In a software development methodology called Extreme Programming [1], it has been learned that two programmers working together at a single keyboard are more productive that those same two working alone. Additionally they seem to have higher job satisfaction as well. People of course are social animals. But we also make mistakes and have imperfect knowledge. Pair programming helps in all of this. Laurie Williams of North Carolina State University [3] has also shown that this effect extends to the class room. Students learn more and produce better artifacts when they create computer programs in pairs.
In education we often team students for particular tasks, but have seldom been willing to go beyond this. In particular, we think of intellectual activity as being individual and not shared. This in spite of the fact that many professionals write joint scholarly papers.
Most of us are also faced with the problem of large classes of students with a variety of skill levels, making it hard to target discussions at the right level. Some of our students need additional help that is hard to give. On the other hand, many (most) professors have at one time or another had the experience that they didn't thoroughly understand something until they had to teach it.
This paper is a proposal for an experiment to bring all of this together. In short, the experiment will attempt to measure if it is effective to make every classroom experience in the first Computer Science course a shared one, by teaming students for all activities. This experiment will be carried out by the main author with the assistance of two undergraduate students (to be selected). The design will entail choosing two sections of the first programming course, one experimental and the other control, and conducting the experimental group using a Pervasive Pair Learning methodology.
Statistical measurements will be taken, measuring overall performance of the two groups. The low quartile in each group will be especially important to this study. In addition, instructor satisfaction and evaluation of the process will be captured via questionnaire. A special (blind) exam will be given to the students that will not be part of their grading and will contain additional course evaluation questions. These questions will include an evaluation of student satisfaction in addition to those measuring student achievement.
The student aides will switch back and forth between the sections, with one attending each lecture in each section. They will record observations for later analysis. They will also be available to the instructor to help manage the process (of forming teams, for example). The chief investigator will teach neither of the two sections, but will serve as a resource to the instructors throughout the process. An attempt will be made to use similar materials in the two sections.
The particular form of Pervasive Pair Learning that will be used is the following. Every activity will be undertaken by at least two students--buddy system. Each student will have a buddy for each activity. The buddies will rotate, so that two students will be paired infrequently and most students will eventually be paired with many others. Some teams may be self selected, but others will be assigned. When a team is paired the grade for the work of the team will be shared equally by the members.
At the discretion of the instructor, larger groups may be formed by combining sets of pairs into larger working groups. Instructors will be encouraged to conduct group activities in the classroom such as "Student Design Sprint," [2] or similar.
Examinations in the experimental group will be given according to the following rule. Two weeks before any exam, pairs are formed for the exam. Pairs are encouraged to work together in preparation. The exam is taken as usual, but the grades of a team are shared using a 90/10 rule. Each student in a team will earn 90% of the mark on their own paper plus 10% of the mark on that of the buddy.
At the end of the process the chief investigator will prepare a report for publication jointly with the two student investigators. If this is at a conference, the students will be encouraged to present the paper.
The goal of this experiment is to encourage students to be responsible for each other's learning and to determine if this can be done effectively while achieving high student satisfaction. The main worry of most observers is that the better students would see this as a drag on their grade and an imposition on their time. It is hoped, and expected, that increased learning will make up for this.
Another worry is that the poor student will get a free ride. This can only be determined by measuring student outcomes, which is the purpose of the experiment. If the poorer students learn more and get a better grade, it will be to everyone's benefit. If the effect does occur, an investigation will attempt to find compensating pedagogy to improve this process and lessen the effect.
The experiment will report both statistical and non-statistical results. While it will not give a definitive answer to the question posed, due to the small sample set, it will guide further study. It will also introduce the student co-authors to the methodology.
Note that while this is being submitted to the Learning Assessment program, it has elements appropriate for the Eugene M. Lang Student Faculty Research Fellowship program as well. The students accepted will meet the requirements of that program. A stipend is also requested for the students as part of the project.
[1] Kent Beck, Extreme Programming Explained: Embrace Change, Addison Wesley
Longman, 2000
[2] Joseph Bergin, Fourteen Pedagogical Patterns, http://csis.pace.edu/~bergin/PedPat1.3.html.
[3] Laurie Williams and Robert Kessler, Experimenting with Industry's "Pair-Programming"
Model in the Computer Science Classroom, http://www.pairprogramming.com/csed.pdf.
Last Updated: April 30, 2002