We argue that the concept and goal of "transparency" in human-automation interactions does not make sense as naively formulated; humans cannot be aware of everything automation is doing and why in most circumstances if there is to be any cognitive workload savings. Instead, we argue, a concept of transparency based on and shaped by delegation interactions provides a framework for what should be communicated in "transparent" interactions and facilitates that communication and comprehension. Some examples are provided from recent work in developing delegation systems.