Skip to main content
Queen Mary Academy
  1. Queen Mary Academy
  2. Educators
  3. Innovation and Scholarship
  4. Innovative Pedagogies
  5. Centre for Excellence in AI in Education
  6. Blog
  7. Items

Beyond AI marking: putting pedagogy at the centre with LearnOuts

Luigi Ventimiglia, Reader in Economics and Finance, and Paulo Oliva, Reader in Electronic Engineering and Computer Science, at Queen Mary University of London, reflect on the development of LearnOuts, an innovative assessment and feedback toolkit designed to strengthen constructive alignment and support the responsible integration of AI in education. Drawing on their experience of developing and piloting the platform across four Schools, they explore how technology can improve feedback quality, reduce academic workload and support a more coherent learning journey. 

Published:
Paulo and Luigi

When we talk about AI in assessment, the conversation often starts at the end of the educational process: can AI help us mark faster, or generate better feedback? Our experience developing and piloting LearnOuts suggests that this is too narrow a starting point. 

Good feedback begins much earlier. It depends on what we want students to learn, how we structure teaching to support that learning, how clearly learning outcomes are communicated, how assessments are designed, and whether marking criteria genuinely reflect those outcomes. Only then do we reach marking and feedback. 

This broader understanding has shaped the development of LearnOuts. 

More than a marking tool 

LearnOuts is an in-house, scholarship-informed educational toolkit developed at Queen Mary around the principle of constructive alignment. Its purpose is to make the connections between learning outcomes, assessment, marking criteria and feedback explicit and usable for both educators and students. 

An educator can use LearnOuts from the beginning of the learning journey: planning a module and its teaching weeks, developing and sharing learning outcomes with students, designing assessment tasks and associated marking criteria and rubrics, and then using the same framework for marking and feedback. LearnOuts links assessment criteria directly to learning outcomes and feedback, enabling educators to provide structured and personalised feedback to students. Increasingly, this information can also be brought together to support reflection at module and programme level. This reflects the wider project aim of extending alignment beyond individual assessments and modules. 

AI can support parts of this process, but it does not define it. Depending on the educator’s preference, LearnOuts can operate without AI, provide light-touch assistance, or offer more extensive AI support. For example, AI can help classify learning outcomes using Bloom’s Taxonomy, suggest assessment formats and task verbs, support the construction of criterion-linked rubrics, and assist with drafting personalised, outcome-specific feedback. Crucially, these outputs remain editable and under academic control, allowing educators to review, adapt and validate AI-generated suggestions before they are released to students.

From efficiency to better feedback 

Of course, workload matters. Providing detailed, personalised feedback is particularly challenging with large cohorts. Across participating Schools, academics using LearnOuts reported reductions in marking time of around 40% on average, while individual implementations achieved savings of up to two-thirds. 

But saving time is only valuable educationally if feedback remains meaningful. 

Our student evaluation across three modules produced 77 responses. Ninety-six per cent of respondents reported reading their feedback carefully, suggesting a high level of engagement with the feedback provided. Compared with feedback received elsewhere, students frequently described LearnOuts-supported feedback as clearer, more detailed and more useful. Particularly important was their emphasis on feedback being personalised, actionable and transferable to future work. 

One student captured this particularly well, describing the value of receiving: 

“Personalised to me, not general marking” 

This points to something more important than efficiency: feedback should not simply explain a mark. It should help students understand what they have learned, where they need to develop and how they can use that understanding in their next assessment. 

From individual modules to programmes 

The next step takes LearnOuts beyond individual assessments. 

Modules can be brought together at programme level, allowing learning outcomes, assessments and related information to be considered through a programme dashboard. This creates opportunities to identify patterns, gaps, duplication and progression across a degree programme rather than considering each module in isolation. This visibility can help programme teams ensure that learning outcomes are appropriately covered, avoid unnecessary duplication and support developmental progression across levels of study. 

This programme-level perspective also aligns with Queen Mary’s 2030 Strategy and the move towards a programme-level approach to AI in assessment. By making the relationships 

between learning outcomes, assessment and feedback visible across modules, LearnOuts can support programme teams in taking a more coherent and strategic approach to both curriculum design and the use of AI in assessment. 

This is where we see considerable potential: the same framework that helps an individual lecturer design an assessment or provide meaningful feedback can also support conversations about curriculum coherence, progression and programme design. 

What have we learned? 

LearnOuts has now been implemented across four Queen Mary Schools, reaching more than 900 students across undergraduate, postgraduate and degree-apprenticeship provision. It has supported very different forms of assessment, including dissertations, quizzes, essays, presentations and group work. 

Perhaps our most important lesson, however, concerns the relationship between pedagogy and technology. 

The development of LearnOuts has been iterative and collaborative, involving feedback from both academics and students. This has helped ensure that technological development responds to authentic teaching and learning needs rather than driving assessment practice itself. 

That distinction becomes even more important as generative AI develops. The question is not simply how much of assessment and feedback AI can perform. It is where AI adds educational value, where academic judgement must remain central, and how technology can support a coherent learning journey. 

For us, that means starting not with AI, but with pedagogy.

Luigi Ventimiglia: School of Economics and Finance

https://www.qmul.ac.uk/sef/staff/luigiventimiglia.html 

Dr Paulo Oliva: School of Electronic Engineering and Computer Science 

https://www.qmul.ac.uk/eecs/people/profiles/olivapaulo.html 

 

 

Back to top