Abstract:

Researchers generally agree there are numerous and noteworthy aspects of learning that occur within video games (Schrader et al., 2017; Serrano-Laguna et al., 2014). Video games are authentic, though non-traditional, learning environments and are typically characterized by their interconnectedness, unbounded nature, and nested/encapsulating contexts (Schrader et al., 2019b; Steinkeuller, 2005; 2006). Players acquire and demonstrate skill, execute strategies, exhibit teamwork, communicate, and more while engaged in games. Although these environments may lack a predetermined educational goal, they often provide valuable research contexts for the study of human behavior, psychology, and education. Video games continue to garner attention as a serious context for instruction, as well as context to study learning and develop theory. The games, their data, and the social ecosystems in which they exist (e.g., eSports), provide a valuable digital Petri dish for educational researchers.

A principal benefit of leveraging games in research is the ability to select a system among numerous types, classes, and platforms. In the case of this proposed paper, the League of Legends was selected for three key reasons (LoL, Riot Games, 2009). First, LoL is a complex system, but one that is bounded spatially and temporally. In this way, player movement in the game can be monitored and each game has a discernable beginning and end. Additionally, the initial conditions for each game are equivalent across sessions, meaning that player choices, skill, and actions determine the outcome of the game. Second, LoL generates incredibly robust and large-scale data that are accessible via an application programming interface (API) provided by the developer. Typically, these types and structures of data are often difficult to access, organize, or analyze. LoL and the associated API helps resolve this. Third, LoL is enjoyed by roughly 115 million people worldwide (Spezzy, 2021). This translates into a robust social ecosystem in which players exchange ideas, share experiences, and communicate about LoL.

Although LoL was selected for its data capabilities, dealing with the breadth and scope of these data is often challenging. Fortunately, Data Science and Educational Data Mining (EDM) techniques are increasingly valuable for researchers in educational psychology (Baker & Inventado, 2014; Baker & Siemens, 2014). These methods have been used to examine a wide variety of psychological constructs and behaviors, such as student procrastination in massive open online courses (Yao et al., 2020), collaborative problem solving (Hur et al., 2020), or games-based stealth assessments (Henderson et al., 2020). As a result, this work applies EDM procedures to the eSports game League of Legends. Thousands of cases were mined using these techniques and they generated more than one million data points. These data were analyzed agnostically to establish the viability of an EDM approach in non-traditional learning environments that generate large-scale data.

Keywords:

Video games, educational data mining, league of legends, data science.