New Framework Links Teaching Rigor and Teacher Support to Student AI Use in Academic Writing

Generative artificial intelligence can now produce fluent, well-structured academic prose on demand, and that capability has unsettled one of the oldest assumptions in writing education: that a polished final text is evidence of the thinking behind it. A paper published in Frontiers in Education argues that whether students use these tools to deepen their learning or simply to cut corners depends less on the technology itself than on the classroom conditions surrounding it. The article, a hypothesis and theory contribution by Eman Al Nafjan of King Saud bin Abdulaziz University for Health Sciences in Riyadh, proposes a conceptual framework called the Rigor–Commitment Alignment Framework, or RCAF, for understanding student engagement with AI in English for Academic Purposes (EAP) writing.

The framework crosses two dimensions of the pedagogical environment: academic rigor and teacher commitment. Academic rigor is defined as the extent to which instructional tasks demand sustained cognitive effort and engagement with disciplinary reasoning. It is operationalized through three subdimensions: cognitive demand, meaning the degree to which tasks require higher-order processes beyond recall; epistemic authenticity, meaning the extent to which tasks approximate real disciplinary practices of constructing and evaluating knowledge rather than decontextualized school exercises; and performance accountability, meaning the requirement that students demonstrate their own understanding through visible processes of drafting, justification, and revision. Teacher commitment, in this framework, refers not to organizational or occupational attachment but to a sustained professional-relational orientation toward students’ learning, enacted through instructional responsiveness, relational support, and critical scaffolding.

Intersecting these two axes yields four instructional configurations. Low rigor combined with low commitment is labeled Neglectful; low rigor with high commitment is Indulgent; high rigor with low commitment is Punitive; and high rigor with high commitment is Rigor with Support. The author is careful to note that these labels are practitioner-facing shorthand rather than stable classifications of individual teachers. Configurations are understood as context-dependent positions that may vary across tasks, courses, or phases of instruction, and teachers may move between them depending on the demands of a given activity, institutional constraints such as class size and assessment regimes, and the signals they receive from student behavior.

Each configuration is linked, through a set of explicitly testable propositions, to four prototypical engagement profiles. Deep-agentic engagement combines mastery-oriented motivation and sustained cognitive investment with critical, autonomous use of generative tools, in which students evaluate and revise AI output in relation to their own developing understanding. Strategic engagement involves deliberate, selective allocation of effort toward assessed outcomes, with GenAI used as a productivity tool to maximize efficiency on graded components rather than primarily to support learning. Surface engagement involves task compliance with minimal cognitive processing, in which students may accept AI-generated content without adequate evaluation or revision. The disengaged profile involves low motivation, withdrawal, avoidance, or defensive delegation, which in AI-mediated contexts may include covert reliance on AI-generated text or complete avoidance of tasks.

The predictions are explicitly probabilistic rather than deterministic. The framework proposes that rigor will be more positively associated with deep-agentic engagement when teacher commitment is high, but more strongly associated with strategic-defensive engagement or disengagement when commitment is low. High rigor with high commitment is predicted to carry the greatest probability of deep-agentic engagement and critical, evaluative GenAI use, although strong performance orientations may still produce strategic engagement. High rigor with low commitment is predicted to increase the probability of strategic-defensive engagement, with disengagement becoming more likely when students perceive demands as unattainable. Low rigor with high commitment may produce surface-compliant engagement, expressed through positive affective or social participation but comparatively limited cognitive investment, although highly self-regulated students may still engage deeply. Low rigor with low commitment is predicted to increase the probability of disengagement, minimal-effort completion, or extensive AI delegation, though high assessment stakes may instead produce strategic completion.

A central claim of the paper is that generative AI acts as an amplifying condition rather than a deterministic cause. Unlike earlier writing-support technologies that handled wording, grammar, and fluency, large language model-based systems extend to more advanced composing activities such as planning, reasoning, source integration, and revision. When assessment emphasizes final-product quality and the composing process remains invisible, extensive delegation can yield acceptable text without corresponding cognitive or rhetorical engagement. In high-rigor, high-commitment contexts, by contrast, process accountability, responsive guidance, and support for evaluative judgement are more likely to direct AI use toward questioning, verification, and revision. The framework formalizes this amplifying mechanism in a proposition stating that differences among configurations will be stronger when GenAI is readily available, assessment emphasizes final products, and composing processes remain unseen.

