New study finds machine learning can spot future guide dogs with 80% accuracy
When it comes to choosing a puppy who could grow up to become a dependable guide dog, even the most experienced trainers admit it’s more of an art than a science. But what if a computer could do better?
That’s exactly what a team of researchers from the U.S. and U.K. set out to test—by teaching an artificial intelligence (AI) model how to identify the makings of a perfect service dog. Their findings, recently published in Applied Animal Behaviour Science, suggest that AI may now be better at predicting future guide dog success than seasoned human professionals.
“This could be a real game-changer for organizations training assistance animals,” said Dr. Mohammad Amirhosseini, Associate Professor at the University of East London, who co-authored the study. “Not only could it save time and money, but it also reduces the emotional burden on dogs and trainers when a dog doesn’t make the cut.”
From Puppy Playtime to Predictive Modeling
The researchers partnered with The Seeing Eye, Inc.—the oldest guide dog school in the world—and several universities, including the University of Pennsylvania and the University of Nottingham. Trainers meticulously logged details about each dog’s temperament, energy, focus, and other traits throughout their development.
The AI model was trained using this data to recognize patterns that correlated with dogs who eventually graduated the program. When tested, the AI predicted successful service dogs with up to 80% accuracy—significantly outperforming trained humans.
Dr. James Serpell, a co-author from the University of Pennsylvania, noted: “We’re not replacing trainers. What we’re doing is giving them a very powerful new tool to help them make better decisions early on.”
Reducing Training Failures and Saving Resources
Training a service dog is costly, often exceeding $50,000 per animal. When a dog doesn’t make the cut, it’s not just a financial burden—it’s emotionally taxing for trainers and disappointing for families awaiting a match.
As TechRadar reports, the use of AI in early-stage selection could mean fewer heartbreaks and better outcomes. By identifying traits linked to future success, the system helps trainers focus their energy and resources where they matter most—potentially saving time, money, and sparing dogs from being removed from training and rehomed after developing strong bonds and routines.
If AI can catch patterns even experienced trainers overlook, it won’t replace human judgment—but it could become an invaluable partner in the world of animal training.
What’s Next for Canine Training and AI?
The project was a collaboration between institutions across two continents, including the University of Arizona and Canine Companions. It shows how emerging technology can work hand-in-hand with human expertise to improve animal welfare and training outcomes.
“This isn’t about replacing intuition—it’s about enhancing it,” Amirhosseini added.
The full study, Prediction of assistance dog training outcomes using machine learning and deep learning models, can be found on ScienceDirect.
