AI-Augmented Cognitive Learning for Adaptive Anti-Scam Training in Older Adults
Older adults remain disproportionately targeted by increasingly sophisticated online scams, yet existing cybersecurity education typically adopts a uniform instructional model that overlooks the diverse characteristics of this population. We introduce an adaptive anti-scam training approach that centers on multi-dimensional persona modeling and explainable, personalized coaching. The system constructs detailed learner profiles incorporating cognitive ability, Big Five personality traits, and technology familiarity, then maps these characteristics to targeted adaptation strategies such as scam-type prioritization and personalized explanation styles. By integrating a scam detector, XAI-based cue extraction, and LLM-driven instructional generation, the framework delivers tailored guidance with explanations calibrated to each persona’s cognitive and psychological needs. The system is implemented as a web application to ensure accessible, real-time training for older adults. To evaluate this framework, we plan a randomized study in which older adults complete a pretest, engage in either generic or persona-based AI coaching, and then take a post-test with unseen scam messages, allowing us to quantify the added value of adaptive, persona-aware training. This approach aims to create more accessible, context-aware learning pathways that better support older adults’ ability to identify and resist online scams.
Publications
In Preparation