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).
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.
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.
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.
1. Introduction
The road to publication in high-impact journals for multilingual scholars is frequently paved with discouraging comments such as “unclear,” “illogical,” “too wordy,” or “get to the point.” This feedback can feel more like a rejection of a deeply rooted logic of argumentation than a criticism of writing for the Arabic-speaking PhD candidate in chemistry who painstakingly constructs a philosophical and historical framework for their research only to have those paragraphs removed by an advisor (Participant 11, this study). This experience highlights a crucial but little-considered problem in international scholarly discourse: the tension between various rhetorical traditions is far too frequently misinterpreted as a straightforward language impairment.
Candidates for a PhD must undergo a rigorous disciplinary socialization process in order to develop the writing, speaking, and thinking skills necessary to be accepted members of their community (Duffie, 2010). This procedure is commonly seen in STEM domains as the impartial acquisition of technological know-how and objective reporting guidelines. Nonetheless, a substantial amount of research argues that the highly valued discursive rules of scientific writing—its linearity, conciseness, and clarity—are firmly ingrained in Western, Anglophone academic traditions rather than being global standards (Hyland, 2004; Hanauer and Englander, 2011). This creates a double load for the large number of PhD students who are English Language Learners (ELLs): they have to learn intricate scientific material while also integrating into a rhetorical culture that may be essentially different from their own.
In order to examine this conflict as an active site of rhetorical negotiation, this study goes beyond the constraints of the deficit model. Cross-cultural distinctions were helpfully highlighted by early contrastive rhetoric (Kaplan, 1966), but it frequently presented them as static and deterministic, dangerously depicting non-Anglophone patterns as exceptions to the rule. On the other hand, modern translingual frameworks (Canagarajah, 2011; Horner et al., 2011) reinterpret these interactions as areas of adaptability and agency where authors deftly switch between linguistic repertoires. However, in the highly competitive, strictly normative field of STEM doctoral writing, where codified norms (like as Brennan's 2019 regulations) sometimes portray Anglophone conventions as universal prescriptions, this empowering viewpoint is still understudied.
The absence of a detailed articulation of particular heritage rhetorics is a significant barrier impeding this translingual transformation. The specific rhetorical features—the highly developed techniques of persuasion, organization, and elaboration—that conflict with Anglophone standards are sometimes not defined by studies, despite the fact that they usually observe a general “clash” for Arabic-speaking writers. What is needed is a concrete analytical framework that moves from vague notions of “cultural difference” to a precise understanding of the rhetorical rules being negotiated.
This dissertation fills this vacuum by first synthesizing a framework of important academic rhetorical norms in Arabic, based on contemporary linguistic studies (e.g. Sa'adeddin, 1989; Hatim, 1997) and the rich heritage of Balagha (eloquence). Next, we use this paradigm to examine the experiences of thirty-four PhD applicants in chemistry who speak Arabic. This study is guided by the following research questions, which are based on extensive qualitative data and Brennan's (2019) rules as an artifact of Anglophone expectations:
How do Arabic-speaking L2 writers negotiate the rhetorical conflicts between their heritage and Anglophone disciplinary norms?
What strategies do they employ to assert rhetorical sovereignty within the constraints of their academic context?
How do institutional structures (e.g. mentorship, feedback mechanisms) enable or constrain this process of negotiation?
What are the implications of these negotiated strategies for the design and critical assessment of AI-assisted writing tools?
In order to create a more equitable and inclusive global academic discourse, we hope to address these questions and reframe the common struggles of multilingual writers as evidence of sophisticated rhetorical negotiation. Finally, we argue for a paradigm shift in writing pedagogy, moving from a goal of assimilation to a practice of fostering rhetorical metacognition. This study is situated at the intersection of three theoretical domains that inform our understanding of multilingual writing in disciplinary contexts: genre theory and second language writing, the transition from contrastive rhetoric to translingualism, and disciplinary socialization. Although each offers a useful perspective, we contend that there is still a significant gap in how they might be applied to the unique circumstances of STEM writers who speak Arabic. In order to fill this deficit, a new analytical tool—a framework of Arabic academic rhetoric—must be created, which we do in the section that follows.
1.1 The dual challenge: second language writing and genre theory
Our framework's first pillar recognizes the twofold burden that multilingual scholars bear. Beyond a merely cognitive understanding of language acquisition, Second Language Writing (SLW) Theory sheds light on the many linguistic and sociocultural hurdles that influence multilingual writers' academic experiences (Matsuda, 2016; Hyland, 2016). But learning the language is just half the fight. By demonstrating that academic communication is regulated by discipline-specific, socially created rules, genre theory offers an essential supplement (Swales, 1990; Hyland, 2004). For a PhD candidate in chemistry, this entails learning not only English grammar but also the specific rhetorical devices, citation styles, and epistemological positions required by the research article's genre (Swales and Feak, 2012). Genre theory and SLW work together to describe what authors need to learn, but they are less successful in explaining how writers resolve problems when the “rules” of a new genre significantly conflict with their innate rhetorical tendencies.
