New study finds AI dog personality algorithm could match you with your new ‘best friend’

A multidisciplinary research team specializing in dog behavior and artificial intelligence has developed an AI algorithm that automates the high-stakes process of assessing the personality of potential working dogs. They want to enable dog training agencies to more quickly and accurately assess which animals are likely to succeed long-term in careers such as assisting law enforcement or assisting people with disabilities. .

Personality tests can also be used to match dogs and humans, allowing shelters to assist with appropriate placements and reducing the number of animals returned because they are not a good fit with adoptive families.

Scientists from the University of East London and the University of Pennsylvania conducted the study on behalf of Dogbutter, a dog technology startup based in Miami, Florida. They published the results of their dog personality testing algorithm in their paper “An Artificial Intelligence Approach to Predicting Dog Personality Types.” scientific report.

The AI ​​algorithm utilizes data from approximately 8,000 responses to the widely used Canine Behavioral Assessment and Research Questionnaire (C-BARQ) to train the AI ​​algorithm. For over 20 years, the 100-question C-BARQ survey has been the gold standard for evaluating potential working dogs.

“C-BARQ is highly effective, but many of its questions are also subjective,” said co-principal investigator James Serpel, professor emeritus of ethics and animal welfare in the UPenn School of Veterinary Medicine. Ta. “By clustering data from thousands of surveys, we are able to adjust for outlier responses specific to subjective survey questions in categories such as dog rivalry and fear of strangers. ”

The researchers’ experimental AI algorithm works in part by clustering answers to C-BARQ questions into five major categories, ultimately forming the digital personality thumbprint that a particular dog receives.

These personality types are identified and described based on an analysis of the most influential attributes in each of the five categories. They include: excitable/obsessive, anxious/fearful, aloof/predatory, reactive/aggressive, and calm/cooperative.

Data points that feed into these ultimate clusters include behavioral traits such as “I get excited when the doorbell rings,” “Aggression towards strange dogs that come into my home,” and “I chase birds or chase them when given the opportunity.” It is included.

Each attribute is given a “feature importance” value. This essentially represents how much weight that attribute receives when the AI ​​algorithm calculates your dog’s personality score. “This is pretty remarkable. These clusters are very meaningful and very consistent,” Serpel said.

Dogvatar and his collaborators plan to conduct further research into the potential applications of canine personality testing algorithms.

“Alpha Pack Leader” Piya Pettigrew, Dogvatar CEO, said: “This algorithm can significantly improve the efficiency of the training and placement process for working dogs and could help reduce the number of companion dogs returned to shelters because they are not compatible. is a win for both the dogs and the people they serve.”