Parallel Session 2

Thursday 3rd September, 14:45 - 15:45

Room A - 20-minute Presentations

A reproducible data science driven approach to course size estimation at the University of Glasgow

Joseph Crispell, University of Glasgow

Abstract: Course size estimates are required for the University of Glasgow’s timetabling process. Building the next term’s timetable starts in February, almost eight months before student enrolment is completed. Course size estimates aim to estimate what course enrolment numbers will be, to enable accurate matching of course events to appropriate rooms.
The course size estimation process has recently transitioned from manually updating last term’s sizes, to an automated machine learning based approach. The automated estimation process is built using the open-source coding languages R and SQL, based on the UK government’s principles of Reproducible Analytical Pipelines.
The estimation model uses historic student enrolment data going back to 2013, summarising these data into features describing the flow of students between courses, historic course size trends, and course ownership information. These features are then used within an automated and robust modelling flow based on a Random Forest time-series regression model. The modelling flow splits the course features data into training/test datasets, uses automated searching to identify optimal model parameters, trains a Random Forest model with those parameters using the training data, and then generates course size estimates using the trained model with the test dataset. The modelling process was validated by generating course size estimates for previous terms and comparing the estimates to the known course sizes. The estimation model generates course size estimates alongside measures of uncertainty, to help stakeholders understand how much variation might be expected around an estimate.
The presentation will provide an overview of the course size estimation process and summarise how we regularly engaged with stakeholders to deliver model-based estimates for the 2026/27 term. We’ll also reflect on the value of reproducible data science practices in delivering trusted data processes within a higher education context.

Keywords: course sizes, machine learning, data science

 

Beyond Descriptive Metrics: Using Holistic Statistical Modelling to Inform Equitable Student Continuation and Completion

Fabio Arico, University of East Anglia

Abstract: Higher education institutions increasingly rely on continuation and completion metrics to evaluate institutional performance and student success. However, institutional reporting often remains dominated by descriptive or bivariate analyses that identify disparities without adequately explaining the pathways through which inequalities emerge. This paper argues that more holistic statistical modelling approaches can generate more actionable insights for institutional enhancement and targeted support.
Using administrative data from multiple undergraduate cohorts at a mid-sized UK university, this study examines continuation and completion across the student lifecycle, including progression between years of study and degree completion outcomes. The analysis focuses on variation across gender, ethnicity, disability, mature status, and socio-economic deprivation, while accounting for prior attainment and degree characteristics.
The paper adopts a layered empirical strategy. Descriptive comparisons are combined with Generalised Structural Equation Modelling (GSEM), alongside linear probability, probit, and logit models, to examine both direct and indirect pathways associated with continuation and completion outcomes. Rather than treating withdrawal as a singular endpoint, the study analyses how risk develops differently across stages of study.
Findings indicate that disparities are especially pronounced in first-year continuation and cumulative completion measures. Students reporting mental health-related disabilities, mature students, and students from more deprived backgrounds display systematically weaker outcomes, while several ethnic minority groups, particularly Asian students, demonstrate comparatively stronger continuation and completion patterns. Importantly, the modelling demonstrates that some disparities operate indirectly through academic attainment and failure thresholds, while others persist even after attainment controls are introduced.
The paper argues that institutional analytics should move beyond retrospective accountability reporting toward more nuanced modelling capable of identifying where and through which mechanisms inequalities emerge. The findings demonstrate how holistic analytical approaches can support more targeted and equitable institutional enhancement strategies, particularly in relation to transition support, wellbeing interventions, and proactive student success initiatives.

Keywords: Student continuation, Degree completion, Learning analytics, Educational equity, Student success

Room B - 20-minute Presentations

Student Survey; beyond the numbers, listening to voices.

