Should every transcript and protein be faithfully reconstructed from data, or is it enough to include only those elements for which we are familiar? Or is it enough for the model to be metaphoric in nature, only loosely resembling the problem domain? When considering a biological system, it is important to consider what is the most appropriate level of fidelity. I will discuss these issues from the perspective of representational thinking. This will involve us to think about model biological systems not as they are observed, but as having an epistemological basis that corresponds to other types of systems. We will discuss both the benefits and drawbacks to this type of modeling endeavor. This approach also allows us to think about biological systems and associated processes in terms of universal behaviors. To demonstrate this, we will consider more exotic types of representations such as graphical, cellular automata, game-theoretic, and hybrid approaches (e.g. multiple models simultaneously).