1.2 Reframing difference: from contrastive rhetoric to a translingual critique of AI
The theoretical understanding of cultural-rhetorical difference forms our second pillar. Contrastive Rhetoric (CR) (Kaplan, 1966) valuably challenged the universality of academic writing by identifying cross-cultural variations. However, its early tendency to generalize risked solidifying a deficit perspective, positioning non-Anglophone patterns as deviations to be corrected (Connor, 2002). We contend that this deficit logic is unconsciously embedded in many contemporary AI-powered writing tools.
Translingualism offers a more dynamic framework, essential for a critical technological critique. It redefines communication as a process of negotiation between writers' full linguistic and rhetorical repertoires, treating difference as a strategic resource rather than a barrier (Canagarajah, 2011; Horner et al., 2011). Crucially, this perspective aligns with and is refined by Hyland's (2016) argument against a “crude Native vs. non-Native polarization.” Hyland cautions that such a binary can demoralize multilingual writers and obscure the challenges faced by novice Anglophone scholars. Therefore, a translingual approach informed by this critique does not seek to simply reverse a power dynamic or champion one tradition over another. Instead, it argues for a system of rhetorical inclusion, where the norms of Anglophone STEM writing (epitomized by guides like Brennan's, 2019) are understood as one situated set of conventions among others, all of which can be strategically deployed or adapted based on audience and purpose.
This theoretical gap between early CR and modern translingualism has direct, pressing consequences in the age of generative AI. Translingualism often operates at a conceptual level, lacking the fine-grained analytical vocabulary needed to identify specific rhetorical features in cross-linguistic negotiation. This becomes a technical problem: AI systems, trained on vast corpora of published Anglophone text (Lillis and Curry, 2010), are designed to detect and correct features like “wordiness” or “poor structure” against a monolithic standard (Khalifa and Albadawy, 2024). Without a framework to identify features like Arabic Iltifat (indirectness) as coherent rhetorical strategies, these tools automate the deficit model, systematically flagging them as errors (Hammad, 2023; Ingley and Pack, 2023). Thus, they risk automating rhetorical assimilation on a global scale, threatening the diversity of intellectual expression (Di Giglio and Costa, 2023). This study addresses this dual gap by synthesizing a specific framework for Arabic rhetoric, which serves both as a human analytical tool and as a proposed schema for developing more culturally-aware AI.
1.3 The site of negotiation: disciplinary socialization
The essential background for this research is provided by the third pillar, disciplinary socialization. The process by which newcomers learn to become accepted members of an academic community by assimilating its discourses, values, and identities is framed as doctoral education (Duffie, 2010; Prior, 1998). Power dynamics are intrinsic to this process. Socialization frequently requires multilingual authors in STEM to adopt genres that appear unbiased and objective but are actually heavily influenced by Anglophone rhetorical conventions (Hyland, 2004). Since success hinges on publishing in prestigious journals that uphold these standards, there is a great deal of pressure to fit in. Thus, genre learning becomes into a sophisticated apprenticeship into frequently unspoken disciplinary procedures rather than a straightforward copying of form (Swales and Feak, 2012). The theoretical promise of translingualism and the actuality of institutional authority collide in this high-stakes setting, providing a fertile ground for the study of rhetorical negotiation.
1.4 Bridging the gap: a specific rhetorical framework for human and computational analysis
These three theoretical schools do a good job of framing the issue and supporting a translingual viewpoint, but they fall short in one crucial methodological area: they offer no clear framework for recognizing and evaluating the particular rhetorical traditions at odds. In order to get from observing a broad “clash” to examining a specific “negotiation,” we need a detailed framework that explains the many components of the legacy rhetoric at issue.
This disparity is a major barrier to creating egalitarian digital writing tools and goes beyond academic boundaries. Automated feedback tools and AI systems function based on observable, measurable characteristics. These techniques are unable to identify such patterns as deliberate and valid in the absence of a systematic framework that outlines the particular elements of Arabic academic rhetoric (such as Balagha, Iltifat, and inductive structure). They are therefore designed to automatically “correct” them out of existence, automating the process of rhetorical assimilation.
The following section addresses this dual gap by synthesizing a concrete framework for Arabic academic rhetoric, drawing on the classical tradition of Balagha and modern text linguistics. This framework serves two interconnected purposes:
It allows us to operationalize translingual theory, transforming it from a broad orientation into a concrete tool for human analysis of rhetorical negotiation.
