Where's the Science in Software Engineering

  • Tagung:

    Zeitschriftenartikel 

  • Autoren:

    Walter F. Tichy 

  • Summary

    The article is a personal account of the methodological evolution of software engineering research from the seventies to the present. In the seventies, technology development dominated. An attitude of "seeing is believing" prevailed and arguing the pros and cons of a new technology was considered to be enough. With time, this advocacy style of research began to bother me. I thought that something was missing from a scientific approach, and that was the experiment. Controlled experiments and other empirical methods were indeed rare in software research. Fortunately, the situation began to change in the nineties, when journals and conferences began to demand evidence showing that a new tool or method would indeed improve software development. This evolution was gradual, but the situation today is that software research is now by and large following the scientific paradigm. However, the cost of experiments can be staggering. It is difficult to find professional programmers willing to participate in controlled studies and it is time-consuming and expensive to train them in the methods to be studied. I suggest an alternative to experiments with human subjects that can help produce results in certain situations more quickly.

  • Jahr:

    2014 

Beteiligte Mitarbeiter (zufällige Reihenfolge)
Titel Vorname Nachname

Bibtex

@article{Tichy14a,
author={Walter F. Tichy},
title={Where's the Science in Software Engineering},
year=2014,
month=March,
publisher={ACM},
volume={2014},
doi={http://dx.doi.org/10.1145/2590528.2590529},
abstract={The article is a personal account of the methodological evolution of software engineering research from the seventies to the present. In the seventies, technology development dominated. An attitude of "seeing is believing" prevailed and arguing the pros and cons of a new technology was considered to be enough. With time, this advocacy style of research began to bother me. I thought that something was missing from a scientific approach, and that was the experiment. Controlled experiments and other empirical methods were indeed rare in software research. Fortunately, the situation began to change in the nineties, when journals and conferences began to demand evidence showing that a new tool or method would indeed improve software development. This evolution was gradual, but the situation today is that software research is now by and large following the scientific paradigm. However, the cost of experiments can be staggering. It is difficult to find professional programmers willing to participate in controlled studies and it is time-consuming and expensive to train them in the methods to be studied. I suggest an alternative to experiments with human subjects that can help produce results in certain situations more quickly.},
pages={6},
journal={ACM Ubiquity},