Grant Braught

The Case for Pair Programming in the Computer Science Classroom

Previous studies indicate that the use of pair programming has beneficial effects on student learning. In this article, we present a controlled study that directly measured students’ acquisition of individual programming skills using laboratory practica (in which students programmed individually under exam conditions). Additionally, we analyzed other measures of student performance, attitudes, and retention. Our results provide direct evidence that pair programming improves the individual programming skills of lower SAT students, and that students who pair program are more confident in their work and are more likely to successfully complete the course. Results from the four other major studies of the effects of pair programming are reviewed and compared with those presented here in order to draw broader conclusions.

The effects of pair-programming on individual programming skill

Previous studies have reported significant educational benefits from the use of pair-programming, in which two students program together at the same computer. Here, we report the results of a controlled study designed to measure the effects of pair-programming on the development of individual programming ability. Our findings indicate significant improvements in individual programming skill for students with lower SAT scores. Additionally, we found that all students are more likely to complete the course successfully when using pair-programming.

The benefits of pairing by ability

An analysis of data from 259 CS1 students is performed to compare the performance of students who were paired by demonstrated ability to that of students who were paired randomly or worked alone. The results suggest that when given individual programming tasks to complete, lowest-quartile students who were paired by ability perform better than those who were paired randomly and those who worked alone.

Core empirical concepts and skills for computer science

Educators are increasingly acknowledging that practical problems in computer science demand basic competencies in experimentation and data analysis. However, little effort has been made towards explicitly identifying those empirical concepts and skills needed by computer scientists, nor in developing methods of integrating those concepts and skills into CS curricula. In this paper, we identify a core list of empirical competencies and motivate them based on established courses outside of computer science, their potential use in standard CS courses, and their application to real-world problems. Sample assignments that facilitate the integration of these competencies into the CS curriculum are also discussed.

Empirical investigation throughout the CS curriculum

Empirical skills are playing an increasingly important role in the computing profession and our society. But while traditional computer science curricula are effective in teaching software design skills, little attention has been paid to developing empirical investigative skills such as forming testable hypotheses, designing experiments, critiquing their validity, collecting data, explaining results, and drawing conclusions. In this paper, we describe an initiative at Dickinson College that integrates the development of empirical skills throughout the computer science curriculum. At the introductory level, students perform experiments, analyze the results, and discuss their conclusions. In subsequent courses, they develop their skills at designing, conducting and critiquing experiments through incrementally more open-ended assignments. By their senior year, they are capable of forming hypotheses, designing and conducting experiments, and presenting conclusions based on the results.

The knob & switch computer: A computer architecture simulator for introductory computer science

The Knob & Switch Computer is a computer architecture simulatordesigned to teach beginning students the basics of computerorganization. It differs from existing simulators in twosignificant ways: (1) it incorporates cognitive hooks in the formof knobs and switches that encourage exploration and discovery onthe part of the student; and (2) it can be presented one componentat a time, starting with a simple interactive data path andbuilding incrementally to a full-featured stored program machine.Both of these features make it possible to engage beginningstudents and effectively convey an understanding of how computerswork. The Knob & Switch Computer simulator can also motivatethe study of other computing topics such as data representation,assembly language programming, and RISC vs. CISC architectures. Inaddition to describing the Knob & Switch Computer, we discussexperiences using the simulator in breadth-based introductorycourses both at Dickinson College and Creighton University.

Teaching empirical skills and concepts in computer science using random walks

An argument is made for integrating the study of empirical skills and concepts into the computer science curriculum. With reference to past work an incremental approach is advocated for the study of these skills and concepts. A unique assignment that exemplifies the advocated approach is presented. This assignment, based on the study of random walks, is intended to introduce empirical investigation as early as is possible, during the first week of the first course. Two extensions to this assignment, one for the first course and one for a programming languages course, are discussed and used to illustrate the advocated incremental approach.

dLife: a Java library for multiplatform robotics, AI and vision in undergraduate CS and research

dLife is a free and open-source Java library that supports undergraduate education and research involving robotics, artificial intelligence, machine learning and computer vision. The design of dLife addresses many concerns raised by experience reports in the CS education literature including a shortened code/test/debug cycle, ready access to robot sensor information and close integration with a robotic simulation system. Full support is currently provided for the following robots: Finch, Hemisson (or Khepera Jr.), Sony Aibo, Khepera 2, Khepera 3, and Pioneer 3, with more in development. Easily extensible packages support classroom and research applications using neural networks, genetic algorithms, reinforcement learning and computer vision.

Learning and lineage selection in genetic algorithms

Lineage selection is a process by which traits that are not directly assessed by the fitness function can evolve. Reported here is an investigation of the effects of individual learning on the evolution of one such trait, self-adaptive mutation rates. It is found that the efficacy of the learning mechanism employed (its potential to increase individual fitness) has a significant effect on the number of generations required for self-adaptive mutation rates to evolve. When highly efficient learning mechanisms are used the evolution of self-adaptive mutation rates requires a greater number of generations than in the absence of learning. Conversely, when less efficient learning mechanisms are used fewer generations are required, as compared to the non-learning case.

Evolving Evolvability: Evolving both representations and operators

The behavior of an evolutionary system incorporating both an evolving genetic representation (a learning mechanism) and an evolving genetic operator (mutation) is explored. Simulations demonstrate the evolution of evolvability through the co-adaptation of these two mechanisms. It is also shown that this co-adaptation produces a transmission function that becomes more conservative as the strength of the learning mechanism increases.