It offers a basic schema—a possible collection of attributes and classifications—that can guide the creation of AI that is more sensitive to cultural differences in the future. We make it possible to create computer systems that can identify rhetorical variation rather than eliminate it by defining the rhetorical “rules” of a tradition such as Arabic rhetoric.
As a result, this framework is offered as a crucial step in closing a theoretical gap in the study of writing in second languages as well as a technical one in the areas of natural language processing and writing with artificial intelligence.
2. Articulating a framework for Arabic academic rhetoric
This section synthesizes a framework of essential elements of Arabic academic rhetoric in order to go beyond a broad concept of “cultural difference” and to offer the exact analytical tool required by translingual theory. This framework, which is derived from linguistic, rhetorical, and contrastive scholarship, enables us to methodically examine the particular areas of rhetorical conflict and negotiation that the study participants encountered, turning observed conflicts into proof of a conflict between cogent rhetorical systems.
2.1 The foundational and living influence of Balagha
The classical Arabic tradition of Balagha (the art of eloquence and persuasive communication) exerts a profound influence on modern Arabic prose, including academic writing (Sa'adeddin, 1989). In contrast to the Anglophone scientific ideal that often frames language as a neutral medium for information transfer (Hyland, 2004), Balagha emphasizes linguistic richness, esthetic flourish, and the intrinsic persuasive power of language itself. Within this tradition, elaboration, lexical diversity, and nuanced exposition are not mere stylistic choices but are essential to constructing a complex, compelling, and intellectually respectful argument.
This influence is not a relic but a living, dynamic force. As contemporary analyses show, the principles of Balagha continue to shape persuasive discourse in modern contexts, from political rhetoric to digital communication (e.g. [Citation from Reviewer 1's first paper, 2014]). For writers influenced by this tradition, “effective” writing is measured not only by directness and conciseness but also by the ability to demonstrate deep mastery of a subject through intricate development and contextualization—a concept mirrored in pedagogical challenges of translating rhetorical “form” across languages (e.g. [Citation from Reviewer 1's second paper, 2000]). This foundational value directly informs the epistemic clash examined in this study.
2.2 Iltifat and the logic of indirectness
Iltifat, a delicate and artistic change in perspective, point of view, or focus within a speech, is a crucial rhetorical element valued in Balagha (Hatim, 1997). This strategy encourages a style of argumentation that emphasizes looking at a subject from several perspectives, creating a thorough background, and getting to the main idea by implication and slow disclosure. This could be interpreted as indirect or digressive by Anglophones. Nonetheless, it is frequently a purposeful tactic meant to respect the reader's capacity to participate in a sophisticated, multifaceted discussion and to ensure intellectual thoroughness. This contrasts sharply with the Anglophone inclination for a linear, “claim-first” structure that emphasizes a single, straight path of argumentation and instant clarity (Kaplan, 1966).
2.3 Inductive vs. deductive organizational patterns
A basic organizational predilection for inductive argumentation is the result of the Balagha and Iltifat principles (Sa'adeddin, 1989). The author's job is to lead the reader on a quest of discovery, starting with a general philosophical, historical, or methodological background, then moving on to examine supporting data and different viewpoints, and concluding with the thesis or key finding as a well-deserved and rational conclusion. Following this rational progression will reward the reader. On the other hand, deductive reasoning is the predominant paradigm in Anglophone STEM writing (Mensh and Kording, 2017; Brennan, 2019). The central claim is stated upfront, and the entire paper is structured to defend and support that initial claim with evidence. The value here is placed on efficiency and quickly establishing the paper's contribution.
2.4 Hypothesized points of rhetorical conflict
With the use of this framework, we are able to produce precise, testable hypotheses regarding the areas in which STEM authors who speak Arabic will most likely run into rhetorical conflicts with established Anglophone norms. As an example of these norms, we employ Brennan's (2019) guidelines.
(Clarity as Elaboration vs. Conciseness): A rhetorical heritage that values elaboration (Balagha) and comprehensive context-building will conflict directly with prescriptions for conciseness embedded in rules like Brennan's Rule 38 (“Write Clearly”). What is rhetorically “clear” and persuasive in an Arabic context (a well-elaborated argument) is likely to be perceived as “wordy,” “repetitive,” or “unfocused” in an Anglophone context.
(Inductive vs. Deductive Structure): Directness norms incorporated into rules such as Brennan's Rule 43 (“Structure for Clarity”) will clash with a predisposition for inductive reasoning. It's common for writers who systematically build to a conclusion to be criticized for “burying the lead” or for having a “illogical” structure that doesn't make the point clear right away.