Kieran McNally, Higher Education Authority

Abstract: The Irish National Student Survey (StudentSurvey.ie) represents a comprehensive approach to measuring student engagement, that supports long-term planning across public and private Higher Education in Ireland. It is co-sponsored by the Higher Education Authority (HEA), the Irish Universities Association (IUA), the Technological Universities Association (TUA), and the student’s union Aontas na Mac Léinn in Éirinn (AMLÉ).
In 2026, the Irish National Student Survey ran after a three-year hiatus for a period of strategic reflection and review. The review process aimed to deepen understanding of student engagement within the context of a rapidly evolving higher education landscape, as well as resolve concerns that were emerging from co-sponsors.
Notably criticisms of StudentSurvey.ie, reflected in declining response rates as well as decreased student and students’ unions engagement, included concerns that the survey lacked context, reducing diverse experiences at different institutions to simple metrics; the 2023 survey had allowed for over 60 questions that did not adequately address student priorities.
The 2026 survey redesign by contrast, had a shorter set of 36 questions and more open text response options that prioritised the learner and student concerns, facilitating new insights and compassionate analysis to improve both institutional effectiveness and the student experience. Key areas of inquiry include time commitments including commuting (associated with accommodation), employment caring; financial stressors, withdrawal consideration; sense of belonging; having someone to talk to; and perceptions of feedback processes. The new improved survey is now able to feed a more nuanced student voice back to each institution when it thinks about quality enhancement and curriculum design.
This presentation offers learnings on how the Student Survey redesign changed the Irish education system’s capacity for long-term planning and how it now facilitates an enhanced compassionate approach to understanding student engagement across diverse institutional contexts.

Keywords: Student Survey compassionate design analysis

 

Understanding Intellectual Stimulation: finding from institutional research project

Elena Zaitseva, Liverpool John Moores University

Abstract: Universities increasingly seek evidence that curricula promote curiosity, critical thinking and intellectual stimulation, yet capturing these dimensions at institutional level remains challenging. Although the National Student Survey asks students ‘how often their course is intellectually stimulating’, this measure provides limited insight into how students recognise intellectually stimulating teaching, whether perceptions differ across student groups, or how such experiences relate to broader indicators of educational quality.
This presentation reports findings from an institutional research project that introduced a bank of optional module evaluation questions focusing on intellectual challenge and stimulation. The study explored relationships between these questions and core evaluation measures, including teaching quality and overall module satisfaction, using correlation analysis. Free-text responses were analysed using Leximancer concept mapping followed by thematic analysis to develop a richer understanding of students' perceptions of intellectually stimulating teaching.
The findings demonstrate a strong positive relationship between perceptions of intellectual stimulation and both teaching quality and overall module satisfaction. However, some differences were observed across demographic groups, including mature and younger students, students from different IMD quintiles, and students with disabilities, suggesting that different aspects of intellectually stimulating teaching may be valued by different learners. Qualitative findings indicate that students value opportunities to question assumptions, engage with multiple perspectives, participate in discussion and authentic learning activities, and receive sufficient scaffolding to support deeper thinking.
The paper discusses the implications for institutional research, curriculum enhancement and quality assurance, arguing that targeted evaluation questions can provide rich evidence of educational quality.

 

Room C - Lightning Talks

Supporting Staff Through Pedagogical Transition - Practice Development in the Age of GenAI

Dr Lenia Kouneni, University of St Andrews, Dr Sharon Leahy

Abstract: The rapid integration of Generative Artificial Intelligence (GenAI) into higher education is a significant professional transition for academic staff and one requiring institutional responses that are both compassionate and evidence-informed. This lightning talk shares early insights from a year-long, differentiated Continuing Professional Development (CPD) programme on AI in teaching and assessment, launched at the University of St Andrews in summer 2025.
Grounded in practice development principles, the programme combined a curated resource hub, peer-learning workshops, and an Education Prize recognising innovative practice, with staff feedback used to refine delivery throughout the year. Around 100 staff engaged across these strands, with over 10 staff-authored blog posts and 7 podcasts featuring staff contributors, alongside nominations for the Education Prize.  The programme was conceived as an evolving environment rather than a fixed intervention, and these numbers signal genuine cultural traction; staff did not not just attend, but actively produced and shared practice.
This talk asks what counts as evidence of successful transition this early in a change process. We suggest that staff choosing to produce and share their own practice (through blogs, podcasts, and prize entries) may be as telling as attendance figures or satisfaction scores. Ultimately, it returns to the question we started with: what does "meeting staff where they are" actually require in practice and how do we know when it's working?

