In this paper we present a learning design approach that leverages perspective-taking to help students learn about complex systems. We define perspective-taking as projecting one's identity onto external entities (both animate and inanimate) in an effort to predict and anticipate events based on ecological cues, to automatically sense the affordances of objects in the environment and take advantage of these affordances, and to understand the mental states of other individuals; an ability crucial for socialization and communication. We introduce one key construct; "phenomenological connectors", which are essentially embodied perspective-taking activities across micro and macro levels of a system. Phenomenological connectors enable the learner to step inside the system at various levels, thereby having first-person experiences of multiple agents and components, as they attempt to make sense of the system. We consider agent-based modeling activities as exemplar of this design approach. We argue that agent-based models (ABMs) present unique perspective-taking challenges, and learning opportunities, to learners. Informed by the learning design approach proposed, we present a curricular agent-based modeling unit on Particulate Nature of Matter (PNoM) and present data from the implementation of this unit in two 10th grade science classes. The discussion of data focuses on how students make sense of their first-hand experiences in a diffusion of odor experiment, where students collectively act as sensors to track the diffusion of odor, by building and reasoning with agent-based models, and taking perspectives of different agent- and aggregate-level elements of the system.