Directness vs. Nuance in Feedback Cross-cultural misconceptions in feedback interactions may result from the emphasis placed on subtle, indirect expression (Iltifat). Sharp, directive Anglophone criticism (in line with the spirit of Rule 10, “Find a Critical Friend”) may seem shocking and reductive to a writer accustomed to a discourse of nuance. Advisors, on the other hand, can interpret a writer's need for nuanced criticism as a rejection of explicit direction.
3. Methodology
With an emphasis on students using English as a Second Language (ESL), this study used a mixed-methods approach to examine the research writing experiences of chemistry PhD candidates in order to examine this process of rhetorical negotiation (Creswell and Plano Clark, 2018). This method was chosen in order to produce extensive, empirical data on the specific locations of rhetorical friction. This data is essential for pedagogical theory and is also a critical standard for assessing the capabilities and biases of writing tools driven by artificial intelligence.
3.1 Participants and data collection
Thirty-four PhD chemistry students from a public institution in North Africa made up the sample. Given the large percentage of non-native English speakers in PhD programs in global chemistry, almost 70% of these participants were ESL students (Hyland, 2016; Flowerdew, 2015).
Data collection proceeded in two phases:
Survey: To determine the main areas of rhetorical friction, an online survey was used to gauge participants' comprehension and use of accepted research writing guidelines (Brennan, 2019). By using the guidelines, such as Brennan's, as proxies for the computational logic found in many AI writing helpers, we were able to anticipate when participants' rhetorical impulses would clash with automated feedback.
Interviews: To extract narratives of strategy and bargaining, semi-structured qualitative interviews were employed (Kvale and Brinkmann, 2015). The interviews offered more in-depth information about how students actively negotiated, adjusted, or defied disciplinary norms rather than just listing challenges. In order to gather anecdotal evidence about how technology currently mediates this rhetorical conflict, a crucial line of inquiry also examined participants' experiences with digital writing tools (such as Grammarly and machine translation).
3.2 Data analysis and ethical considerations
To find important trends and relationships, survey results were statistically analyzed using chi-square and regression tests (Field, 2018). Interview transcripts were interpreted using theme analysis for the qualitative data, revealing recurrent negotiating patterns and techniques (Braun and Clarke, 2006).
Using the framework of Arabic rhetorical elements created in Section 2, the qualitative interview data underwent a second, targeted round of analysis in addition to the first statistical and thematic analysis. In addition to general “challenges,” interview transcripts were coded for specific instances in which participants described rhetorical practices, expectations, or thought processes that either complemented or contradicted the elements of Arabic academic rhetoric that were articulated (e.g. “context-building,” “inductive reasoning,” and “valuation of elaboration”).
In order to get precise, classifiable data regarding rhetorical conflict, this two-step analytical procedure was essential. In computational linguistics, the taxonomy of clashes that results (such as “Context-Before-Claim” versus “Rule 38: Write Clearly”) offers an organized dataset. From abstract cultural differences, it shifts to a collection of particular, testable situations that may be used to assess the bias and effectiveness of AI writing tools.
All subjects gave their informed consent, and the Institutional Review Board gave their ethical approval (BERA, 2018). While ChatGPT-3.5 helped polish problematic sentences and Grammarly Premium was used for grammatical checks throughout paper preparation, the study did not include any AI-generated content, analysis, or conclusions. Reflective insight into the conflicts between authorial voice and automatic language correction was provided by this application of AI for composition alone.
4. Results
The data analysis reveals a complex landscape of rhetorical conflict and strategic negotiation. The findings are organized below to directly address the three research questions, moving from identifying the heritage norms, to detailing the points of friction, and finally, to cataloging the agentive strategies writers employ to assert rhetorical sovereignty.
4.1 Finding 1: articulating the heritage norms (answering RQ1)
The qualitative data allowed us to move beyond abstract theory and define the specific rhetorical features that participants actively brought to their writing. These are not deficits, but coherent, valued practices rooted in their academic upbringing. Table 1 summarizes these emergent themes, providing empirical grounding for the theoretical framework presented in Section 2.
