IMPACT & KNOWLEDGE EXCHANGE

  • IMPACT STORY
    Positive change for children who tube-feed and their families
    1 Jan 2018 - 1 Jan 2025
    Lead contributors for this impact or knowledge exchange:
    • Hopwood N
  • IMPACT STORY
    Feedback nobody can use isn’t feedback: why feedback literacy is also a design problem
    Lead contributors for this impact or knowledge exchange:
    Picture a first-year UTS student at 11 pm, a few weeks into their first semester. A Canvas notification says their grade is posted, and they see the mark straightaway. Their tutor’s comments sit behind a small icon in the sidebar, this may include annotations that are inside the submission viewer, with the rubric being further down the page, and the suggestions from the (generative) AI writing tool they used (e.g., Copilot or Grammarly) are in another tab. They read their mark and go no further. Moments like this will be familiar to anyone who teaches with a learning management system (LMS). In the week after marks are released, many student questions are about the mark rather than the comments. A student asks where to find the feedback, or what “needs more critical analysis” actually means. Another student pastes the comments into a generative AI tool to have them explained. Tutors, meanwhile, write similar comments across dozens of submissions and wonder who reads them. Research suggests these are not isolated cases. In one large study, when students could see their mark before opening their feedback, 42% of feedback was never opened (Mensink & King, 2020). As generative AI makes feedback cheaper to produce, the challenge may shift from how much feedback students are given to whether they can locate it in the LMS, make sense of it and then act on it. Scattered feedback affects each student differently, which makes this an issue of equity, diversity and inclusion (Dickson-Deane, et al., 2024). Using feedback well often depends on knowing the unwritten rules of university: which comment matters most, what a phrase like “develop your argument” means, when it is worth emailing a tutor. First-in-family students, students from culturally and linguistically diverse backgrounds, including those studying in an additional language, students with disability, students with carer responsibilities and those working long hours may have less time and insider knowledge to piece together feedback spread across several places. When they do not engage, it can look like low motivation rather than a design issue. Students’ circumstances and expectations have also changed. Everyday apps have shaped expectations that information will be immediate, in one place and easy to use on a phone. Yet many LMS features still assume a student sitting at a desk, clicking through several pages. Many students now check their results in short gaps between work, study and care, often on a phone. LMS design needs to reflect how students actually live and learn today. Feedback that is easy to find and use is an inclusive design choice: what helps students with the least time is likely to help everyone. Feedback now reaches students through LMS such as Canvas, analytics dashboards and AI writing assistants. These tools are often designed to make marking efficient, not to help students find and use what is written. This matters because feedback is a critical checkpoint in learning. It comes after students have tried to apply what they know, and it is when they find out what they got wrong and can correct their understanding before the next task. When feedback is hard to find, that checkpoint can be skipped and misunderstandings may carry forward. Feedback is not only what we say but how it appears on the screen, and that design can decide whether a student ever reads it. Feedback is not only what we say but how it appears on the screen, and that design can decide whether a student ever reads it. Feedback literacy research, including influential work by David Boud and colleagues at Deakin’s CRADLE and UTS, has helped reposition students as active participants in feedback rather than passive recipients (Carless & Boud, 2018). In practice, though, universities often respond by focusing on the student, for example with an online module, a workshop or a few videos on using feedback. They regularly report how quickly feedback is returned but check far less often whether students actually use it. Promising tools already exist, including OnTask , UTS’s own AcaWriter and the University of Surrey’s FEATS, which brings a student’s feedback together in one place. Yet these remain separate add-ons or pilots, and in the platform students use every day, finding feedback can still take several clicks. We suggest that feedback literacy is also a design problem. Students do need skills to use feedback, but the learning environments around them can make this easier or harder. Recent CRADLE work makes a similar point: how students engage with AI feedback depends on both their feedback literacy and the context the technology creates (Zhan et al., 2025). For today’s students, that context includes the LMS and the AI tools around it, yet university guidance on generative AI remains uneven (Atif et al., 2026). AI can make feedback faster and easier to obtain, but whether students act on it depends on how it is designed. A recent study from the University of Queensland shows this clearly. Hassan Khosravi and colleagues tested three ways of giving AI feedback to more than 13,000 students (Alsaiari et al., 2026). When students had to ask the AI for help themselves, they acted on almost none of its suggestions. When they received a standard AI review, they acted on about 14%. When they were asked to pick the suggestions, they wanted to work on and then talk them through with the AI, they acted on about 26%. The AI was the same each time; only the design changed. The study has not yet been peer reviewed and covers one platform, but it suggests that how feedback is designed can shape whether students use it. Drawing on our fields of learning experience (LX) design and learning analytics, we see three practical starting points. First, test feedback for usability, as we would any other user experience (Dickson-Deane & Chen, 2018): can a student find it in Canvas, understand it and name one next step? Second, look at uptake as well as delivery, by checking whether students open feedback, return to it and use it in their next task. Learning analytics can make this visible, much as early alert systems already detect patterns of student engagement (Atif et al., 2020). When students co-designed such a tool with Monash’s Centre for Learning Analytics, they also asked for feedback from different sources in one place, with prompts to act (Jin et al., 2024). Third, design feedback for focus rather than volume, whether it comes from a tutor or an AI tool, as more feedback can quickly become information overload (Sharif & Atif, 2024): fewer, prioritised comments, timed to the next task, with a step that asks students what they will do next. If feedback cannot be used, it cannot do its job. Many of the barriers students face are design choices, not student failings. Some are built into the LMS. In Canvas, the mark is usually the first thing a student sees, while the comments sit behind an icon., and rubric notes, annotations and AI suggestions are spread across different places. Others sit in assessment design. Feedback often arrives with no link to the next task, and if the next assessment does not build on the last, students have little reason or opportunity to act on it. Each of these makes feedback harder to use, and each could be designed differently. Small design changes may make a real difference. In Mensink and King’s (2020) study, placing the mark inside the feedback file, so students had to open it to see their grade, reduced unopened feedback from 42 to 17 %. In a sense, this recreates what paper marking did by default, when the grade and comments came back together on the same page. Technology has, in some ways, made a simple process more complicated. Similar ideas could be tested at UTS: bringing feedback from different sources into one place, designing assessment so each task builds on the last, and asking students to choose one thing they will act on. None of these changes requires AI; they apply to any feedback, in any LMS. We welcome colleagues at UTS, and partners at CRADLE, Monash, UQ and in industry, to explore a pilot that tests these changes with students in a large first-year UTS subject and follows what they do with their feedback. Our hope is simple: that the next time a student opens Canvas at 11 pm, the feedback is as easy to find as the mark.