Although pair programming has been studied since the late 1990s, it is only recently that the results of earlier studies are being fine-tuned. For example, the results of a 2007 study conducted in Europe suggests that although pair programming might not always increase software quality, it can do so when the software being built is relatively complex and junior developers are on the team. This paper describes two experiments involving the development of small systems (10,000 to less than 100,000 lines of code) and very small programs (less than 10,000 lines of code) by student programmers in an Asian university. The results involving the small systems, which were actually complete systems designed to support the Personal Software Process (PSP), showed that defect densities of systems written by the pair programming teams were significantly lower than those written by the teams that used the traditional approach of individual coding and testing of units, followed by integration testing. On the other hand, results involving the very small (and therefore much less complex) programs did not show any significant differences between the defect densities of the programs written by the pair programmers and of those written by the solo programmers, though they did show significantly greater productivity of the solo programmers when writing simpler code. The combined results of the two experiments suggest that pair programming might increase software quality without decreasing productivity when projects are sufficiently large (or complex) for the programmers working on them.

Despite the growing attention given to agile methods and practices in developed countries, there is an exiguity of empirical studies showing their successful (or unsuccessful) use in the developing world. After reviewing the literature on agile practices in developing countries, especially in Asia, this paper adds to the scant literature by describing an experiment on the use of pair programming by students in an undergraduate software engineering course at a Philippine university. Results show that defect densities of programs written by pair programming teams were significantly lower than those written by teams that followed the traditional approach of individual coding of units followed by unit and integration testing. The results, which tend to support the positive outcomes of similar studies in the U.S. despite cultural differences, underscore the importance of seriously investigating the use of pair programming and other agile practices in academic as well as industrial settings in Asia.

Agile methods are increasingly gaining attention in many developed countries; however, there is a dearth of empirical studies showing their successful use in developing nations. This needs to be addressed because most software offshore outsourcing destinations are in the developing world. This paper describes experiences in the use of agile methods and practices by software development firms in the Philippines, a developing country that has consistently appeared in lists of top software outsourcing destinations. The results tend to confirm those found in the literature concerning the positive impact of agile methods on software development, and how software organizations tend to adapt agile practices to suit their own organizational and project contexts. One case study in this paper also describes a particular approach to agile offshore software development.

There is a lack of published studies on software development in Southeast Asia, which is fast becoming an IT outsourcing haven. This paper presents exploratory survey and case study results on software practices of some software firms in five ASEAN countries (Malaysia, Philippines, Singapore, Thailand and Vietnam), and provides directions for further research on software practices in the ASEAN/Southeast Asian region.

The personal software process (PSP) is a process and performance improvement method aimed at individual software engineers. The use of PSP has been shown to result in benefits such as improved estimation accuracy and reduced defect density of individuals. However, the experience of our institute and of several others is that recording various size and defect data can be onerous, which in, turn can lead to adoption and data quality problems. This paper describes a system that we have developed that performs automatic size and defect recording, aside from providing facilities for viewing and editing the usual PSP logs and reports. Moreover, the system automatically classifies and ranks defects, and then consolidates schedules and defect lists of individual developers into a schedule and defect library for the developers' team.

This paper describes a learning process for the tutor of an intelligent tutoring system (ITS) to automatically learn models of student categories and self-improve its instructional plans on the level of these categories. Using real-world teaching scenarios as experiment data, we empirically show that for every category the tutor is able to efficiently learn effective instructional plans. Our experiment results also show that the absence of category background knowledge decreases the tutor's learning performance as well the effectiveness of the learned instructional plans.

Educational systems such as intelligent tutoring systems (ITSs) are computer programs that provide customized and adaptive learning according to learner's needs, behavior, and learning situation. Most implementations of these systems are built from scratch and are set to tackle a specific domain or a programming concept (e.g. Johnson, 1990). As a result, ITSs limit the use of learner's profile in one tutoring system. We propose a learner agent concept that allows a profile to encompass multiple learning activities. We discuss how the learner agent is used as the student model of an ITS. We also focus on how the learner agent collects and stores information into this profile.

This paper presents how multistrategy error detection and discovery (MEDD), a student modeling system using machine learning can be applied to the domain of object oriented programming. Java is the language used in learning object oriented programming. MEDD detects the learner's errors and discovers the misconceptions based on the presence (or absence) of errors.

This paper discusses a machine learning framework that uses extraction, classification, and generalization techniques to classify students according to their cognitive and behavioral learning patterns and to categorize tutoring strategies of expert human tutors. A great deal of the discussion focuses on the use of reinforcement learning techniques, specifically the ?-greedy and temporal difference TD(0) methods in deriving tutoring policies over a class of students. Future works will deal on incremental learning and modification of learned policies while the tutor performs on-line in real-time, and extracting and learning the way expert tutors execute their tutoring activities.

Instructional systems are computer programs that provide educational support to individual learners. One such system is the intelligent tutoring systems (ITSs). ITSs center on individualized learning and do not consider the situation where learners can learn beyond the teacher-learner relationship. Several learning companion systems (LCS) were implemented to create social learning environments. However these systems are merely extensions of tutor agents and act as "learners". In this paper, we introduce the use of learner agents both as student modeling systems and as peers.

CBR-Tutor is an Internet agent-based tutoring system that uses case-based reasoning approach in providing adaptive instruction to its learners. Cases can quickly recognize whether a teaching strategy is relevant to apply in a given situation. It is composed of four types of agents. These are the system agent (SA), case facilitator agent (CFA), case-based information agent (CIA) and case-based tutor agent (CTA). The CTAs interact directly with the learner. It has a set of local cases, which are commonly used. If the local cases are not useful for the new situation, the CTA will request retrieval from the global set of cases. The CIAs are responsible for storing and retrieving cases from the global case libraries. These cases contain situations experienced by the CTAs in the system. This paper presents how cases are retrieved in CBR-Tutor.