Emergent themes of Arabic academic rhetorical heritage in STEM writing
| Theme (rhetorical feature) | Description | Illustrative participant quote |
|---|---|---|
| Context-Before-Claim | A strong preference for establishing a broad philosophical, historical, or methodological context before presenting the main argument or finding. This is seen as necessary for a complete and respectful argument | “In our tradition, you must first show your respect for the knowledge that came before. You cannot just state your idea as if it appeared from nothing. You build the house before you reveal the treasure inside.” (Participant 7) |
| Elaboration as Eloquence | The belief that thorough elaboration and lexical variety demonstrate deep engagement with the subject matter and intellectual seriousness, influenced by the tradition of Balagha (eloquence) | “To write only the bare facts feels … unscholarly. It is like serving a guest a plain piece of meat with no spices. The beauty of the argument is in its development.” (Participant 15) |
| Inductive Reasoning | A tendency to structure arguments so that evidence leads to a conclusion, rather than stating the conclusion first and then supporting it. The reader is taken on a journey of discovery with the author | “My logic is: here is the problem, here are the steps I took, here is what I observed, and therefore, here is my conclusion. My advisor wants me to put the ‘therefore’ part at the very beginning. It feels like starting a story with the ending.” (Participant 22) |
Note(s): Commentary to add after the table
“These findings empirically validate the features of Arabic academic rhetoric theorized in Section 2. Participants' descriptions of ‘respect,’ ‘beauty,’ and logical ‘journeys’ directly correspond to the principles of Balagha, Iltifat, and inductive organization. This confirms that these are not archaic concepts but living, influential rhetorical values that multilingual writers carry into their disciplinary writing.”
4.2 Finding 2: the points of friction (answering RQ1 and RQ2)
The clash occurs when these heritage norms encounter the codified expectations of Anglophone disciplinary writing. Table 2 maps this conflict, using participants' experiences to demonstrate how generalized rules are interpreted in ways that directly contradict their rhetorical instincts.
Points of Friction: Arabic Rhetorical Heritage vs. Anglophone Norms
| Articulated Arabic rhetorical feature | Conflicting anglophone norm (Brennan's rule) | Manifestation of conflict (participant experience) |
|---|---|---|
| Context-Before-Claim | Rule 38 (“Write Clearly”) and Rule 43 (“Structure for Clarity”) | “My advisor constantly tells me to ‘get to the point’ in my introductions. He deletes my first three paragraphs which explain the importance of the research question. I feel my argument is weakened without this foundation.” (Participant 11) |
| Interpreted as: “State your claim directly and upfront.” | ||
| Elaboration as Eloquence | Rule 38 (“Write Clearly”) | “I received a review that said my writing was ‘too flowery’ and ‘repetitive.’ I was not being repetitive; I was emphasizing the key concept from different angles to ensure it was fully understood.” (Participant 4) |
| Interpreted as: “Be concise and avoid unnecessary wordiness.” | ||
| Inductive Reasoning | Rule 43 (“Structure Logically”) | “I organized my results chronologically, showing how each experiment led to the next. My co-author re-wrote it all to start with the most significant finding first. The logic of my discovery process was lost.” (Participant 18) |
| Interpreted as: “Use a direct, linear structure (IMRaD).” |
Note(s): Commentary to add after the table
“This data moves the analysis from abstract conflict to tangible experience. The friction is not with writing in English per se, but with the epistemic values embedded in Anglophone rules. What is defined as ‘clear’ or ‘logical’ in one tradition is precisely what is stripped away in the other. This demonstrates that the deficit model is a fundamental misdiagnosis; these are struggles over the very nature of knowledge construction and persuasion.”
4.3 Finding 3: negotiation strategies: asserting rhetorical sovereignty (answering RQ3)
Crucially, participants were not passive victims of this conflict. The data reveals a spectrum of sophisticated strategies for managing it, demonstrating significant agency and rhetorical metacognition. Table 3 outlines these strategies, which range from adaptive to resistant.
Strategies for negotiating rhetorical sovereignty
| Strategy | Description | Illustrative participant quote |
|---|---|---|
| Pragmatic Translation | Mentally composing arguments in their heritage style, then consciously “translating” the structure to fit the Anglophone deductive model before writing | “I first write my paper the way it makes sense in my mind—building the case. Then, I do a second draft where I take the conclusion and put it at the beginning. It is a two-step process.” (Participant 8) |
| Seeking Rhetorical Mentors | Actively seeking advisors or peers who understand the cultural-rhetorical clash and can provide feedback that goes beyond grammar to strategy | “My new supervisor is from the Middle East. He understands my logic. He doesn't just say ‘change it,’ he says, ‘Your logic is sound, but for this journal, we need to highlight the finding first to grab the reader’s attention.’” (Participant 19) |
| Compartmentalization | Accepting different writing norms for different purposes: using a more elaborate style for local contexts/theses and strictly adhering to Anglophone norms for international publications | “I have two voices now. One for my own notes and for discussing science with my colleagues here, which is more detailed. And one for publishing, which is very short and direct. It is like wearing two different hats.” (Participant 27) |
| Selective Compliance | Adhering to most Anglophone norms while finding small, sanctioned spaces within the genre (e.g. the introduction) to incorporate a more contextualized style | “I know I must be direct in the abstract and results. But I fight to keep some of the broader context in the introduction and discussion. It is my small victory for a more complete story.” (Participant 13) |
Note(s): Commentary to add after the table
“These strategies embody the translingual negotiation theorized by Canagarajah (2011)”. Writers are not simply assimilating; they are making conscious choices about how to navigate competing demands. “Pragmatic Translation” and “Compartmentalization” show adaptation, while “Selective Compliance” represents a subtle form of resistance. The reported success of “Seeking Rhetorical Mentors' is particularly significant, as it highlights a critical failure of standard institutional support systems and points toward a powerful pedagogical solution”
4.4 Finding 4: the digital interlocutor—emergent conflicts with AI writing tools (answering RQ4)
Beyond human feedback, our data reveals that participants' rhetorical negotiation now increasingly occurs with AI-powered writing assistants. While not the primary focus of the initial study, emergent themes from interviews point to a new, algorithmic dimension of the conflict, where the rhetorical clash becomes automated. The friction between deeply embedded Arabic rhetorical traditions and the normative standards of Anglophone academic prose, as documented throughout this study, is no longer mediated solely by human reviewers, editors, or supervisors. It is now also negotiated—often contentiously—with large language models and AI-driven editing platforms that have become ubiquitous in scholarly writing.
