Emotions are expressed through facial expressions. Just by closely observing a partner's face, one can prevent minor arguments from escalating into world wars. Anna Zamansky, a professor at the University of Haifa in Israel, discovered that cats also read facial expressions to avoid conflicts. Like socially adept individuals, cats that effectively mimic the expressions of others tend to get along well with their peers.
Scientists are challenging themselves to read the facial expressions of animals like humans do. They have sequentially developed artificial intelligence (AI) that detects expressions starting from pets such as dogs and cats to farm animals like pigs, horses, and sheep. If they can read animals' expressions, it is expected that conflicts between pets and humans can be reduced, and livestock farming environments can be optimized.
◇Capturing muscle changes that affect expressions
Professor Zamansky announced in November of last year in the international journal 'Scientific Reports' that AI revealed cats have a higher sociality than previously thought. Cats express their feelings by wrinkling their noses or slightly raising their upper lips, and when their counterpart mimics them, 60% play together or groom each other. This was a form of connection through facial expressions.
The research team identified expressions by detecting how the positions of specific marked points on the facial muscles of cats changed in videos. This method has been validated in other animals as well.
Pia Haubro Andersen, a professor at the Swedish University of Agricultural Sciences, announced the development of AI that detects pain through facial expressions of horses late last year. The research team marked points on the horse's face that influence expressions. When experiencing pain, the points marked above and below the eyes come closer together, while the points near the nostrils move apart. When a horse feels pain, wrinkles appear around its eyes, and its nostrils widen (see graphic below).
Scientists have known for a long time that animals convey emotions through facial expressions like humans do. Charles Darwin, who established the theory of evolution, argued in his 1872 published work that expressions serve as a type of 'common language' among mammals.
Darwin based his theory on anatomy, stating that mammals share a large portion of the facial muscles used to make expressions. Humans share 38% of their facial expression muscles with dogs and 34% with cats. Horses also shared 47% of their expression muscles with primates.
Darwin's theory has been proven at the level of nerve cells. Researchers at the Max Planck Institute for Biological Cybernetics in Germany presented evidence in Science in 2020 that brain nerve cells are associated with facial expressions. They confirmed through animal experiments that facial expressions reflect emotions.
◇AI reads animal emotions better than humans
Until now, scientists have read emotions from facial images manually. They checked with their eyes how the marked points influencing expressions changed. AI has automated this process. When videos are input into the computer, it can immediately detect the emotions reflected in the expressions. Professor Zamansky from the University of Haifa announced that in 2023, AI determined whether cats were experiencing pain with an accuracy of 77% based on their expressions.
The AI that reads expressions can assist in new drug development. It can detect whether laboratory animals experience pain through their expressions after administering medication. Researchers at the University of North Carolina in the United States published findings in 2022 that AI read whether mice were in pain from images 100 times faster than humans. Therefore, AI is well-suited for deciphering large-scale animal experiment results. Scientists believe AI reduces the errors of manual work while allowing easier comparisons of experimental results from various laboratories.
Similarly, it can also assist in livestock farming. Brazilian scientists trained AI with 3,000 images of horses before and after surgery and medication for pain management. Later, it determined with 88% accuracy whether the horses were in pain based solely on facial images. AI was able to detect pain that veterinarians might have missed.
In the United Kingdom, AI is being used to improve pig farming environments. Researchers from the University of West England Bristol and Scotland's Agricultural College developed an AI system called Intellipig, which detects pain or stress from pigs' facial expressions. The AI distinguished individual pigs with 97% accuracy, surpassing humans. Pets like dogs and cats can also benefit from AI. For example, by determining how much a cat mimics another's expression, it is possible to predict which companions it will get along with.
Of course, there are limitations to AI reading animal expressions. There is a lack of training data compared to AI that reads human expressions. While there is an abundance of images showing human faces on the internet, there are few images depicting animals' emotional expressions. Scientists believe that if AI also learns other physical information, such as body temperature or movement, these limitations can be supplemented. Furthermore, if combined with AI that deciphers animal sounds, it may become possible to communicate with animals through smartphones in the near future.
References
Scientific Reports(2024), DOI: https://doi.org/10.1038/s41598-024-79216-2
Scientific Reports(2024), DOI: https://doi.org/10.1038/s41598-023-50383-y
bioRxiv(2022), DOI: https://doi.org/10.1101/2022.08.12.503790
PLoS ONE(2021), DOI: https://doi.org/10.1371/journal.pone.0258672
Science(2020), DOI: https://doi.org/10.1126/science.aaz9468