Precision livestock farming (PLF) technologies are advancing rapidly, offering opportunities to monitor behavior at unprecedented scales and resolutions. However, ensuring that these data are meaningful for animal welfare requires a clear understanding of the behaviors being measured. Many behaviors commonly captured by sensors, such as feeding, drinking, or grooming, can reflect a range of internal states depending on context. For instance, grooming may reflect environmental cleanliness, social stability, frustration, or anhedonia, while drinking could indicate hydration or abnormal polydipsia. Without grounding these measures in behavioral research, PLF risks misinterpreting normal variation or missing signs of compromised welfare. Abnormal repetitive behaviors, often used as welfare indicators in traditional research, remain underutilized in PLF, likely due to their perceived rarity or subtle expression. To build systems that reliably capture such behaviors and use them for real-time welfare interpretation, we need rigorous and repeatable observations to produce robust ground truth datasets. At the same time, PLF tools themselves may alter the behaviors they aim to monitor or introduce new welfare risks, such as injuries, requiring caution in interpretation. Integrating behavioral science early in PLF development can lead to more meaningful and accurate welfare interpretations. This talk will present a roadmap for bridging behavioral research and PLF applications, using cattle behavior, including insights from years of data collection on abnormal behavior and 24-h patterns of behavioral development, as central examples.

cattle,abnormal behavior,animal welfare,animal behavior