Pen, Paper, and Artificial Intelligence: A Hybrid Architecture for Academic Assessment in the Era of Generative AI

The rapid expansion of generative artificial intelligence (GenAI) is transforming theconditions under which academic knowledge is produced and assessed. Contemporary large languagemodels can generate coherent texts, arguments, and solutions that are often difficult to distinguishfrom human work, challenging the reliability of conventional digital assessment formats.This study explores an alternative strategy that does not rely on detecting AI-generatedcontent but instead redesigns the architecture of academic assessment. We propose a hybridhandwritten examination model combining controlled paper-based responses with post-examdigitisation and AI-assisted semantic analysis, while the instructor retains the role of final evaluator.The model was implemented in an undergraduate course where handwritten exam responseswere digitised and analysed using a large language model under structured prompting conditions. Theworkflow enabled automated semantic structuring of extended answers, generation of preliminaryanalytical feedback, and partial reduction of instructor grading workload while preserving humanoversight.The findings indicate that hybrid assessment architectures integrating analogue studentknowledge production with AI-assisted evaluation may offer a resilient and scalable approach toacademic assessment in the era of generative artificial intelligence. The study contributes to theemerging discussion on AI-mediated assessment by proposing a hybrid evaluation architecture thatintegrates handwritten knowledge production with AI-assisted semantic analysis under instructorsupervision.