ABSTRACT
Context: Pair programming is a software development approach which involves a pair of programmers working on the same task at the same time using the same set of computing resources. Pair programming is useful for practitioners as well as for students learning programming at different levels of education: K-12 to higher education. However, there is no recent study, to the best of our knowledge, reviewing the empirical research on the pair programming practice in higher education. We believe that such a study is important to give a prudent view of the practical implications to educators, and research implications to encourage new interventions in this field.
Objective: The objective of this chapter is to present a systematic literature review of the current empirical evidence of using pair programming as a pedagogical tool in Computer Science education in universities/colleges. A scientific approach of literature review is followed to specifically provide to the educators and researchers an organized view of the Pair Programming practice as per the research reported in the literature. Despite this, it will limit researcher bias, and also improve quality of analysis.
Method: The authors performed the systematic literature survey on five electronic databases resulting in 1744 papers which were reduced to 68 as per the inclusion and exclusion criteria mentioned in the review protocol. Snowballing, forward as well as backward, was also used as a search technique. Thematic analysis helped to identify the objectives and the underlying themes in the related literature.
Results: The chapter provides answers of five research questions to address the role of pair programming in an education setting. We could identify six measures of interest and their related metrics to understand the existing empirical work. The most common measure is students’ performance. Pair programming has mixed effect on students’ performance, but almost universally positive effect on students’ attitudes (i.e. enjoyment) toward programming. However, there are also studies where it has no effect.
Conclusion: This chapter indicates many open problems in this research field. A few are: (i) Small sample size in majority of the studies hinders generalization of the results; (ii) Lack of context details while reporting PP experiments for publication obscures validity of the results, (iii) Longitudinal analysis of PP experiments with learning tasks of increasing size and complexity is important to build real-world evidence of the practice.
