What if we had an AI-like system instead of representatives? Imagine this: "Pennsylvania-One" is an AI. Every day, from their smartphones or another device, PA citizens vote on issues, ranging from small to big. Using this data, "Pennsylvania-One" creates a model of what its constituents want, and then, using its vast knowledge of political theory, history, economics, etc., makes a decision. This decision can be double-checked by constituents or by some other "watchmen"-like individuals.
Of course, this would never happen (people trust people much more than machines), and it's more of a sci-fi idea, but it's an interesting thought nonetheless.