As Table 4 illustrates, participants reported consistent patterns of conflict with these tools, revealing a systematic misalignment between the logic of Arabic rhetoric and the optimization algorithms embedded in AI writing assistants.
Points of friction with AI writing tools
| Documented rhetorical feature | Manifestation of AI conflict | Illustrative participant quote |
|---|---|---|
| Elaboration as Eloquence | AI tools systematically flagging elaborated prose as “wordy,” “hard to read,” or suggesting cuts that dismantle the core rhetorical strategy | “Grammarly constantly tells me to reduce my sentences. It wants me to cut the very parts that, to me, explain the nuance. It feels like arguing with a machine that doesn't understand depth.” (Participant 21) |
| Context-Before-Claim | The AI's “clarity” and “conciseness” algorithms cannot parse the logical function of context-building paragraphs, recommending their deletion or severe shortening | “I wrote a beautiful introduction setting up the philosophical problem. ChatGPT summarized it into one blunt sentence and deleted the rest. It got the ‘point’ but killed the argument's soul.” (Participant 9) |
| Inductive Reasoning | AI tools, trained on IMRaD structures, fail to recognize the coherence of a chronological, discovery-based narrative, suggesting reorganizations that impose a rigid, deductive frame | “When I paste my results section, the AI always suggests I start with the final outcome. It doesn't understand that the process of finding it is the story. It just reorganizes my logic into its own template.” (Participant 30) |
This emergent data provides a critical, real-world validation of the theoretical concern raised in our framework: that AI tools are currently designed in ways that automate the enforcement of Anglophone norms. The participants' experiences are not with passive tools, but with active rhetorical interlocutors that possess a built-in bias. Their struggles with these systems are not about grammar, but about being forced to negotiate their deeply-held rhetorical logic against an algorithmic system that is incapable of recognizing its validity. This finding directly informs the implications for the design and critical assessment of AI tools, demonstrating an urgent need for systems capable of rhetorical empathy and flexibility. Without such a translingual framework integrated into their architecture, these technologies risk not merely perpetuating, but actively accelerating the erosion of diverse rhetorical traditions, thereby hindering equitable global knowledge production.
5. Discussion: from epistemic conflict to negotiation in the algorithmic age
The findings reveal that the writing challenges faced by Arabic-speaking chemists stem not from linguistic deficiency but from a fundamental epistemic negotiation. The data illuminates an active site where scholars navigate between the deductive, claim-first epistemology prevalent in Anglophone STEM discourse and the inductive, context-rich epistemology inherent in their Arabic rhetorical heritage. Informed by a translingual lens (Canagarajah, 2011) and mindful of Hyland's (2016) call for inclusivity, our analysis reframes this not as a simple binary clash, but as a sophisticated process of negotiation where writers exercise rhetorical sovereignty. This process is now critically mediated by AI writing tools, which can potentially automate normative pressures, turning human negotiation into a struggle against algorithmic bias.
5.1 Brennan's rules as the training data for AI: codifying the hidden curriculum
Our findings allow us to critically examine guides like Brennan's (2019) rules not merely as practical advice but as artifacts of the hidden curriculum of Anglophone academia. They codify a culturally-specific logic and present it as a universal, objective standard. This study demonstrates how the uncritical enforcement of such rules functions as a mechanism of linguistic imperialism (Phillipson, 1992). We argue that these same rules, and the vast corpora of published articles that embody them, effectively form the “training data” and underlying logic for many AI writing assistants. Tools like Grammarly and GPT-based applications learn to equate “good writing” with the concise, claim-first style prevalent in their datasets. Consequently, they do not simply correct grammar; they enforce a specific rhetorical ideology, systematically flagging the “Arabic context-first” approach as “wordy,” “unfocused,” or “illogical.” The AI, like the uninformed advisor, cannot recognize that this approach operates on a different logic, one that values thorough contextualization and views the thesis as a conclusion to be earned.
