Algorithmic assimilation or digital sovereignty? AI writing tools and the negotiation of Arabic rhetoric in global science

Abstract

Purpose The deficit model, which is prevalent in L2 writing pedagogy, is contested by this study since it portrays multilingual scholars' struggles with Anglophone norms as a language weakness. We contend that these conflicts are better understood as a negotiation of deeply embedded rhetorical traditions, drawing on translingual theory (Canagarajah, 2011). Design/methodology/approach We first synthesize a framework of Arabic academic rhetoric using a mixed-methods approach, identifying salient characteristics including inductive structure, elaboration as eloquence (based in Balagha), and context-before-claim argumentation. Next, we use this approach to analyze data from 34 PhD applicants in chemistry who speak Arabic. Findings Results show that rather than a lack of ability, their difficulties with normative standards (such as Brennan (2019) “Write Clearly”) are the result of an epistemological conflict. To establish rhetorical sovereignty, participants used advanced techniques such as pragmatic translation, seeking rhetorical mentors, and selective compliance. A technique for auditing AI writing helpers for rhetorical bias is provided by the synthesized framework. Originality/value According to the study's findings, encouraging metacognitive negotiation rather than imposing assimilation is the new paradigm for effective academic support. These findings have urgent implications for the design of AI-based writing tools, suggesting that without a translingual framework, such technologies risk automating rhetorical bias and hindering equitable global knowledge production.