The framework is grounded in Bandura’s triadic reciprocal model of causation, in which personal factors, observable behavior, and the environment reciprocally influence one another. Within the RCAF, academic rigor and teacher commitment jointly constitute the instructional environment; personal factors include goal orientation, motivation, self-efficacy, language proficiency, and GenAI literacy; and behavior covers observable engagement and AI-use practices such as planning, drafting, verifying, delegating, and avoiding. The reciprocal structure matters because engagement is not only an outcome of instruction but also an input to it. The paper cites evidence that student engagement predicts changes in teacher practice over time, and it proposes that students’ engagement and AI-use behavior will predict subsequent adjustments in rigor and commitment, with the direction of adjustment depending on how teachers interpret student behavior and on institutional constraints. A teacher confronted with evidence of extensive delegation might increase process accountability and scaffolding, or might instead lower demands, tighten surveillance, or withdraw support, potentially reinforcing surface or defensive engagement.

The methodological approach is a structured conceptual synthesis rather than an empirical study. The author conducted a purposive, conceptually driven search of peer-reviewed sources indexed in Scopus, Web of Science, and ERIC covering 2000 to 2026, with seminal earlier works included through backward citation searching, organized around three domains: teacher and instructional conditions, student engagement, and generative AI. Key constructs were charted, compared for overlap, and grouped according to their primary function, then integrated within the Bandura-based architecture. The paper draws on approaches-to-learning research for cognitive depth, self-determination theory for motivational quality, agentic engagement research for proactive learner contribution, and multidimensional engagement models for the cognitive, affective, behavioral, and social manifestations of engagement. The author also positions the framework against related dimensional models, including the demandingness-by-responsiveness framework from parenting styles research and models of motivating teaching, arguing that the RCAF’s added value lies in explaining how GenAI’s alternative pathway to acceptable performance makes the relationship between instructional conditions and engagement contingent on assessment design, process visibility, and accountability.

The practical implications are framed diagnostically rather than as generic advice. In Neglectful environments, where neither demand nor guidance is present, the intervention is to redesign tasks and feedback structures so that writing is treated as a process. In Indulgent environments, the recommendation is to preserve relational support while increasing cognitive demand, for example by asking students to compare AI-generated claims with source evidence or justify citation choices. In Punitive environments, where high demands meet weak support and AI use is experienced mainly through surveillance and detection, the paper warns that prohibition may intensify covert use, and recommends increasing responsiveness, relational support, and scaffolding through writing conferences, process portfolios, and guided revision. In Rigor with Support environments, AI can be used to generate alternatives, test clarity, examine counterarguments, and refine structure, with the teacher’s role becoming more rather than less important as students need guidance in judging when AI output is useful, insufficient, or misleading. For institutional policy, the framework supports pedagogically contingent rules that specify the instructional conditions under which AI use is likely to support learning, rather than treating it as uniformly permissible or prohibited.

The author is explicit about the limits of the contribution. The RCAF is propositional, an integrative synthesis of established literatures rather than an empirically validated theory, and its diagnostic value is presently heuristic rather than tested. The engagement profiles are prototypical patterns rather than mutually exclusive learner categories, and students may display mixed or shifting patterns depending on task type, assessment pressure, and prior experience. The framework was developed primarily for English-medium EAP writing in higher education, and its applicability to secondary education, community-based ESL, or EFL settings with limited AI access remains untested. It may also underrepresent institutional assessment systems, disciplinary conventions, socioeconomic differences, and learner factors. Future research directions include operationalizing and measuring the core constructs, longitudinal studies of reciprocal movement between engagement and instruction, process-oriented studies using think-aloud protocols, keystroke logging, and revision analysis, and design-based research on whether the framework functions as a professional development tool. The paper’s closing argument is that generative AI has not changed the fundamental purpose of EAP writing instruction, but has made the quality of pedagogical design more consequential.

Subject of Research: A conceptual framework linking academic rigor and teacher commitment to student engagement with generative AI in English for Academic Purposes writing

Article Title: Rigor, commitment, and generative AI: a conceptual framework for student engagement in EAP writing

Article References: Al Nafjan, E. (2026). Rigor, commitment, and generative AI: a conceptual framework for student engagement in EAP writing. Frontiers in Education, 11, 1944529. https://doi.org/10.3389/feduc.2026.1944529

Image Credits: AI Generated

DOI: 10.3389/feduc.2026.1944529

Keywords: generative AI, academic writing, English for Academic Purposes, student engagement, teacher commitment, academic rigor, higher education, conceptual framework, self-determination theory, AI policy, writing pedagogy, Frontiers in Education