Scientists Are Testing AI That Could Detect Pain in Goats Just By Their Facial Expressions

Detecting pain in animals isn’t exactly as easy as asking them how they’re feeling; they can’t talk, after all. There are ways of identifying signs of pain, however—animals may vocalize pain through unpleasant shrieks, or may not eat as much as expected. Measuring whether they’re in pain isn’t always straightforward though, and can be subjective, requiring decades of experience to make a judgement call about whether an animal is in distress.

Thankfully, for all its flaws, one area where artificial intelligence performs well is in pattern recognition. ChatGPT is fundamentally just really good at recognizing how to produce logical sentences after seeing many examples of them; feed Google Lens an image of a cat and it can find millions others, just by isolating all the unique characteristics of a cat.

Scientists believe the same idea can be applied to recognizing pain in the faces of animals. Decades of experience from veterinarians identifying pained faces could be ingested by AI algorithms to automate this process for farmers. The technology could eventually be applied to other non-verbal patients, particularly young children.

Phys.org reported this week on the work coming out of the University of Florida’s College of Veterinary Medicine. Researchers there focused on goats, filming the faces of 40 subjects, some that were in pain and others that were comfortable. They then fed the data into an artificial intelligence model that learned to distinguish pain by facial expressions alone.

The algorithm they created is said to be anywhere from 62% to 80% accurate at identifying pain in faces. More data would be needed to improve accuracy, but the implications for clinicians are significant if they can more easily detect pain in non-verbal patients without relying on gut feeling.

“It’s not just an animal-welfare issue,” said Ludovica Chiavaccini, a clinical associate professor of anesthesiology at the University of Florida’s College of Veterinary Medicine. “We also know animals that are in pain don’t gain weight and are less productive.” You could imagine farmers having some sort of program on a mobile device that would quickly let them evaluate large numbers of goats to see which ones need further evaluation without needing to stop and slowly assess each animal manually or wait until they notice severe problems.

It’s one of those areas where AI has obvious utility automating something that normally would require humans extensive time and experience to master. Computers are very good at ingesting and studying lots of information in a way humans just can’t easily.