AI in Software Testing

Abstract

Software Testing is a process in the software development lifecycle in which developers verify and validate that a program is free of bugs, meets the requirements and satisfies the user’s needs and expectations seamlessly. With the recent developments in Artificial Intelligence, software testing has increasingly begun to incorporate AI-driven approaches. This thesis explores the role of AI in the domain of software testing as of today. This is done mainly through a literature review that explains how AI is used in software testing, provides the benefits and challenges of using AI, and evaluates some of the tools being used. Amongst the analysed literature, it is evident that AI has had a positive impact in software testing where it is seen through improved metric like testing time, coverage and encountering post launch defects. However, there are still challenges present, mainly the availability of data to implement models used for making these software tests. Additionally, pynguin which is a testing framework and copilot which is an LLM are compared in terms of line coverage to evaluate how two different AI-based approaches perform in generating software tests.