This proposal investigates the role of feedback in communication networks, with the ultimate goal of providing architecture-level guidance in the design of robust and efficient communication systems. Many common communication situations are over inherently two-way channels, such as wireless networks, digital subscriber lines (DSL), and the Internet, even when the information transfer is only in one direction. Thus we can receive feedback from the other end of the channel, which we can use to improve the quality of communication. Even though feedback is present in many communication systems and is used in certain primitive forms, such as channel estimation and automatic repeat request (ARQ), the theory underlying its use has not been completely developed.We propose new approaches for several communication settings, such as single-user channels with memory, multiple-user channels with feedback, compound channels with feedback, and two-way channels, in which two or more users simultaneously exchange information through a shared channel. The proposed research is based on integrating new mathematical and engineering tools into the communication problems. In particular, we will integrate the idea of causality and directed information and use optimization tools from operational research to derive schemes that achieve that capacity.Overall, the proposed research aims to advance a theoretical understanding of feedback in communication and to give practical guidance for robust feedback communication systems. While positive results will provide a novel communication system design, negative results will prevent over-engineering and allow more confidence in simple and modular implementations. Since feedback and causality also constitute a pivotal concept in biology and economics, a deeper understanding of the role of feedback in communication will lead to a better understanding of the role of feedback in a broader interdisciplinary context.