Never mind trouncing humans at video games and the ancient pursuits of chess and Go. Researchers have now harnessed the power of artificial intelligence to infer how pigs are feeling on the basis of their grunts.
Scientists believe that the AI pig translator – which turns oinks, snuffles, grunts and squeals into emotions – could be used to automatically monitor animal wellbeing and pave the way for better livestock treatment on farms and elsewhere.
“We have trained the algorithm to decode pig grunts,” said Dr Elodie Briefer, an expert in animal communication who co-led the work at the University of Copenhagen. “Now we need someone who wants to develop the algorithm into an app that farmers can use to improve the welfare of their animals.”
Working with an international team of colleagues, Briefer trained a neural network to learn whether pigs were experiencing positive emotions, such as happiness or excitement, or negative emotions, such as fear and distress, using audio recordings and behavioral data from pigs in different situations, from birth through to death.
Writing in the journal Scientific Reports, the researchers describe how they used the AI to analyze the acoustic signatures of 7,414 pig calls recorded from more than 400 animals. While most of the recordings came from farms and other commercial settings, others came from experimental enclosures where pigs were given toys, food and unfamiliar objects to nose around and explore.
The scientists used the algorithm to distinguish calls linked to positive emotions from those linked to negative emotions. The different noises represented emotions across the spectrum and reflected positive situations, such as huddling with littermates, suckling their mothers, running about and being reunited with the family, to negative situations ranging from piglet fights, crushing, castration and waiting in the slaughterhouse.
The researchers found that there were more high-pitched squeals in negative situations. Meanwhile, low-pitched grunts and barks were heard across the board, regardless of their predicament. Short grunts, however, were generally a good sign of porcine contentment.
“There are clear differences in pig calls when we look at positive and negative situations,” Briefer said. “In the positive situations, the calls are far shorter, with minor fluctuations in amplitude. Grunts, more specifically, begin high and gradually go lower in frequency.” According to the researchers, the algorithm correctly classified 92% of the calls as positive or negative emotions. With more recordings, the pig translator may be able to learn to distinguish a broader repertoire of emotions and shed light on the mental wellbeing of other animals.
While farmers tend to recognize that the mental health of animals is important for their wellbeing, the majority of animal welfare efforts focus on physical health. Briefer and her colleagues believe their algorithm can pave the way for new automated systems in the livestock industry that monitor sounds on farms and other sites to assess the animals’ psychological wellbeing.