Systematizing AI-Assisted Web Development with the C.O.M.P.A.S.S. Framework and D.I.A.L. Cycle

Abstract

The integration of Large Language Models (LLMs) into software engineering promises to accelerate development. However, their adoption is often ad-hoc, lacking a systematic methodology to bridge high-level requirements with the precise instructions required by AI. This paper addresses this methodological gap by proposing a novel framework consisting of two parts the C.O.M.P.A.S.S. Framework, which provides structured prompt specifications, and the D.I.A.L. Cycle for iterative code development refinement. The utility of this methodology is evaluated through a qualitative action research study involving six interns developing six web applications tailored to professional needs. Our findings indicate that the framework provides essential scaffolding for developers, shifting their workflow from a reactive, bug-fixing paradigm to a proactive, specification-driven engineering process. The framework is observed to systematize their approach, resulting in more consistent and comprehensive prompts. A key contribution of this work is the formalization of this replicable methodology, which offers a fundamental step towards a more systematic and rigorous approach to collaborative human-AI-assisted software development.

Ähnliche Arbeiten