COPENHAGEN, Denmark – Anyone who has spent enough time around animals knows that they display their emotions in various ways. Horses pinning their ears back is a warning and dogs wiggling their bodies as a sign of excitement. Animals also express emotion through vocalization or the sounds that they make.

Researchers have been studying animal vocalizations for decades, attempting to decode the various sounds to understand what animals are communicating to each other and to other species. While it’s not possible to truly determine animal emotions like happiness or sadness, it is possible to understand if vocalizations are positive or negative and how intense the emotional valence is, said Élodie Floriane Mandel-Briefer, associate professor in the department of biology at the University of Copenhagen, Denmark. For example, animals can be calm or excited at the positive end of the scale and bored or stressed at the negative end.

“The way we assess (vocalizations) is to place the animals in contexts that are very clearly positive or negative, and then we record them,” she continued. “We also often measure heart rates to understand the intensity of the emotion or look at their behavior to validate if indeed the context seems to be positive or negative for them.”

In studies investigating pig vocalizations, negative contexts the animals are exposed to include castration, isolation and missed nursing. Positive contexts include huddling, social reunion and nursing. Other studies just observe and record sounds from animals as they occur naturally in their day-to-day life.

In the lab, Mandel-Briefer said that the team analyzes the sounds and extracts several parameters from the calls like duration, frequency and amplitudes, exploring how they change over time. Studying these parameters alongside understanding the context in which the animal made the sound helps researchers understand the meaning.

“We can even go one step further, which is to ask the animals if we are correct (about the meaning of the sound),” she said. “We use big speakers and hide the speakers somewhere and play the recorded sound to the animal. Then we can see if they react the way we would expect them to based on the assumed meaning of the sounds.”

Now, Mandel-Briefer and her team have turned their attention to using AI to decode animal sounds. In a recently published study, the team developed a machine learning model that was successfully able to distinguish between positive and negative emotions in seven different species, including cows, pigs and horses. The model, trained on thousands of vocalizations from animals, can accurately distinguish between positive and negative emotion 89 percent of the time. Interestingly, key predictors of positive or negative emotion among the species studied were consistent, indicating that emotional expression could be an evolutionarily conserved system.

“We’re hoping that (AI models) can be used as a tool for farmers to know about the welfare of the animals because nowadays there are not systems, at least that I know of, that track the emotions of animals on farms,” Mandel-Briefer said.

She added that many people today believe that it is more important for animals’ welfare if they are experiencing positive mental health. While physical health is still important and should be maintained, experiencing positive emotions also supports animal welfare. Mandel-Briefer said the use of AI models to detect and track positive emotions on farms could be useful for farmers and allow them to adjust the environment to create more positive moods. Additionally, other researchers have been investigating the use of AI to detect and decode animal facial expressions.

Even without AI, humans can learn to understand the emotions of animals, Mandel-Briefer said. Without any training, people can determine if animal vocalizations are positive or negative and their intensity, with an accuracy greater than random chance. 

“Anecdotally, when I explain to students in my class what to pay attention to in the sounds, they really get much better (at identifying the meaning of animal sounds),” she said. “Farmers already know their animals pretty well, but if they pay attention to the sounds, they can get to understand the animals even better without any monitoring systems.”