For those who dream of talking to their pet like Dr. Dolittle or who often wonder what is going on in its head as it stares at and barks nonstop at that forgotten slipper in the corner of the room, artificial intelligence (AI) may be closer to making that wish come true. The technology is already helping scientists investigate animal communication through sounds, identifying patterns that may reveal clues about what the animals are trying to express.
With the ability to analyze thousands of recordings and identify patterns that are difficult to detect by human ear, the technology may reveal information about emotions, behaviors, and interactions between different species. But translating animal communication the way we do with human language is still a challenge.
According to Rytis Maskeliūnas, a professor at the Faculty of Informatics of Kaunas University of Technology in Lithuania, the field is already the subject of research. The goal, however, goes beyond recognizing barks or identifying when an animal is agitated. Scientists want to discover what information sounds carry and how these messages relate to the context in which they are produced. Information from Phys.org.
For those who are in a hurry
- AI can identify acoustic patterns associated with the emotions and behaviors of dogs;
- A study analyzed German Shepherd barks associated with four states: happiness, anger, crying, and loneliness;
- The Dog2vec model was trained on more than 6,000 hours of bark recordings;
- Recognizing sound patterns does not mean understanding what animals are trying to communicate;
- Whales, dolphins, elephants and birds are also subjects of research into animal communication.
What do a dog’s barks reveal?
In one study, Maskeliūnas and his colleagues investigated whether dogs’ sounds could reveal information about their emotional state. To do so, the researchers compiled recordings of German Shepherds and analyzed four categories defined based on assessments by animal behavior specialists: happiness, anger, crying and loneliness.
The team divided the audio into frequency and duration components and extracted acoustic features to train machine learning models. The results showed that the sounds carry structured information, allowing the researchers to identify barks associated with anger with a relatively high level of reliability.
Your dog is trying to tell you something and you may not be understanding – Image created by artificial intelligence – Image: ChatGPT / Olhar Digital
The main finding, however, was not simply recognizing whether a dog was irritated or lonely. The study indicated that animals’ voices exhibit patterns that are not always perceptible to people. This does not mean, however, that the researchers have deciphered everything a dog is trying to communicate.
More recent models may expand this investigation. One example is Dog2vec, trained on more than 6,000 hours of bark recordings.According to Maskeliūnas, the system outperformed general-purpose models in some canine sound recognition tasks.
These tools can also combine audio with images, movement, the reactions of other animals, and information about the environment. Joint analysis makes it possible to identify relationships between sounds and specific situations, offering clues about canine behavior.
Why does identifying a sound not mean understanding it?
One of the main obstacles is distinguishing the identification of patterns from the understanding of their meanings. The AI may detect that a particular sound occurs repeatedly when an animal encounters danger or is cared for by its owner. This association alone does not prove what the animal is trying to communicate.
To confirm a hypothesis, scientists must observe animal behavior and conduct controlled experiments. One possibility is to present a particular signal to another individual and check whether the reaction matches the prediction made by the model.
Study analyzed German Shepherd vocalizations associated with four distinct emotional states – Image: milatiger/Shutterstock
Maskeliūnas compares this process to Sherlock Holmes’s work: the AI finds an important clue, but it is biological experiments that make it possible to verify whether the clue has been interpreted correctly.
The difference from human language helps explain the difficulty. Language models are trained on large volumes of text that already contain words, grammatical rules, and known meanings.In animal communication, researchers still do not know whether an equivalent symbolic system exists or how their signals are organized.
Therefore, it is not enough to feed an algorithm millions of recordings. Scientists also need to know what was happening when each sound was produced, who it was intended for, and how other animals reacted. Without this information, even a sophisticated model may find statistical relationships that do not correspond to a real message.
Which animals may have their communication deciphered?
Whales, dolphins, elephants, some primates, and certain bird species are among the promising candidates for this type of research. These animals exhibit a variety of forms of communication, often important to their social relationships.
Parrots, for example, can imitate different sounds, while crows are capable of making relatively complex decisions. Nevertheless, intelligence or the number of vocalizations alone do not determine which species will be easier to study. The availability of long-term recordings, accompanied by information about behavior, is also fundamental.
In the future, researchers may also be able to test artificial signals created based on patterns identified by AI. If an animal reacts to a synthetic sound in the same way it reacts to a natural signal, this may strengthen the hypothesis that scientists have understood part of its function.Even so, it will be necessary to verify whether the response is linked to the meaning of the signal or merely to acoustic characteristics, such as frequency and intensity.
How could this change our relationship with animals?
A more precise understanding of animal communication could help owners identify signs of stress, fear, and discomfort in their dogs, including when these signs are not evident in their behavior. Technology could also help assess animals’ housing and transportation conditions and monitor wild populations without directly interfering with their lives.
Read more:
- 6 science tips to help your dog live longer – Olhar Digital
- The silent signs your dog gives before becoming ill
- Why do Pinchers tremble? Phenomenon encompasses other breeds
The consequences may go beyond practical applications. If studies show that animal signals carry more complex information than was imagined, it will be necessary to reconsider what we understand by intelligence and communication.
This possibility also raises ethical questions. The more humans are able to recognize animals’ needs and states, the harder it will be to ignore them. For now, however, AI mainly functions as an investigative tool, rather than as a translator capable of talking directly with other species.
Matheus Labourdette is a writer at Olhar Digital.
Ana Figueiredo is a technology editor at Olhar Digital. She graduated in Journalism from the Federal University of Uberlândia (UFU).
