A Student-Centered Framework for Responsible Use of Generative AI in Higher Education

Synthesis: Alsammani (2026) proposes a student-facing framework for responsible Generative AI use in Higher Education: ten actionable guidelines organized around three pillars — Learning and Growth, Ethics and Integrity, and Awareness and Safety. Drawing on a structured interpretive synthesis of 2023–2026 research and major policy frameworks, the paper argues the three pillars are jointly necessary because each addresses a distinct, empirically documented failure mode, and that portable student-facing guidelines can support responsible use by externalizing metacognitive prompts that unguided AI use tends to weaken. The framework is distinctive in being student-facing, course-agnostic, and actionable at the point of decision, and it specifies five empirically testable propositions.

The Gap Between Policy and the Moment of Decision

Institutional guidance on generative AI has expanded rapidly, yet it overwhelmingly addresses administrators, instructors, and assessment designers rather than the students who make most decisions about AI use. Adoption has raced ahead of design: UK national surveys record undergraduate generative AI use rising from 66% in 2024 to 95% in 2026, with 94% reporting using it for assessed work. Students report that the boundary of acceptable use remains unclear precisely at the moment they must act — interviews describe an indistinct "line" between acceptable and unacceptable use, a difficulty rooted not in ignorance of policy but in the absence of anything that translates policy into a judgment about the task at hand.

Three Pillars, Ten Guidelines

The framework organizes student-facing guidance around three jointly necessary pillars:

1. Learning and Growth — guidelines that help students use AI to deepen learning rather than bypass it, supporting Self-Regulated Learning and Metacognition.
2. Ethics and Integrity — guidelines addressing Academic Integrity, attribution, and the honest disclosure of AI use.
3. Awareness and Safety — guidelines covering critical evaluation of AI output, privacy, and awareness of the limits and risks of generative tools.

The argument is that the three pillars are jointly necessary because each addresses a distinct, empirically documented failure mode of unguided AI use. Because unguided use can weaken the metacognitive prompts that support responsible decisions, the portable set of guidelines works by externalizing those prompts into explicit, actionable checks at the point of decision.

Distinctive Positioning

Compared with policy guidance, institutional frameworks, assessment scales, and AI literacy curricula, the framework is distinctive in being student-facing (written for the person making the decision), course-agnostic (portable across disciplines and assignments), and actionable at the point of decision (translatable into a judgment about a specific task). This directly addresses the documented gap between institutional policy and the student's moment of choice. The paper specifies five empirically testable propositions to guide evaluation of the framework's effectiveness.

What this means for practice

• Learners. Use the ten guidelines at the point of decision, where the boundary of acceptable use is unclear, and treat them as explicit checks that restore the metacognitive prompts unguided AI use tends to weaken.
• Learners. Have a route to human help when a task's acceptability is genuinely ambiguous — Guideline 7 assumes an instructor who can be asked — rather than resolving the ambiguity alone.
• Educators. Circulate student-facing guidance alongside institutional policy, since existing policy speaks to administrators, instructors, and assessment designers while students make the frequent, independent decisions.
• Instructional designers. Embed the guidelines as a course-agnostic decision aid and expect the largest benefit for students with the least prior AI exposure, testing that directly rather than assuming it.
• Instructional designers. Evaluate the framework with behavioral or artifact-based measures wherever both are available, because self-reported AI use is vulnerable to social desirability bias.

Limitations

• The framework is conceptual and unvalidated and reflects one interpretive reading by a single author; a different reading could yield a different set or number of pillars, and the synthesis was interpretive and structured rather than systematic, with no claim of exhaustive coverage.
• It has not been validated through expert consensus, student consultation, or classroom trial, and its five propositions are proposed for testing rather than tested.
• The survey evidence underpinning the problem statement is drawn largely from the United Kingdom and the United States, and the interview evidence from a single Australian institution (19 students and 12 teachers in Corbin et al., 2025), so applicability in other systems is untested.
• The framework assumes conditions that will not always hold — lawful access to at least one generative AI tool, an instructor who can be asked under Guideline 7, and assessment that still involves work produced outside supervised conditions — and it has little to add where AI use is wholly prohibited and invigilated or wholly required; it also leaves institutional procurement, tool selection, and the environmental and labor questions of AI supply chains unaddressed, and its ten guidelines may exceed what students hold in mind under time pressure.

Citation

Alsammani, A. (2026). A Student-Centered Framework for Responsible Use of Generative AI in Higher Education. EdArXiv preprint.