As the skills that constitute literacy evolve to accommodate digital media, computer science education finds itself in a sorry state. While students are more in need of computational skills than ever, computer science suffers dramatically low retention rates and a declining percentage of women and minorities. Studies of the problem point to the over-emphasis in computer science classes on abstraction over application, technical details instead of usability, and the stereotypical view of programmers as loners lacking creativity. In spring 2003, Georgia Institute of Technology trialed a new course, Introduction to Media Computation, which teaches programming and computation in the context of media creation and manipulation. Students implement PhotoShop-style filters and digital video special effects, splice sounds, and search Web pages. The course is open only to non-computer science and non-engineering majors at Georgia Tech, such as liberal arts, management and architecture students. The course is supported through the use of a Web-based collaboration environment where students actively share and discuss their digital creations. The results have been dramatic. 120 students enrolled, 2/3 female, and only three students withdrew. By the end of the semester, the combined withdrawal, failure and D-grade rate had reached 11.5% — compared to 42.9% in the traditional introductory computer science course. 60% of the students who took media computation reported that they would be interested in taking an advanced version of the course; only 6% reported that they would otherwise be interested in taking more computer science. Results of the trial indicate that media computation motivates and engages an audience that is poorly served by traditional computer science courses.