Keywords: GenAI; CPD; differentiated support

 

GenAI Use in Assessment at RGU: a Content Analysis

Dr Lindsey Masson, Robert Gordon University, Dr Rachel McGregor (RGU), Dr Elliot Pirie (RGU), Dr Mark Zarb (RGU)

Abstract: The AI Assessment Scale (AIAS) was introduced to offer a practical framework to support the appropriate and ethical use of GenAI in assessment design (Perkins et al., 2024) and a version of the framework was implemented at Robert Gordon University (RGU) in academic session 2025/26.  There have been limited empirical studies examining the impact of the scale on educators’ practice in respect to assessment design. RGU is strongly positioned to contribute to this evidence base thanks to the level of institutional consistency in assessment practice resulting from the introduction of an Assessment Brief template in 2022 used across all summative assessments. This has enabled an institutional approach to be taken towards the specification of appropriate GenAI use in Assessment Briefs, moving from a binary choice (authorised/not authorised) in 2024/25 to the introduction of RGU’s five categories of GenAI Use, modelled on the AIAS, in 2025/26. This study aims to explore the application of RGU’s categories of GenAI use in summative assessment. Assessment Briefs will be downloaded directly from Moodle (RGU’s virtual learning environment) for 1728 modules in 2025/26 and 1623 modules in 2024/25. Data from the ‘Key Information’ section of Assessment Briefs will be extracted and entered into the Statistical Package for the Social Sciences (SPSS). The percentage of all summative assessments using each of RGU’s categories of GenAI use in assessment will be presented by academic year, School, SCQF level and assessment type. Ethics approval was granted by the RGU School of Computing, Engineering and Technology Ethics Committee and the project was approved by RGU’s Deputy Vice Chancellor – Academic. Preliminary data will be available to present by September.

Keywords: GenAI, assessment

 

Offshoring the Student Voice: AI, Cognitive Displacement and the Erosion of Thinking in Higher Education

Dr. Candela Delgado Marin, King's College London, NA

Abstract: This paper argues that the current debate on generative AI in higher education has misdiagnosed the problem. The issue is not primarily academic integrity, but a shift in where thinking happens. Increasingly, students produce fluent, well-structured work that appears analytically competent, yet is underpinned by minimal cognitive engagement. What we are observing is not poor writing, but the offshoring of the student voice.

Emerging evidence suggests this is a structural development. In courses most exposed to AI, top grades have risen significantly, particularly in unsupervised written assessments, raising concerns about the extent to which grades still reflect independent intellectual effort. At the same time, institutional responses remain inconsistent and predominantly compliance-driven, with many universities lacking accessible policies and framing AI through detection rather than learning. This creates a paradoxical environment: students are both encouraged to use AI and required to prove that they have not relied on it, generating new forms of performative authorship.

I propose a three-layer framework. Cognitively, students outsource the labour that transforms abstract knowledge into situated argument, bypassing the friction through which thinking develops. Generationally, AI intensifies an existing pattern of surface-level production, prioritising fluency and correctness over depth. Pedagogically, institutions risk reinforcing this displacement by positioning AI as a substitute for core academic practices, despite warnings that such substitution may hollow out the educational process itself.

The consequence is the emergence of what I term thinkers’ anxiety: a reluctance to engage in sustained, independent reflection. If higher education continues to prioritise outputs over processes, it risks deprioritising its core function: not writing, but thinking.

Keywords: Cognitive displacement; Thinkers’ anxiety; AI-mediated learning; Surface-level criticality; Independent thinking

 

Competitors or colleagues? Supporting institutional pedagogical development through cross-institutional collaboration and co-creation

Dr Gordon Spark, University of Dundee, Dr Julie Blackwell Young, Abertay University. 

Abstract: In an increasingly challenging and competitive Higher Education landscape, this lightning session presents a provocation: true institutional success is best unlocked not through competition, but through informal, cross-institutional collaboration.

This claim is grounded in the presenters' experiences as part of the International Learning and Teaching Network (ILTN), and more specifically as part of a sub-group of that network dedicated to supporting institutional pedagogical development. While other network strands stalled, this cross-institutional group thrived. We will outline the structural and cultural features that allowed that progress, exploring how members created a safe, collaborative space to share expertise, mitigate time constraints, and alleviate—rather than multiply—workplace pressures.

This session bridges data-driven practice and the social aspects of co-creation and collaboration. We will demonstrate how our sub-group moved from initial mapping and benchmarking of our institutions’ collective educational development offerings to tangible sector outputs, such as presenting on learning excellence at Advance HE’s Teaching Excellence Symposium. We argue that these collective activities offer a replicable, collaborative model for evidence-based institutional enhancement.

Ultimately though, we contend that the group’s value far exceeds just its outputs. Its defining feature is a culture of compassion—a space to pause, connect, and support. By recognising each other as colleagues rather than competitors, we model a compassionate approach to sector-wide resilience, enhancement, and institutional success.

Keywords: collaboration, compassion, co-creation