Artificial intelligence is increasingly more capable of making sense of the world. This helps behavioural researchers who wish to automate their time-consuming behavioural annotation. Existing systems mainly measure basic behaviour such as attitude, location and activity, or are specific and hard to adapt. This PhD thesis investigates recognition of rodent behaviour using various machine learning methods. It presents an automated behavioural recognition (ABR) system that recognises 10 specific behaviours and is now used in more than 300 academic and industrial laboratories worldwide. It describes three aspects of behaviour that make recognition with current models difficult. This PhD thesis puts forward a hybrid tool for annotating new behaviour that makes it possible to annotate with AI assistance. Finally, this thesis looks ahead with fully data-driven analysis of behavioural effects, and reflects on the ethical implications of automated behavioural monitoring. The PhD thesis thus shows that combining the power of AI and human expertise can lead to responsible, usable tools that are useful to society.
Elsbeth van Dam was born in 1969 in Doetinchem. After a year at the Dutch Circus School, she enrolled in a study programme in Mathematics at Utrecht University in 1988, but graduated in Philosophy of Mind at Radboud University in 1994. Her professional life began at Human Inference in Arnhem, a company specialising in name and address recognition. In 2004, she joined Noldus IT in Wageningen, where she worked on high-throughput applications for collecting and analysing behavioural data and, from 2007, on developing algorithms for behavioural recognition. In 2018, she enrolled at Radboud University as an external PhD candidate. Not long after, she became AI team leader at Noldus IT and was made responsible for developing the new generation of behavioural recognition tools. The common thread in her life has been understanding how living beings perceive and interpret the world around them and how this can be mimicked using AI.