Teaching and Learning Development Grants

From Bottleneck to Breakthrough: AI‑Enhanced Feedback in Finance

High-quality, timely feedback is fundamental to student learning in finance, where assessments require nuanced evaluation of analytical reasoning rather than simple correctness. However, delivering such feedback is resource-intensive in large-enrolment subjects, creating tension between quality, consistency, and turnaround time. Emerging research suggests that AI can enhance formative feedback when guided by structured rubrics and human oversight, yet evidence remains limited in discipline-specific contexts such as finance. This project evaluates an AI-enhanced feedback approach designed to improve the quality, consistency, and timeliness of assessment feedback while maintaining academic rigour. Aligned with the University of Melbourne's Framework for Educational Excellence and Advancing Students and Education Strategy 2023–2030, the project will be piloted across finance subjects using authentic case-based assessments. Findings will inform evidence-based guidelines for integrating AI into assessment practices and support broader adoption of scalable, high-quality feedback across higher education.

AI‑Assisted Marking Rubrics & Prompt Design Agent for Assessment

FNCE90057 (Ethics in Finance) is a core postgraduate Winter subject with typical enrolments of 200–300 students; in Winter 2026 about 150 students are enrolled. The unit now uses a 2,500‑word individual assessment plus a case/essay final exam, so a detailed, clear and instructive marking rubric is essential to ensure fair, formative assessment and transparent expectations.

Our extensive literature review identifies persistent problems: rubrics are often applied mainly for summative grading, risk over‑standardization, content‑validity and inter‑rater reliability are problematic in authentic tasks, and rubric development imposes substantial staff workload (Bloxham et al., 2011; Humphry & Heldsinger, 2014; Sadler, 2009). Empirical work on Generative AI for rubric development is limited and discipline‑specific models for Business/Commerce are scarce (Li et al., 2024; Vatankhah et al., 2026).

We will develop and pilot a secure AI‑assisted rubric agent that drafts outcomes‑aligned criteria, performance descriptors and criterion‑linked feedback templates, constrained by university assessment and AI policy and Bloom’s taxonomy. Academics will review and finalise all outputs (human‑in‑the‑loop). The project aims to improve feedback timeliness and actionability, reduce rubric drafting time, enhance marking consistency and produce robust evidence on reliability and validity. This work aligns with the Faculty’s Advancing Students and Education Strategy by improving assessment quality, feedback for learning and inclusive digital capability, and is designed to be transferable across Commerce subjects, including quantitative units.

Automated Personalised Feedback System to Enhance Student Engagement

This project aims to enhance the student learning experience by developing an Automated- Personalised-Feedback System that provides timely and tailored feedback on and for learning. It addresses a consistent gap identified in student evaluations: lack of specific and useful feedback to guide improvement (Iraj et al., 2021). Personalised feedback has been documented to provide improved learning experience, engagement, and performance (Cavalcanti et al., 2021; Pardo et al. 2019). Our system enables detailed, rubric-aligned feedback to be delivered at scale within our faculty and, potentially, across the university. Aligned with the ASE Strategy, this project promotes ‘digital technology to boost feedback provision’ (ASE, p.17) and ensures ‘all students can access the academic support they need for learning regardless of subject and class size’ (ASE, p.17).

The system can be applied to practice activities and formal assessments across a wide range of question types. It generates personalised feedback that identifies students’ strengths and weaknesses against the assessment criteria and provides targeted recommendations for improvement. This supports both feedback for learning during practice activities and feedback on learning in formal assessments.

By combining automation with rubric-aligned personalisation, this project establishes a scalable and transferable approach to delivering meaningful feedback, enabling timely and actionable feedback to be provided consistently in large classes and across multiple subjects.

Providing Targeted Feedback to Activate Self‑Regulated Analytical Thinking

This project aims to address a significant challenge: Supporting students become active, self-directed learners who can monitor their progress, and sustain their learning over time, thereby improving academic performance and developing lifelong learning capabilities.

In redesign of Assignment 1 for Business Decision Analysis (MGMT20005), considerable work has been done over the last two years to develop a bank of case-based questions and implement them in online Canvas-based platform, in which students can apply analytic techniques to real-life problems. Whilst that helps ensure academic integrity in response to generativeAI and allows them to learn from mistakes by providing ‘automatic’ feedback, this project aims to go beyond this. It aims to enhance students' opportunities to translate assignment learning to final exam by activating/driving “cognitive self-regulation” behaviours through personalised tutor feedback, online automated feedback, non-assessed practice problem sets and feedback on reflective activities.

Therefore, a suite of innovative teaching and learning initiatives will be developed, to support students:

• recall techniques to address areas for improvement (rehearsal),
• connect techniques to apply in new situations (elaboration),
• organise their thinking to identify priorities (organisation),
• engage with techniques in more meaningful ways (critical thinking),
• feel prepared for final exam (exam preparedness).

