Abstract:

Ensuring equitable participation for students with invisible disabilities remains a persistent challenge. Many existing approaches rely on reactive support mechanisms that do not fully address the unpredictable nature of certain conditions, such as epilepsy, which affects more than 65 million people worldwide (GBD Epilepsy Collaborators, 2025). Beyond the physical risks of seizure events, students may experience persistent anxiety and reduced confidence that undermines their autonomy and sense of belonging. Current institutional responses are designed to manage events after they occur, rather than to anticipate them.

This paper presents a proof-of-concept study exploring an AI-enabled assistive ecosystem integrating three components: a trained seizure alert dog with scent detection capabilities, a smart sensor collar with inertial motion sensors, and a mobile alert system. When the dog detects seizure-associated scent and performs its trained alert behaviour, the collar automatically recognises this signal and transmits a notification, including the student's location — to a designated contact such as a caregiver or disability support officer. This enables timely intervention without requiring the student to self-report or interrupt their participation.

Technical feasibility of the collar system has been established and peer-reviewed (Brady et al., 2025). A supervised machine learning classifier achieved 92.4% accuracy under a cross-dog evaluation protocol, with 19 of 23 complete alert sequences correctly identified. This cross-dog generalisation indicates the system is not dependent on per-animal calibration, making it scalable for real-world deployment. A structured 13-week scent detection training programme further confirmed that a trained dog can reliably distinguish seizure-related from control samples under controlled conditions.

Framed within Universal Design for Learning (UDL), this work reconceptualises inclusive participation in higher education. While traditional UDL frameworks address barriers to content access, this ecosystem extends inclusion into student safety and physical presence, the conditions necessary for confident participation. The paper also examines institutional and ethical implications, including data governance, student agency, and the importance of technology complementing rather than substituting for human support.

The contributions are threefold: a novel conceptualisation of assistive technology as a multimodal human-animal-AI ecosystem; technically validated findings situated within an educational inclusion framework; and a critical discussion of implications for student autonomy, institutional responsibility, and assistive system design. Together, these suggest a pathway toward proactive assistive infrastructures that support safe, confident, and autonomous participation in learning.

Keywords:

Digital inclusion, assistive technologies, AI, education.