Generative artificial intelligence as an enabler of student feedback engagement: a framework

Synthesis: A conceptual framework paper from feedback-literacy leaders Boud, Dawson and Yan arguing that GenAI can be an enabler of student feedback engagement — but only when positively aligned with students' feedback literacy. Analyzing feedback across three stages (eliciting, processing, enacting), it proposes a cyclical self-regulation feedback model (feedback forethought → feedback control → feedback retrospect) grounded in an ecological perspective, as a pathway for students to productively engage with GenAI feedback while mitigating over-reliance and Academic Integrity risks.

The framework

The paper is a theoretical synthesis (deductive reasoning, four steps: literature search, analysis, adaptation, refinement) that adapts Chong's (2021) ecological framework to the GenAI context. Its core argument is ecological: GenAI affordances create opportunities for feedback engagement, but affordances do not guarantee engagement — students must possess the feedback literacy to perceive and act on them (van Lier's "match between the environment and agent").

GenAI as an enabler at each feedback stage

• Eliciting (seeking feedback). GenAI is on-demand, always accessible, and less intimidating than human instructors — helpful in large classes, time-poor contexts and power-hierarchical cultures. But effectiveness hinges on prompt quality; vague prompts yield generic, useless output. Constraints include version/access divides (subscriptions worsen the Digital Divide), English-centric training data, and data-privacy/intellectual-property concerns.
• Processing (making sense). GenAI reduces cognitive barriers (immediate clarification of academic jargon) and emotional barriers (less anxiety than human interaction). But output can be hallucinated, biased or general/overlapping, so students must exercise evaluative judgment and emotional reflexivity rather than blindly trusting GenAI (see AI Feedback Quality).
• Enacting (using feedback). GenAI enables iterative, cyclical feedback loops, self-monitoring toward self-defined goals and personalized non-judgmental suggestions. The risk is uncritical over-reliance and boundary confusion over what constitutes acceptable academic-integrity use.

The cyclical self-regulation feedback model

Adapting Zimmerman's (2000) self-AI Regulation in Education phases, the proposed pathway comprises three interwoven phases that cycle continuously:

1. Feedback forethought — setting feedback goals and planning strategically, based on learning needs and understanding of GenAI's capabilities/limitations.
2. Feedback control — self-observation and self-monitoring of one's prompts and GenAI's responses, assessing prompt and information quality, and refining interaction.
3. Feedback retrospect — self-reflection on the feedback received, interaction quality, strategies used and outcomes, feeding insights back into the next cycle's forethought phase.

These phases are interwoven with the eliciting/processing/enacting stages of the feedback process.

The interaction of GenAI context and feedback literacy

Two contrasting illustrative cases (IELTS writing with ChatGPT 3.5) ground the argument: Student A (low feedback literacy) used a vague prompt, received generic feedback, blindly trusted or over-copied output, and showed superficial engagement; Student B (high feedback literacy) used a specific, criteria-referenced prompt, exercised evaluative judgment (following up, cross-checking sources), monitored revisions and achieved deep engagement. This illustrates that feedback literacy specific to the GenAI context — including prompt engineering, evaluative judgment and self-regulation — must align with the GenAI environment for GenAI to be an effective enabler.

What this means for practice

• Instructors. Help students set individual feedback goals before they query GenAI, and model metacognitive planning, monitoring and evaluation in class, so that seeking feedback becomes a planned act rather than a reflexive one. Train prompt engineering and evaluative judgment alongside it.
• Instructors. Do not treat access as the intervention: design stage-specific Scaffolding — prompt templates for eliciting, evaluative-judgment tasks for processing, revision monitoring for enacting — because affordances do not by themselves produce engagement.
• Instructors. Build an explicit calibration step into the processing stage by requiring students to check at least one GenAI claim against an independent source before acting on it, since hallucinated or generic output is most dangerous to the students who trust it uncritically.
• Learners. Work the cycle deliberately: write a specific, criteria-referenced prompt, monitor and refine the interaction as it unfolds, then review what the feedback and your revision strategy achieved before beginning the next cycle.
• Researchers. Use the forethought–control–retrospect cycle as a coding and design framework, and test whether it predicts engagement in discipline-specific contexts and how engagement evolves over time.

Limitations

• This is a theoretical synthesis with no primary data: the framework is produced by deductive reasoning in four steps (literature search, analysis, adaptation, refinement), so it can propose relationships but cannot show that they hold.
• Its empirical grounding is two illustrative cases drawn from a single trial use of ChatGPT 3.5 on IELTS writing — one tool version and one task — used to illustrate low- and high-literacy behavior rather than to measure it.
• The cyclical model itself is untested: the paper offers no evidence that feedback forethought, control and retrospect predict engagement or learning, and the pathway is not yet validated against feedback engagement in real courses.

Citation

Zhan, Y., Boud, D., Dawson, P., & Yan, Z. (2025). Generative artificial intelligence as an enabler of student feedback engagement: a framework. Higher Education Research & Development, 44(5), 1289–1304. (CC BY-NC-ND 4.0 — license flagged: non-commercial, no-derivatives.)