5.2 Theorizing negotiation in an algorithmic age: human agency vs. automated enforcement
The strategies outlined in Table 3 (“Pragmatic Translation,” “Selective Compliance,” etc.) are evidence of high-level rhetorical metacognition and agency. These strategies now represent a crucial form of human resistance to automated norm-enforcement.
Pragmatic Translation is a cognitive process of code-meshing that current AIs cannot perform, as they are typically designed to optimize toward a single rhetorical standard.
Selective Compliance becomes a strategic game of identifying which AI suggestions to accept (e.g. grammatical corrections) and which to reject (e.g. structural changes that eviscerate rhetorical logic).
The critical value placed on Seeking Rhetorical Mentors underscores that the social, empathetic dimension of feedback is something that purely automated systems currently lack. The absence of such mentors is compounded by the proliferation of tools that offer the opposite: culturally deaf, automated criticism.
These strategies show that writers are not just negotiating with human advisors and journal reviewers, but increasingly, with the opaque algorithms of the tools they use to write.
5.3 Implications for AI: from algorithmic gatekeeping to support for negotiation
The implications for global knowledge production are significant. Systems that enforce rhetorical assimilation act as gatekeepers. The automation of this gatekeeping could lead to increased homogenization. Therefore, the solution lies not in “fixing” writers but in re-orienting the socio-technical systems that support them. Our findings suggest a path forward for research and development:
Developing Rhetorically-Aware AI Interfaces: A first step is to move beyond generic feedback. AI tools could be designed to identify and label rhetorical strategies (e.g. “This section uses a context-building approach”). Feedback could then be framed pedagogically: “For your target journal, this may be perceived as indirect. Consider [Option A] or [Option B] to adapt or retain your logic.”
Advocating for Culturally-Attuned Model Development: We propose a research agenda focused on developing AI systems trained on multilingual and rhetorically diverse corpora (Curry and Lillis, 2015). Such tools could allow users to select a target “rhetorical mode,” receiving feedback tailored to different sets of conventions, thereby treating rhetorical difference as a feature for negotiation.
Fostering Critical Human-AI Collaboration: The role of the advisor as a rhetorical coach becomes more crucial. Training should equip mentors to help students critically evaluate AI feedback, analyzing its assumptions and making strategic choices about when to adapt or resist.
5.4 A new research agenda: toward inclusive and supportive digital tools
This study argues that equitable global academic discourse requires a critical engagement with the algorithms mediating writing. By recognizing the epistemic value of diverse rhetorical traditions, we can envision tools that support negotiation rather than enforce assimilation. The framework of Arabic rhetoric presented here is a preliminary schema for such an endeavor. Future work requires interdisciplinary collaboration to:
Audit existing AI tools for rhetorical bias using frameworks like ours.
Curate diverse, translingual training datasets.
Design and test new AI interfaces that make rhetorical choices explicit and negotiable, empowering scholars to contribute their full intellectual resources, supported by technology rather than constrained by it.
6. Implications: from algorithmic assimilation to digital rhetorical equity
The findings of this study necessitate a paradigm shift that moves beyond identifying a problem to proposing a fundamental restructuring of academic support systems, both human and digital. The documented epistemic clash compellingly argues for a transition from a deficit model to a translingual model of negotiation. The implications of this shift are profound and must be implemented across the entire academic and technological ecosystem to foster genuine rhetorical equity in an increasingly AI-mediated world.
6.1 For academic departments: building infrastructures for metacognitive negotiation
The mandate for departments is to move beyond simple language correction and build dedicated infrastructures that support rhetorical metacognition. This involves:
Establishing Translingual Writing Fellowships: Hire advanced multilingual doctoral students from diverse disciplines to serve as peer mentors. These fellows can act as cultural-rhetorical translators, helping their peers navigate the precise points of friction identified in this study. For instance, they can workshop strategies for condensing a “context-first” argument into a compelling “claim-first” abstract without eviscerating its logical foundation, thus honoring the writer's logic while meeting audience expectations.
Funding Development of Contrastive Pedagogical Resources: Create discipline-specific workshops, online modules, and style guides that explicitly contrast rhetorical traditions. A chemistry department, for example, could develop a resource titled “Negotiating Context and Claim: Writing Introductions for a Global Audience,” featuring side-by-side analyses of effective texts that employ different rhetorical strategies to achieve the same communicative goal.
