Towards personalized peer-matching for learning programming.

Abstract

Peer review has been used as a proven process for sharing knowledge and improving skills among the learners. This became more popular by the use of ICT. Peer review provided effective, qualitative and summative feedback. It is challenging when it comes to the peer reviw of programming languages where there is no specific single solution to the problem. Programming languages students need peer review to learn new techniques while sharing their ideas to solve the problems they face. This helps the programming language teachers in the marking process of the assignments and hence mininizes their workload. Students receive feedbacks from reviewers which helps them to learn the concepts from different perspectives by looking at others’ work and techniques. The main challenge becomes the students’ needs in different sub-sections of the problem solving. In this thesis, we used the personalized need for review by introducing a personalized peer matching algorithm. It is used to assign reviewers to students based on their personal needs and performances. Peer matching is used in combination with the assessment criteria that are used in the assessment of the submitted solutions. This helps the student get the most needed help on the assessment criteria where s/he is weaker compared to others. This is done by assigning reviewers that are not on the same performance level on those criteria. The algorithm groups the students according to their performances in three levels (low, medium and high). The algorithm assigns each student with three reviewers one from each level. These students must have scored different performances on the same criterion. In order to use the algorithm and show its benefits to students, we developed a peer review system for programming language based on teachers’ and students’ perspectives. This system uses peer matching algorithm on an introductory programming course being offered at the University of Agder with the first year students of computer science. Students were given a programming problem to solve and submit the solution for review. Students were assigned reviewers using the personalized peer matching algorithm and they reviewed others’ solutions. The algorithm showed significant benefits to the students as they received improving feedback on criteria they had not performed well in their previous assignment.