GroupWriter

GroupWriter addresses a persistent challenge in tutorial teaching: how to enhance student group work, making it more inclusive, visible, and manageable - for students and tutors. Typically, participation is uneven, collaboration is difficult to coordinate, and valuable student thinking is lost because it is spoken between group members, written on paper, or never shared beyond a small table group. The use of personal devices/laptops in classes and the variation in classroom infrastructure prevents current systematic approaches to student presentations. GroupWriter aims to provide a shared digital workspace where students can collaborate visually in real time, organise ideas collectively, and present their work to the wider class through a common platform. This particularly benefits students who may find verbal participation challenging, by creating alternative pathways for contribution through writing, annotation, structuring, and visual interaction. At the same time, tutors will gain practical tools to manage in class presentations, monitor live group activity, and review logged individual contributions, making it easier to gauge engagement, support quieter students, and recognise participation more fairly. Accordingly, GroupWriter will encourage engaged learning and benefit the academic’s ability to provide tutorial-based formative feedback and assessment. Existing software options (e.g. Miro) do not provide the full range functionality.

HiLRA: Human‑in‑the‑Loop Research Agents for Authentic Assessment

The rapid growth of AI has intensified concerns about cognitive verification when the cognitive process becomes hybrid, i.e. students complete tasks with AI-agents, making it harder to determine what students authentically understand (Sotiriadou, 20261).

Building on authentic assessment research, this project introduces HiLRA (Human-in-the-Loop Research Agent): a harness-engineered AI framework that structures complex research workflows through human oversight at critical decision points. Situated in ECOM90025 Advanced Data Analysis, HiLRA enables graduate students to collaborate with AI agents within documented, verifiable processes.

The project aims to:

• develop practical curriculum materials for the structured design and use of AI agent in complex research tasks with human oversight;
• design and implement interactive oral assessments (IOAs) that support authentic learning within AI-assisted assessment; and
• examine the human-AI hybrid cognition and how it shapes individual and collective learning.
• demonstrate a cost-effective model for scaling HiLRA implementation.

Library of Interactive Learning Artefacts for Threshold Concepts

Quantitative reasoning subjects in FBE rest on threshold concepts such as sampling distributions, the bias–variance trade-off and omitted-variable bias. These are inherently abstract and resist conventional explanation. A robust mechanism for teaching such "invisible" objects is scaffolded interactive simulation (Garfield & Ben-Zvi, 2007; Usmeldi, 2026). Existing third-party libraries (Seeing Theory, StatKey) do not match the concepts, sequencing, and notation of commerce subjects and are not readily adaptable. Bespoke per-subject authoring has been historically infeasible on cost grounds; LLM-assisted authoring has changed this calculation by an order of magnitude. CIs have already developed working prototypes covering neural networks, CI vs PI geometry, and dummy-variable interactions, evidencing the workflow and its feasibility (see, for example, [Neural Network Playground]).

This project aims to develop, evaluate, and disseminate a library of scaffolded interactive learning artefacts aligned to the threshold concepts of CMCE30003 and ECON20003, together with a replicable LLM-assisted production workflow that other FBE academics can adopt for their own subjects. The work overlaps with content in other subjects, providing pathways for faculty- and University-wide dissemination and is the first phase of a wider programme.

The project advances the ASE strategy through active, inclusive, AI-resilient learning.

An AI‑Supported Case‑Based Smart Platform for Competition and Merger Analysis in Economics Teaching

The aim of this project is to develop an AI-supported interactive case platform that links real competition cases to core economic concepts, enabling students to engage in structured, applied, and policy-relevant learning. More broadly, the project supports innovation in how knowledge is delivered in the age of AI and contributes to broader curriculum development in Industrial Organisation (IO) – the study of market power in the economy -- and competition economics in response to Australia’s 2026 merger reforms and federal productivity roundtables.

The project responds to an important pedagogical shift. While teaching has traditionally focused on theory, techniques, and methods, students are increasingly drawn to problem-driven learning and to understanding how theory applies in real-world settings. With the growing availability of AI tools, the value of learning is shifting away from memorising content towards developing the ability to structure problems, ask the right questions, apply economic reasoning in unstructured and ambiguous contexts to inform policy or business decision-making.

The project also addresses a clear gap between how competition economics is taught and how it is practised. Real-world merger and antitrust analysis require integrating theory into structured, case- based reasoning in legal and policy contexts. Additionally, Australia’s merger reform and the Australian Competition and Consumer Commission’s call for stronger training in competition economics highlight the need to equip graduates with these applied skills.

AI‑Assisted Simulation as Authentic WIL Assessment

Aligned with the university’s Advancing Students and Education Strategy 2023–2030, this project responds to the growing challenge that widespread access to generative AI poses to traditional assessment practices by prioritising active, student-centred learning experiences that prepare students for complex professional contexts. We will trial the integration of AI-tools in Work-Integrated Learning (WIL) assessment design to support authentic, practice-based learning, aiming to:

a) investigate how this approach can reshape student engagement and authentic demonstration of learning, while strengthening students’ professional communication, reflective skills and readiness for employment, and;

b) generate an evidence-based framework for responsible AI integration in assessment and capturing students’ and staff’s responses to these emerging practices.

Research has shown the impact and value of authentic, real-time decision-making exercises on student learning (Bishop & Verleger, 2013), and recently it is becoming increasingly more common the use of AI-assisted tools for this purpose (Aster et al., 2025; Jones et al., 2025). However, assessment integration with such tools remains a significant barrier due to privacy risks and questions of acceptability within the academic community. Building on the University’s work in asynchronous digital observation, we will conduct a comparative study of two AI assisted simulation platforms embedded in assessment.