A team of researchers specializing in canine behavior and artificial intelligence has developed an AI algorithm aimed at streamlining the assessment of potential working dogs’ personalities. The initiative seeks to assist dog training agencies in swiftly and accurately identifying animals with the potential for long-term success in roles such as law enforcement support and aiding individuals with disabilities.
The newly devised personality test also holds promise for facilitating dog-human matchmaking, thereby aiding shelters in making appropriate placements and consequently reducing the number of returned animals due to mismatches with their adoptive families.
Conducted by scientists from the University of East London and the University of Pennsylvania on behalf of Dogvatar, a canine technology startup based in Miami, Florida, the research culminated in the publication of their findings in a paper titled “An Artificial Intelligence Approach to Predicting Personality Types in Dogs,” featured in Scientific Reports.
The AI algorithm draws insights from nearly 8,000 responses to the widely utilized Canine Behavioral Assessment & Research Questionnaire (C-BARQ) to refine its predictive capabilities. For over two decades, the comprehensive 100-question C-BARQ survey has served as the benchmark for evaluating potential working dogs.
James Serpell, co-Principal Investigator and professor emeritus of ethics and animal welfare at the UPenn School of Veterinary Medicine, explained, “C-BARQ is highly effective, but many of its questions are also subjective. By analyzing data from thousands of surveys, we can mitigate the impact of outlier responses inherent in subjective survey questions, particularly in areas such as dog rivalry and stranger-directed fear.”
The experimental AI algorithm developed by the research team functions by clustering responses to C-BARQ questions into five primary categories, thereby crafting a digital personality profile for each dog.
These personality types, delineated based on the analysis of influential attributes within each category, include “excitable/attached,” “anxious/fearful,” “aloof/predatory,” “reactive/assertive,” and “calm/agreeable.”
Key data points informing these clusters comprise behavioral attributes such as “excitability upon hearing the doorbell,” “aggression towards unfamiliar dogs visiting the home,” and “propensity to chase birds when given the opportunity.”
Each attribute is assigned a “feature importance” value, which underscores its significance as the AI algorithm computes a dog’s personality score. Serpell remarked, “These clusters exhibit remarkable coherence and meaning.”
Dogvatar and its research collaborators plan to explore further applications for their dog personality testing algorithm.
“This represents a significant breakthrough for us,” noted Piya Pettigrew, CEO of Dogvatar. “The algorithm holds the potential to enhance efficiency in the training and placement of working dogs, thereby reducing the incidence of companion dogs being returned to shelters due to compatibility issues. It’s a win-win for both dogs and the individuals they serve.”
