This study investigates the effectiveness of an explainable-by-design model in comparison to a system based on a generic foundational model for livestock surveillance, with a focus on calving sign detection. Stillbirth in calves presents a significant risk for beef cattle breeders, highlighting the urgent need for surveillance systems that can detect signs of calving and promptly notify farmers. While foundational models trained on large-scale image and video datasets have recently shown promise in general video surveillance tasks, their applicability to domain-specific scenarios remains limited, particularly when intuitive and context-specific explanations are required, such as in calving assistance. To address this gap, we propose a model that incorporates the decision-making process of farmers into its architecture, thereby enhancing its interpretability while leveraging the structural strengths of a generic foundational model. Experiments using real-world livestock surveillance videos demonstrate that the proposed explainable model outperforms conventional foundational model-based approaches in detecting calving signs, suggesting its potential for practical deployment in precision livestock farming.