"With AI, we can monitor animal behaviour and welfare more effectively"
Researcher and poultry vet commit to animal welfare
Analysing movement in horses, monitoring chicken behaviour or recognising pain in dogs: These are all examples of research projects that seek to improve animal welfare using artificial intelligence (AI). These projects all come under the AI Animal & Welfare Lab, one of Utrecht University's 11 AI Labs. We speak to Mona Giersberg, researcher at the Faculty of Veterinary Medicine and coordinator of the AI Lab, and Willem Dekkers, poultry vet at Royal GD in Deventer. "Our ultimate goal is to jointly develop future-proof AI systems that respect animal welfare."
"We use a sort of QR code on the chicken, so AI can recognise animals individually on camera images from the barn", says coordinator Mona Giersberg enthusiastically as she shows the images to poultry vet Willem Dekkers, who has recently become a collaboration partner of the AI Lab. Giersberg is open to more new collaborations, both within the university and externally. "The condition is that it adds something in terms of animal welfare. The animal has to be at the centre."
How can AI contribute to animal welfare?
"With AI, we can analyse animal behaviour", says Giersberg. "We use behaviour as a measure of well-being. If AI detects abnormalities in behaviour, it could be a sign that something is wrong. In that case, we have to, for example, make changes to the animals' housing. So AI is a kind of signalling system, based on which we can change things that improve animal welfare. We know, for example, that dust baths are important for chicken welfare. If straw is too wet, AI will register that the chickens aren’t taking many dust baths. The farmer can then intervene and change the straw."
Are we already using AI for this in professional practice?
"We’re doing a pilot where we’re using a camera system to monitor broiler behaviour in different types of housing on four farms", says Dekkers. "By using AI to analyse the behaviour of different breeds of chicken, we discover what the differences are and what needs each breed has. Also, after a while, you know what’s normal and you can identify abnormalities caused by health or welfare problems."
What added value does AI bring for vets or livestock farmers?
Giersberg: "AI enhances and refines people's powers of observation. It’s impossible for a livestock farmer to monitor all their animals 24 hours a day. But that is possible with a camera system and AI. There are also certain signs and behaviours that humans don’t see but that AI recognises. This allows us to monitor animal behaviour and welfare more effectively."
Dekkers also thinks that there are definitely opportunities for AI in livestock farming. "It can make life easier for livestock farmers by taking over tasks and alerting them to abnormalities in animal behaviour. A good livestock farmer probably sees anomalies faster than AI but isn’t in the barn 24 hours a day."
What is the purpose of the AI & Animal Welfare Lab?
"We want to bring together university partners, civil society partners and businesses", says Giersberg. "For example, within the university we work with colleagues in computer science. We can learn a lot from their technical expertise on algorithms, for example, while they benefit from our knowledge of animal behaviour. And we’re also looking further afield, such as to companies’ R&D departments to develop new technologies together. By sharing knowledge and experience with each other, we will take AI for animal welfare forward. And, through these collaborations, we’re also laying the foundation for future research." Dekkers adds: "It’s important to involve external partners in your research. If you don't include animal owners, for example, then you’re developing AI systems for nothing."
What are the benefits to professional practice of collaborating with the AI Lab?
"Collaboration with the university is invaluable", says Dekkers. "A good example of this is the behavioural sensor that we developed. The sensor provides us with interesting data but translating this into animal welfare is complicated. For this, we need the expertise of scientists. Collaboration is also useful when developing the technology itself. For example, we can work with an external partner to test whether the sensor does what it is supposed to do within a scientific project. That increases reliability."
How do you approach AI responsibly?
"We have to remain realistic and cautious", says Giersberg. "AI applications are sometimes brought out quickly, in the form of simple apps, for example. Jumping to conclusions can be dangerous. Not only do we have to trust the systems, we also have to take the circumstances into account; you can’t rely on just one aspect. And, in the AI Lab, the interests of the animal must come first. Our ultimate goal is to develop future-proof AI systems that respect animal welfare together with stakeholders."