AI Offers Potential to Eliminate Human Error and Bias from Animal Welfare Assessments

A researcher with the Western College of Veterinary Medicine says the use of AI to assess animal welfare offers an opportunity to eliminate human error and bias during animal welfare assessments.

The Western College of Veterinary Medicine and the University of Saskatchewan's Department of Computer Engineering, in partnership with 14 industry partners with support from the Natural Sciences and Engineering Research Council, are using artificial intelligence to develop tools to assess animal-based indicators of on-farm welfare, such as lesions detected on carcasses at slaughter.

"Automated swine welfare monitoring at the abattoir" was discussed as part of a forum held last month as part of the 2024 Banff Pork Seminar.
Dr. Martyna Lagoda, a Post-doctoral Fellow Swine Behavior and Welfare with the Western College of Veterinary Medicine, says AI technology is becoming more widespread and can be applied to animal welfare assessment.

Quote-Dr. Martyna Lagoda-Western College of Veterinary Medicine:

Measuring welfare of animals is a complex task and often involves subjective indicators which means there is a lot of potential for bias and human error.
Automation of these this process means the scoring of  welfare indicators is performed by a computer which eliminates human error and bias.Performing welfare assessments using an automated system at the time of slaughter in the abattoir means we can eliminate the issues associated with assessments performed on farm.

For example, these assessments are time consuming and as a result of  that they are often only performed once a year.They are also a risk to biosecurity.
We have different people entering the farm to perform these assessments.

Assessment of welfare at the abattoir using automated methods is made possible because there are animal-based indicators of welfare which are visible on the animal's carcass that inform on welfare throughout the pig's life and these are used as the scoring targets of an automated system.These indicators tell a story of welfare retrospectively and can be tracked back to different stages of the production chain.