Programming Expertise during Incremental Software Development: An Empirical Study

This paper studies the expertise differences between intermediate and expert programmers during incremental software development. We conducted experiments on the intermediates and experts by asking them to perform programming on the same problem. We recorded the process and analyzed the data. We identified the differences between intermediate and expert programmers in terms of design decision generation, test case generation, the hypotheses generation when debugging the program and actions at different Bloom's taxonomy levels. It seems that experts generate higher levels, more general and higher quality hypotheses and are more able to evaluate their hypotheses and design decisions. Intermediate programmers seem to pay more attention to the detailed decisions and lack ability a evaluating the hypotheses. Experts are also better able to get rid of discredited hypotheses while intermediates often try to keep hypotheses despite some contradictory evidence. Experts generate an overview in general before beginning to program. Intermediates generate more hypotheses when they are debugging programs while experts produce them quickly and accommodate them with the data without introducing more hypotheses. Experts are better in using the domain knowledge and have better abstract knowledge than intermediate programmers at conceptual levels as well. With regards to Bloom's taxonomy, it seems that experts spend more efforts at higher Bloom's cognitive levels and intermediates have more activities at lower Bloom's levels. This study might have implications for knowledge engineering and provide guidance for software development education.