A Human Centric Requirements Engineering Framework for Assessing Github Copilot Output

TL;DR

This paper proposes a human-centric framework to evaluate GitHub Copilot's ability to adapt to user needs and facilitate collaborative programming, addressing a gap in existing technical-focused assessment methods.

Contribution

It introduces a novel requirements framework with metrics for assessing human factors in AI programming assistants like GitHub Copilot.

Findings

Copilot adapts explanations based on user expertise

The framework reveals areas for improving human-AI interaction

Results highlight importance of human-centered evaluation metrics

Abstract

The rapid adoption of Artificial Intelligence(AI) programming assistants such as GitHub Copilot introduces new challenges in how these software tools address human needs. Many existing evaluation frameworks address technical aspects such as code correctness and efficiency, but often overlook crucial human factors that affect the successful integration of AI assistants in software development workflows. In this study, I analyzed GitHub Copilot's interaction with users through its chat interface, measured Copilot's ability to adapt explanations and code generation to user expertise levels, and assessed its effectiveness in facilitating collaborative programming experiences. I established a human-centered requirements framework with clear metrics to evaluate these qualities in GitHub Copilot chat. I discussed the test results and their implications for future analysis of human requirements in automated programming.