The Fluency Presumption: Probabilistic AI Detection, Second-Language Writing, and the Procedural Abdication of the University
Year: 2026
Type: preprint
Abstract: : Predictability is not evidence of synthetic authorship. University integrity systems increasingly deploy probabilistic software that scores student writing by perplexity and burstiness, flagging texts whose syntax is uniform and vocabulary expected. Those features define the prose of second-language writers still acquiring English. By equating statistical regularity with machine generation, academic institutions establish native-speaker fluency as the baseline for human innocence. Non-native students face an asymmetric evidentiary trap where disciplinary proceedings demand impossible affirmative proof of unassisted authorship. For international students holding F-1 visas, resulting suspensions trigger mandatory federal SEVIS termination and abrupt deportation risks under judicial deference to faculty judgment. The same architecture coerces students into commercial rewriting tools that trigger subsequent concealment flags. Treating detection scores as actionable evidence constitutes procedural abdication. Universities must eliminate probabilistic scoring and relocate proof of authorship to observable educational practice. Keywords: Fluency Presumption, AI Detection, Probabilistic Classifiers, Second-Language Writing, Language Acquisition, Academic Integrity, Perplexity, Burstiness, Turnitin, GPTZero, Disciplinary Due Process, F-1 Visa, SEVIS Termination, Procedural Abdication, Lexical Diversity, Synthetic Remediation, Algorithmic Bias, Higher Education Governance, Title VI, Technological Due Process
Authors: Jefferson, Josie, Velasco, Felix
Language: English
Subfield: Education
Field: Social Sciences
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7233553144
Cites:
Cited by:
Related to:
