I helped teach 6.270 for three years, after taking the class as a freshman. For my first two years, I was a "TA" - I only worked during the contest itself. In 2001 I was an "organizer" - part of the team of students that prepares the contest. This last year was the most exciting. From February to December, I met with the other ten organizers to discuss everything from the circuits that would control the robots to the rules of the contest.
Among the organizers, I found that everyone had beliefs I disagreed strongly with, often so strongly that I couldn't even understand their point of view. So I had to work continuously to win allies on certain issues and evangelize my ideas without polarizing the group or wasting time. Making small compromises also helped me convince the group of things I felt strongly about.
My biggest political success was being co-designer of the contest table itself. The table was full of compromises: it was the normal size and shape, kept the (silly!) requirement that robots orient themselves at the start of a match, included the simplest possible active element (something people really wanted), and used the traditional task of "scoring-object collection" that has always been the goal of the 6.270 contest. However, these compromises were just intended to distract from the main idea; that the contest should be easy to follow and exciting for the audience, like any professional sport. So the scoring rules were simple: the robot with the most scoring objects won. These were large, brightly colored, balls, visible from the back of the audience. Towards the sides, where the robots collected their balls, the table sloped downward, so that there would be fewer borderline cases to judge. The starting positions were moved from the sides to the center of the table, to encourage interaction between robots. It was a simple, fun table, and I was really happy to see it used.
The contest worked. Despite the heaviness of the balls, many robots rolled them easily around the table, a few could push them onto the floor, and one robot was capable of launching them, catapult style, from one side to the other. Using the Pololu IR beacon, the better robots could detect their opponent's location and used this information to find unobstructed paths across the table. A few robots were very simple (one just spun in place), while a few were ridiculously complicated (one came apart into five independent vehicles). The robots employed several very different strategies toward the goal of obtaining balls and made the contest fun to watch.