Engagement is a design problem, not a student problem

Audio briefing based on Student Voice Weekly issue #10.

This Week

This week, the episode discusses engagement design, AI feedback, and assessment change. New research suggests engagement gaps often point to institutional design. The main topics are grouped below by student voice practice, research, sector developments, archive context, and practical application.

Main Topics Discussed

Student Voice Practice

This week I was reading around a deceptively simple question: when engagement is weak, where should universities look first?

Research Spotlight

Sector Watch

From the Archive

Practical Application

Universities rarely use student feedback insight through one small central team.

Subscribe to The Student Voice Weekly: https://www.studentvoice.ai/blog/newsletter/

Transcript

Hi, and welcome to Student Voice Weekly. I'm Dr Stuart Grey, founder of Student Voice, and this week's theme is engagement design, AI feedback, and assessment change: why the gaps you are seeing are often the product of how we design learning, not who your students are.

Today I'd like to talk about a common reflex when engagement looks weak, or when assessment and feedback scores wobble. We reach for a student explanation. Students are time poor. They are disengaged. They are anxious. They are not prepared.

Sometimes that is true. But it is not the most useful starting point, because it pushes you towards motivational fixes: more nudges, more comms, more monitoring.

A more useful interpretation is this. Engagement is often telling you something about your design. The conditions you have created, the friction in the system, the way teaching is delivered, the way assessment is set up, and the practical reality of participating.

And the key thing is, design is something universities can actually change.

I want to cover the main story, one research finding you can use, two sector signals, what this means for student comments, and one thing to try this week.

Main story. A lot of what we label as engagement, assessment and feedback, and AI, is actually the same underlying experience. Students experience one joined up learning system. If that system is slow, unclear, inconsistent, or feels adversarial, you will see it show up as lower engagement, higher anxiety, more AI use, and repeated feedback complaints.

This is why small design choices matter. If a brief is vague, criteria are generic, and students cannot see what good looks like, they will look for clarity somewhere else. If feedback lands after the next deadline, students treat it as box ticking, because that is what we have designed.

On AI, universities often talk as if generative AI is an external force disrupting assessment. But what students are actually saying is simpler. They use whatever reduces risk, gives speed, and helps them interpret expectations. So if your system is slow or unclear, they will fill the gap. Increasingly, that includes AI.

So my judgement this week is: stop treating engagement, assessment and feedback, and AI as separate workstreams owned by separate groups. Students do not experience them separately. If your AI work is not also asking about brief clarity, feedback turnaround, and what happens when students ask for help, you are probably missing the mechanism.

Now, one research finding worth using. Sharma and Garg looked at engagement across 553 students. In plain English, they found that institutional variables explain more of the variation in engagement than personal background variables do. Their model explains around 30 per cent of the variance overall, and most of the explanatory power comes from institutional factors. Demographics were mostly not significant, apart from gender.

This does not mean background never matters. But it does give you permission to start your diagnosis with what you control. Instead of asking, why are these students disengaged, ask: where is the friction in our design.

When I say design, I mean the things that map closely to what students actually comment on.

First, teaching delivery. Do students know what is expected before the session. Is there a reason to attend beyond content transmission. Is the session structured so students can participate without guessing the rules.

Second, curriculum coherence and relevance. Students will tell you it feels disconnected, or not aligned to assessment. That is an engagement issue caused by course design, not a motivation deficit.

Third, course climate. Do students feel comfortable asking questions. Do they experience staff as approachable. Engagement drops quickly in environments that feel judgemental or closed.

Fourth, support that is usable. Not whether support exists in theory, but whether it works in real student lives. Can they access help at the moment they need it. Are responses timely.

Fifth, the practical conditions of participation. Timetables, online systems, access to materials. These are design factors that can look like an engagement problem.

So if you are seeing weak engagement, do not jump straight to an attendance policy conversation. Make sure you do a quick design review first. What barriers have we created. How coherent is the experience between teaching, assessment, and feedback.

Now, a research signal linking AI and feedback. Henderson and colleagues analysed a large multi institution dataset, including open comments. Students reported using generative AI for speed, accessibility, and low stakes iteration. That last phrase matters. It means trying things out and getting unstuck.

Trust is the key point. Around 90 per cent of students rated teacher feedback as trustworthy, and around 60 per cent rated GenAI as trustworthy.

So uptake is not the success metric. Students will use a tool they do not fully trust if it is fast and available. They are making a risk trade off. They still want teacher judgement, especially when tasks are high stakes or ambiguous.

If you are running an AI feedback pilot, make sure your questions are practical. What did students use it for. Did it change what they submitted. Where did it mislead them. What did they still need from a teacher.

Now, sector watch.

First, the University of Glasgow has relaunched its Assessment and Feedback Practice Enhancement Tool. The signal here is not the tool itself. It is the change mechanism. Many universities are good at collecting evidence and discussing results. The failure point is translating evidence into structured staff review, support, action, and tracking.

The key thing is a clear route from student voice to staff action. That is design again. A designed pathway, rather than hoping action happens.

Second, Wonkhe's work on AI and assessment links late feedback, unclear expectations, and AI use. One point that should make people pause is that a significant minority of students say they have submitted work they could not fully explain. That is not only an academic integrity issue. It is an assessment design and learning issue. It tells you some students are operating beyond their confidence, and sometimes beyond their understanding, in a high stakes, time pressured system.

The mistake is to respond with only tighter rules and detection. Clear rules matter, but you also need to audit the conditions driving the behaviour: feedback timeliness, clarity of briefs, and whether students get opportunities for formative practice and safe feedback.

Now, what this means when you are looking at student comments.

The risk is that all of this gets grouped into broad headings like engagement, assessment and feedback, or AI, and the response becomes generic. Instead, make sure your analysis points to a lever someone can pull.

Here are three distinctions that help.

First, separate timing from quality. Late feedback is a process problem. Unclear feedback is a quality problem. They need different fixes.

Second, separate clarity of expectations from fairness of marking. If students say they did not know what you wanted, that is briefs, criteria, exemplars, and assessment literacy. If they say it was unfair, that is calibration, consistency, and moderation.

Third, separate AI rules from needing help. When students say, you told us not to use AI but did not tell us what is allowed, or you did not give us enough support to do it without, that is often a request for clearer boundaries and faster guidance. It is not automatically a confession of cheating.

Also, listen for coherence comments. Lectures do not match seminars. Seminars do not match assessment. Assessment does not match feedback. That design problem quietly drives disengagement because students stop believing effort maps to outcomes.

One thing to try this week. In your next meeting on engagement, assessment, or AI, do a ten minute design audit with three questions.

Question one: where are students using speed substitutes. Where are they reaching for anything fast, including AI, because our system is slow.

Question two: where are students experiencing ambiguity. Briefs, criteria, what good looks like, and the boundary on AI use. Pick one module where ambiguity is high and decide how you will reduce it.

Question three: where is the loop broken between student comments and staff action. Who sees the comments, when, in what format, and what happens next.

If you can answer those three questions honestly, you often do not need a huge engagement strategy. You need a few practical design fixes, and a better mechanism for acting on what students are already telling you.

That is it for this week. The full set of links and summaries is in Student Voice Weekly.