Abstract:

The rapid adoption of Generative Artificial Intelligence (GenAI) in higher education feedback practices has often outpaced the development of well-informed pedagogical design principles. This has created uncertainty about accountability, trust and the role of human judgment in feedback practices. In response to these challenges, the Feedback and Learning: AI-Assisted & Reimagined (FLAIR) project aims to develop a custom chatbot designed as a human-in-the-loop AI feedback assistant that supports, rather than replaces, tutor expertise.

This paper proposes a framework for designing GenAI-assisted feedback systems that engage users, provide pedagogically informed guidance, and enhance the usability and reliability of AI-supported feedback tools in higher education. The framework is empirically grounded in qualitative data collected from student focus groups as part of the FLAIR project, including students’ perceptions of effective feedback practice across Science, Technology, Engineering, Mathematics and Medicine (STEMM) disciplines and their views on the usefulness and limitations of GenAI-assisted feedback. The proposed design framework comprises the following principles: pedagogical role, collaborative design, supportive persona, behavioural boundaries, knowledge base, transparency and quality assurance.

Central to the framework is the conceptualisation of the feedback bot as a pedagogical assistant rather than an autonomous assessor. The custom bot aims to scaffold feedback clarity, structure and engagement while maintaining tutor expertise and relational presence with students. Another design principle for the bot emphasises collaboration between academic researchers and the Information and Communication Technologies (ICT) team, informed by students’ perspectives. This partnership ensures the system’s technical architecture aligns with educational objectives, institutional infrastructure, and responsible AI practices. The paper outlines the design workflow, briefing documentation, and multiphase piloting used to evaluate feasibility and usability in real learning and teaching contexts. In addition, a key principle concerns the persona of the bot, which shapes user expectations and interactions. Rather than adopting the voice of an authoritative evaluator, the bot is framed as an academically grounded and supportive ‘critical friend’. This persona encourages constructive engagement and reflection, supporting staff to critically consider and develop their feedback practices. The bot’s behavioural boundaries are also clearly defined.

Transparency and disclosure form another pillar of the framework. For example, the bot prompts users with a disclaimer indicating that its responses are AI-generated and should be critically reviewed. The framework also integrates mechanisms to mitigate hallucinations and inaccuracies. By presenting a design framework grounded in empirical data and collaborative, pedagogically informed development, the FLAIR project demonstrates how carefully designed human-AI collaboration can enhance feedback quality while maintaining academic judgement and professional responsibility.

Keywords:

Bot, Collaboration, Design, Engagement, Feedback, Pedagogy.