Can AI Tell If Your Goat Is in Pain? What the Latest Research Says
Whether you like it or not or have never used it, AI has a lot of potential. Can you imagine training artificial intelligence to recognize if your goat is in pain? We all know goats are pros at pretending everything’s fine. They can limp only when you’re not looking, act completely normal after a fall, and hide pain until it’s practically unbearable. So what if a machine could catch what we miss?
Two recent studies suggest that AI might have the potential to reveal “Automated Pain Recognition” …..with some training of course. One study focuses on goats, the other on sheep, but together they point toward a future where technology can help take some of the guesswork out of animal care. And if you’re a goat owner who’s ever wondered, “Is she just being dramatic or is she actually hurting?” this research may offer some reassurance that backup is on the way.
The first study, from my favorite goat hospital and by Chiavaccini et al., tested deep learning-based models for detecting acute pain in domestic goats using video footage. Researchers collected thousands of images from 40 goats hospitalized at the University of Florida Large Animal Hospital, some in pain and some not. They used a pre-trained convolutional neural network architecture (VGG-16)1 with a Support Vector Machine classifier to analyze facial expressions from the video stills. Pain status was determined by a validated tool called the UNESP-Botucatu Goat Acute Pain Scale.
Across different technical approaches, including splitting the dataset in different ways and testing how well the system generalized to new animals, the best models reached about 80% accuracy. That’s comparable to human-level performance in many areas of veterinary observation. Notably, even under the more stringent condition of subject-wise validation, where the model had to evaluate goats it hadn’t seen before, accuracy remained above 60%. The conclusion? AI might not be perfect, but it’s already a pretty solid assistant when it comes to spotting signs of goat pain.
The second study by Feighelstein et al. looked at sheep and posed a head-to-head competition: could a machine outperform human experts in evaluating pain? Using video footage of 48 sheep before and after surgery, the researchers tested AI against four veterinary professionals. The AI was trained to evaluate facial expressions using a machine learning pipeline powered by the CLIP encoder and a Naive Bayes classifier. This system only had access to static images, front and side views of each sheep’s face, while the humans used both facial scoring and the full behavioral assessment method known as USAPS.
Here’s where it gets even more interesting. The AI didn’t just match the vets in some areas. It beat them. When comparing facial expression scores (using the SFPES scale), the machine’s accuracy and diagnostic power were significantly higher. And when pitted against the more holistic USAPS method, which looks at behavior over time, AI still held its own. Although it didn’t significantly outperform humans in this category, the difference wasn’t statistically meaningful. In other words, the machine was just as good!
Both studies acknowledge the importance of validation and the complexity of interpreting pain, especially since animals can’t tell us how they feel. But they also point to something really exciting: we might be on the verge of tools that can make our job as caretakers more accurate, especially when it matters most.
Relevance for Us
For goat owners, this research opens the door to future tools that could assist in real-time monitoring. Imagine your barn camera alerting you when a goat might be in distress, or an app that can flag subtle changes in facial tension before you’d notice it yourself. While such tools are not yet available for consumer use, they’re no longer in the realm of science fiction. As the technology improves and becomes more accessible, we may have new allies in the barn, ones that don’t sleep, don’t miss a thing, and never assume your goat is just “being a goat.”
Quick Takeaways
- AI is getting better at recognizing pain in goats and sheep using facial images.
- In one study, AI outperformed human experts in evaluating sheep pain based on facial expression.
- In another, machine learning identified pain in hospitalized goats with up to 80% accuracy.
- These tools aren’t yet ready for everyday farm use, but they’re coming one day!
Citations
Goat Study: Chiavaccini, Ludovica & Gupta, Anjali & Anclade, Nicole & Chiavaccini, Guido & De Gennaro, Chiara & Johnson, Alanna & Portela, Diego A & Romano, Marta & Vettorato, Enzo & Luethy, Daniela. (2024). Automated acute pain prediction in domestic goats using deep learning-based models on video-recordings. Scientific Reports, 14. https://doi.org/10.1038/s41598-024-78494-0
Sheep Study: Feighelstein, M., Luna, S. P., Silva, N. O., Trindade, P. E., Shimshoni, I., van der Linden, D., & Zamansky, A. (2025). Comparison between AI and human expert performance in acute pain assessment in sheep. Scientific Reports, 15, 626. https://doi.org/10.1038/s41598-024-83950-y
- Convolutional neural networks use three-dimensional data for image classification and object recognition tasks. (Dive Deeper here.) ↩︎
Explore More Tags
ADGAADGA geneticsanimal welfareBrix testingcolostrum timingdairy goat breedingdairy goat caredairy goat managementdairy goat pedigreesdairy goatsdam-raised vs bottle-feddelayed milkingDHIDHIRearly rearingfermentationforage feedinggentle handlinggoat behaviorgoat colostrumgoat digestiongoat healthgoat nutritiongoat pregnancygoat psychologygoat salivagoat social skillsgoat stressgoat welfareherd improvementherd managementkid raisingLinear Appraisalmastitismaternal separationmilkingmulti-omicsnatural bufferingneonatal goat carepassive immunitypedigree analysisPTIrumen microbiomeSaanen breedSaanen goats