AI summaryⓘ
The authors studied how well large language model (LLM) assistants help programmers discuss complex, unclear rules (called Non-Functional Requirements or NFRs) related to HIPAA compliance. They had 49 programmers use GitHub Copilot to check these rules against a specific codebase and analyzed the conversations across different aspects like reasoning and code identification. They found that while programmers often agreed with the LLM's answers, the LLM's accuracy compared to expert judgment was low. They also discovered that longer answers lowered user satisfaction, but more proactive help improved it. Their work highlights areas to improve these AI tools for better teamwork on complex software requirements.
Large Language ModelsNon-Functional RequirementsHIPAA ComplianceGitHub CopilotSoftware EvaluationMulti-turn DialogueUser SatisfactionCode LocalizationCollaborative Reasoning
Authors
Ali Pourghasemi Fatideh, Wilder Baldwin, Maria Dhakal, Collin McMillan, Sepideh Ghanavati
Abstract
LLM-based dialogue assistants have become mainstream tools for software developers, yet current evaluation benchmarks focus exclusively on functional correctness. This leaves a critical gap in assessing the quality and accuracy of these conversations when handling Non-Functional Requirements (NFRs), which are inherently vague, context-dependent, and involve many parts of a program. Evaluating how well these systems support collaborative reasoning about NFRs requires methods that go beyond single-turn accuracy to capture both the correctness of the system's outputs and the quality of the multi-turn interaction. In this paper, we investigate the accuracy and quality of multi-turn conversations between developers and an LLM-based agent in the domain of Health Insurance Portability and Accountability Act (HIPAA) regulatory compliance. We hired 49 programmers to interact with GitHub Copilot to assess 148 HIPAA-derived NFRs against the iTrust codebase, a system designed to comply with HIPAA regulations, across three dimensions: requirement satisfaction level, reasoning, and code localization. We find that developers tend to agree with LLM assessments, but accuracy against expert ground truth is low. We model user satisfaction and find that longer system responses and more information-providing turns negatively affect user satisfaction, whereas proactive interactions positively affect it. Our findings provide insights for designing LLM-based dialogue systems that support NFR assessment.