Abstract
Unit testing is a high-effort and high-cost activity in software development. It is estimated that developers spend around 28% of the total development time writing unit tests. Automated unit testing solutions using LLMs have been developed, but they still struggle to generate tests for complex code. To address this challenge, we adopt a Design Science Research (DSR) approach to improve LLM-based unit test generation. In this paper, we present the initial findings towards the development of a nascent design theory that provides prescriptive design knowledge for developing LLM-based unit test generation tools. Based on earlier literature found during the literature review, we define 5 meta-requirements and 5 design principles. Hence, we contribute design knowledge that guides the development of an LLM-based tool capable of generating high-quality unit tests for software development organizations.
Recommended Citation
Gurung, Ram Prasad; Karhu, Katja; Chandio, Sreman; and Siemon, Dominik, "Establishing Meta-Requirements and Design Principles for
LLM-based Unit Test Generation" (2025). ACIS 2025 Proceedings. 56.
https://aisel.aisnet.org/acis2025/56