Deriving Empirically Grounded NFR Specifications from Practitioner Discourse: A Validated Methodology Applied to Trustworthy APIs in the AI Era

Specifying non-functional requirements (NFRs) for rapidly evolving domains such as trustworthy APIs in the AI era is challenging as best practices emerge through practitioner discourse faster than traditional requirements engineering can capture them. We present a systematic methodology for deriving prioritized NFR specifications from multimedia practitioner discourse combining AI-assisted transcript analysis, grounded theory principles, and Theme Coverage Score (TCS) validation. Our five-task approach integrates purposive sampling, automated transcription with speaker diarization, grounded theory coding extracting stakeholder-specific themes with TCS quantification, MoSCoW prioritization using empirically derived thresholds (Must Have ≥85%, Should Have 65–84%, Could Have 45–64%, and Won’t Have <45%), and NFR specification consistent with ISO/IEC 25010:2023 principles of stakeholder perspective, measurable quality criteria, and explicit rationale. Applying this methodology to 22 expert presentations on trustworthy APIs yields Weighted Coverage Score of 0.71 and 30 prioritized NFR specifications across five trustworthiness dimensions. MoSCoW classification produces 11 Must Have requirements (Robustness and Transparency), 9 Should Have, 6 Could Have, and 4 Won’t Have. The analysis reveals systematic disparities where Fairness contributes zero Must Have or Should Have requirements due to insufficient practitioner consensus. Each NFR emphasizes stakeholder perspective, measurable quality criteria, and explicit rationale, enabling systematic verification. The validated methodology with complete replication package enables empirically grounded, prioritized NFR derivation from practitioner discourse in any rapidly evolving domain.