Biological collectives, ranging from social insects like ants, honeybees to vertebrate groups such as bird flocks, fish schools, have long served as a rich source of inspiration for designing and deploying artificial systems, including robotic swarms. However, a persistent challenge lies in bridging the gap between individual-level behavior and the emergent collective dynamics. Classical equation-based models often fall short in this regard, as the micro-to-macro link are hard to interpret or modify, since this link is not explicit. In this paper, we propose the rule-based modeling language Kappa for modelling biological collectives. Unlike approaches that directly model emergent behavior through equations, Kappa allows one to observe the collective behavior emerging from local, mechanistic interaction rules. These rules are both intuitive to interpret and easy to edit, allowing modelers to design and conduct in silico perturbation experiments at the level of individual agents. In addition, once written in Kappa, the collective dynamics can not only be explored through simulation, but also subject to advanced formal analysis techniques, such as model ab- straction or causal queries. We demonstrate our approach through a case study of thermotaxis-driven aggregation in juvenile honeybees. Specifi- cally, we investigate how heterogeneous compositions of agents influence aggregation at spatial areas with optimal temperature.