Reducing human dependency and bias in software design and development tasks

Abstract

The success of software development tasks depends, to a great extent, not only on the availability of the experienced and skillful software engineers, but also on the unbiased inputs from a variety of stakeholders. The experienced and skilled software engineers are in short supply. We also know that there are no born experts in any field; the expertise is gained from the experience of others. In this thesis we present techniques that we have developed to allow us to reduce dependency on human experts in the field of software development. The techniques that we built exploit the vast body of experiential knowledge that is available in the form of natural language text. Further, stakeholders who participate in collective decision making in a project tend to be rational. That is, they always aim to maximize their own benefit, often putting the quality of the overall decision outcomes at a second place. The impact that such rational stakeholders can have, when participating in software design decision making, is significant. The issues pertaining to rational agents is well studied in the field of social choice theory. In this thesis we also present a novel application of the ideas from social choice theory to eliminate stakeholder bias from software design decision making.