate students may also be used to help with recording the data.)
The resulting data will serve two purposes. First, it will provide an account of the kinds of questions that students are asking TAs now that the students have StatTutor as an additional tutor. (Our original expectation was that the questions would change both in nature and frequency.) Secondly, it will give us more information about whether the current iterations of the StatTutor assignments are sufficiently well designed to support the planned reduction without harming students' learning. For example, if there is a difference in performance between the two groups, those with 1 TA and those with 2 TAs, this will be a source of concern about how to redesign of StatTutor or strategies for using it so that the reduction can be successful. Simply put, we feel this experiment is a key formative assessment step along the way to success in our original goal of capital-for-labor substitution.
Having solved some of the technical challenges described below will also allow the use of StatTutor versions of homework assignments. Oded Meyer, the statistics professor currently teaching the introductory courses, reports that the labor associated with homework grading is perhaps more significant that we originally estimated in our proposal. The original proposal conjectured the possibility of labor savings in homework grading, but did not depend on it as a means of labor reduction. We will begin to learn in the spring of 2002 whether there are more savings here than originally anticipated.
Program in Course Redesign Quick Links: