Recent research in Learning Sciences has drawn attention to the affordances of enabling students to learn about scientific phenomena through a complex systems lens. In this study, we adopt a complex systems perspective in helping students learn about a multi- level phenomenon, artificial selection, by using an agent-based participatory simulation Bird Breeder. Our goal is to investigate how design revisions to the activity in the form of 1.) Explicit representations of students' shared experiences, 2.) Access to an underlying third level of alleles, and 3.) Transparent rules of interaction facilitated abstraction of population-level trends in terms of change over time. We draw on data from two iterations of an agent-based modeling curriculum for evolution as part of a design-based research study in three tenth grade biology classes in the mid-west. The findings hold implications for the design of participatory simulations in general, and ways to support meaningful learning about complex multi-level phenomena in particular.