Research in computing education has been criticized as "Marco Polo," e.g., the researchers tried something and reported what happened. Our developing field needs more hypothesisdriven and theory-driven research. We will get there by making clear our goals and hypotheses, testing those goals and hypotheses explicitly, and critically reconsidering our results. My colleagues and I designed and evaluated a mediacentric introductory computing approach ("Media Computation") over the last ten years. We started from a "Marco Polo" style and an explicit set of hypotheses. We have worked to test those hypotheses and to understand the outcomes. Our iterative effort has led us to explore deeper theory around motivation and learning. This paper serves as an example of a ten year research program that resulted in more hypotheses, a more elaborated theory, and a better understanding of the potential impacts of a computer science curriculum change.
SETTING HYPOTHESESValentine critiqued computer science education research papers in 2004 [45] as having a "Marco Polo" style: "I went there and I saw this." He pointed out that such papers play an important role in communicating experience to fellow educators, but did not have the same value as experimental papers, which conducted some theory-driven analysis of a given treatment. In his meta-analysis, he characterized the majority of papers published in the SIGCSE conference were "Marco Polo" papers. I suggest in this paper that "Marco Polo" is a natural condition at the beginning of a research project. Explicit statement of hypotheses, testing against those hypotheses, and iterating to refine and integrate other results and theories creates a progression to producing theory.Since 2002, my students and I have been designing and evaluating an approach to teaching introductory computation with a media-focused context, named Media Computation (or MediaComp). MediaComp introduces computing through exploration of data abstraction related to digital media. Students manipulate pixels to create Photoshop-like image effects, samples to splice or reverse sounds, text to compose or search HTML pages, and frames in a video. At Georgia Tech, MediaComp is taught using Python, but at other institutions (such as UCSD [37]) they teach similar things in Java. Our first paper which proposed the approach was even less of a research contribution than a "Marco Polo" paper, since we merely considered the possibility of a mediacentric approach [25]. The first two papers describing the initial implementation were clearly "Marco Polo" style papers [26,17] describing what we were trying and some of the initial responses from the students.In addition to the reports of "we went there and saw this," those two published papers (the ITICSE 2003 paper [17] and our first design paper [24]) and an internal design document for the course[16] 1 together lay out a series of design goals. The evaluation efforts that followed were explicitly testing those design goals as hypotheses, where the hypotheses were that the implem...