The number of Massive Open Online Courses (MOOCs) and the number of participants therein have exploded within the last couple of years. The literature on MOOCs (according to Clark, Daniel and Kolowich) mentions several characteristics of MOOCs: flexible, free, and providing room for niche subjects of learning. Criticism points to questionable quality and lack of student activation. Few empirical investigations focus on systematic analysis of student interaction in MOOCs. In this paper, the argument is that interaction between participants' of MOOCs on subject matter issues are a sign of collaborative knowledge construction processes and, as such, signs of learning (in line with the CSCL thinking of Laurillard, Stahl, Sorensen, and Salmon). This paper presents a study whose objective is to identify methods for analyzing and evaluating the nature of collaborative knowledge construction in the discussion forums of an MOOC, and relate this to the overall learning objective and educational design for learning of the specific MOOC that is analyzed. The analytical frame builds on two theoretical mindsets on interaction and collaboration. Language Games (LG) (as defined by Wittgenstein and applied in Sorensen, Takle, and Moser in CSCL) are applied as a method to investigate the structural elements. That is, the participants are viewed as the 'players' that engage in LG, where, for example, both openings and closing are viewed as vital in a successful interaction. Here, the volume of structural elements in selected online dialogues and the exemplary LG that is implemented becomes interesting. Salmon's Five-Stage model is the second tool, where the same selected dialogues are analyzed to identify where in the Five-Stage model (5S) the participants are interacting from, and if any progression occurs in the duration of the course. Moreover, the overall educational design is evaluated on the basis of Salmon's ideas. In the winter of 2013, the Coursera course 'A Brief History of Humankind' was analyzed. The findings show that the collaborative knowledge construction process occurs from the first week onward, and that the quality in a Salmon 5S view improved from week 1 to 10, but there are also contradicting findings regarding mutual respect and conversational tone. The analysis using LG finds that structural elements are important for the quality of the game, as responsibility and closing of games is seldom present, but that the games are initiated and many posts are made within the rules of the games. As such, this framework reveals paradoxical trends. There are authentic collaborative knowledge construction processes, although the number of active participants may be very few compared to the number of people enrolled. There are signs of collaboration, where the disruptive nature of MOOCs as a learning innovation argues against the postulate that MOOCs are solely transmissive in nature. Perhaps, MOOCs instead transcend the traditional Computer Supported Collaborative Learning CSCL (CSCL) perspectives for those who want to participate actively. This is discussed further in the paper.
