Digital Twins for Abundant Feedback: Novel Feedback Paradigms via Explainable Multilingual Natural Language Processing

Providing rich feedback to students is a powerful approach to supporting student learning. We will use natural language processing (NLP) approaches to develop digital twins of our assessment and feedback processes, capturing the professional expertise and judgment of our expert markers. In so doing we make abundant a previously scarce and expensive learning resource by providing repeatable, scalable, and instant feedback to students. Succeeding in this effort requires significant methodological advances in multilingual NLP, in terms of providing explainable feedback based on entire documents, in close synergy with educational research.
This unique convergence of multilingual NLP (PI1) and frontier models of engineering education (PI2) will open up new research questions regarding how students engage with and value feedback, as well as the way we provide that feedback, expanding our understanding of what constitutes good educational practice in engineering.
Feedback explainability ensures the validity and reliability of the assessment and feedback process, and it serves as a critical mechanism for establishing trustworthiness in the use of this approach. This represents a frontier for multilingual explainability techniques in NLP.