Researchers Propose Domain-Specific Generative AI Declaration Frameworks

Nicholas Micallef and Olga Petrovska propose a domain-specific Generative AI declaration framework for higher education in a paper posted to arXiv on June 11, 2026 (arXiv:2606.13389, cs.CY). The framework offers two task-specific declaration structures, one for writing-focused activities and one for coding assessments, developed for a Computer Science department and built on an existing taxonomy of GenAI usage types.

Per the abstract, the framework asks students to categorize AI assistance by cognitive and developmental stage rather than declaring AI use as a single undifferentiated fact, for example distinguishing structural planning from textual content generation in writing tasks, and code improvement from code generation in programming tasks. The authors frame this as a design artefact and position paper: it proposes and motivates the framework rather than reporting results from deploying it.

For assessment designers, the practical appeal is that task-specific declarations can align more directly with grading rubrics and academic-integrity triage than a binary checkbox, and they prompt students to reflect on which parts of their own process were AI-assisted. The tradeoff is added complexity: instructors need clear guidance on interpreting multi-stage declarations, and the framework's usefulness depends on students filling it out honestly.

As a position and design paper, this framework has not yet been tested: there is no reported data on student compliance rates, instructor workload, or whether it actually improves academic-integrity outcomes compared to binary disclosure. Watch for empirical follow-up from the authors or peer institutions, and for whether the writing and coding categories generalize to other disciplines.