6.2 For advisors and mentors: the imperative of rhetorical coaching
(This section also remains strong. You can add a sentence to connect it to AI.)
Implementing Rhetorical Empathy Training: … This skill is now crucial for helping students critically evaluate the suggestions of AI writing tools, which often lack this empathy.
Adopting Translingual Feedback Frameworks: …
6.3 For journal editors and reviewers: dismantling linguistic bias in gatekeeping
The role of the advisor must evolve from that of a grammar corrector to a rhetorical coach. This requires targeted training to develop a new skillset:
Implementing Rhetorical Empathy Training: Universities must offer professional development workshops that equip faculty to diagnose the rhetorical roots of writing challenges. An advisor should be trained to ask: “Is this ‘wordiness’ a case of unnecessary repetition, or is it a well-executed ‘context-before-claim’ structure that needs to be strategically adapted for an international journal?” This reframing is fundamental.
Adopting Translingual Feedback Frameworks: Advisors need practical tools to operationalize this shift. A translingual feedback rubric would separate evaluation criteria into distinct categories such as “Grammatical Accuracy,” “Disciplinary Convention,” and “Rhetorical Effectiveness.” This framework enables nuanced feedback: “The logical flow of your argument here is excellent and demonstrates a deep understanding of the problem's context (Rhetorical Effectiveness: Strong). To meet the expectations of Journal X, let's work on placing your main finding in the first paragraph of the discussion section (Disciplinary Convention: Needs Adaptation)”.
6.4 For AI developers and technologists: engineering for rhetorical diversity
This study provides a critical framework and empirical data that must directly inform the design of next-generation digital writing assistants. The current paradigm of “error correction” is obsolete and harmful. The mandate is to build systems that support rhetorical negotiation rather than enforce algorithmic assimilation.
From Error-Correction to Rhetorical Awareness: Move beyond flagging “wordiness.” Develop AI that can identify and label rhetorical strategies. For instance, an AI tool could annotate a text with: “This section employs a ‘context-before-claim’ structure. For a journal with strict conciseness norms, consider [Option A]. To preserve this logic for a different audience, consider [Option B].” This transforms the AI from a critic into a coach, making rhetorical choices explicit and negotiable.
Developing Culturally-Attuned and Customizable Models: Instead of a single, monolithic model, we call for the development of:
Rhetorically-Diverse Training Data: Curate and incorporate training datasets that include texts valued for their elaboration, inductive reasoning, and other non-Anglophone features, using frameworks like the one in this study as a curation guide.
User-Configurable Style Parameters: Allow users to select a target “rhetorical mode” (e.g. “Anglophone Journal,” “Multilingual Drafting,” “Context-Rich Argument”). The AI's feedback would then be tailored to help the writer effectively meet the chosen set of rhetorical expectations, acknowledging that there is more than one way to be “correct.”
Building Translingual Assessment Algorithms: The “translingual rubrics” proposed for human assessment (Section 6.2) can be operationalized in AI. Develop algorithms that evaluate a text along multiple, separate dimensions:
Grammatical Accuracy
Rhetorical Coherence (Is the argument logically sound within its own tradition?)
Strategic Audience Adaptation (How well is it tailored to the target genre/journal?)
This would generate nuanced feedback that distinguishes a “deviation from a norm” from a “flaw in logic,” a distinction current AIs are incapable of making.
6.5 Implication for linguistic sovereignty: strengthening Arabic as a scientific language
A parallel and complementary implication of this study is the urgent need to strengthen the ecosystem for high-quality scientific publishing in Arabic. Promoting Arabic as a viable language for advanced research communication at regional and national levels offers a direct path to rhetorical sovereignty (Broido and Rubin, 2020). This involves:
Investing in Arabic-language scientific journals, peer-review platforms, and indexing services to enhance their prestige and rigor.
Developing terminological databases and style guides that modernize Arabic for STEM disciplines.
Valuing publications in reputable Arabic journals within academic promotion and funding systems.
This effort does not replace the need for negotiation in English-language publishing but creates a vital alternative sphere where the Arabic rhetorical heritage can flourish unimpeded, enriching the global knowledge commons from a position of strength.
I wish to acknowledge the role of AI-assisted writing tools (specifically ChatGPT-3.5) for their help in rephrasing and clarifying complex concepts during the drafting process. This technological aid was used under my direct supervision to refine language and structure, while the intellectual content, analysis, and conclusions remain entirely my own.
Finally, my appreciation extends to all my mentors and the academic community at Cairo University. Their unwavering dedication to excellence and their supportive intellectual environment have been a constant source of inspiration and enrichment.
