University of Arizona scientists at the College of Veterinary Medicine are using multimodal data and machine learning to validate and improve the many ways people benefit from trained service dogs.
Work led by associate dean of research Maggie O’Haire examines the impact of service dogs for veterans with post-traumatic stress disorder (PTSD). The dogs’ training is designed to instill a sense of safety, and self-report survey research shows it works. O’Haire is validating those findings with objective data from veterans with PTSD – some with and some without service dogs.
Wrist and collar wearables track sleep, activity and proximity between veteran-canine pairs, and saliva samples from both provide data on stress hormone levels. Participants without dogs also provide saliva samples and wearables data. Combined, the data paints pictures of daily life for veterans with and without service dogs.
Raising and training a service dog takes considerable time and money. Costs range from $50,000 to $70,000 per dog, and only about half ultimately achieve certification. Research by associate professor Evan MacLean aims to raise that rate and lower costs by identifying biological and behavioral markers correlated with successful canine trainees.
MacLean gathers data from cognitive tests, observed actions, genetic analyses and other biological traits. Machine learning integrates those datasets across a growing roster of subjects to identify patterns that predict which puppies will be the best candidates for service training.
Ultimately, he hopes to also categorize candidates by aptitudes. Matching strengths to specific needs – sniffing out explosives vs. assisting someone with a seizure disorder, for example – helps optimize health and safety outcomes at both ends of the lead.