The long-term goal of our research is to understand how to make adaptive educational games — in particular, how to give individual learners the freedom to learn in ways they prefer while still making sure the personalized learning experiences satisfy necessary pedagogical structure. Through this EAGER project, we gained scientific knowledge and understanding of 1) how individual learners behave in a widely used genre of educational games, 2) how to algorithmically model individual learning behaviors, and 3) how to scaffold individual learning by balancing learning autonomy and pedagogical process. 

We developed Solving the Incognitum, an educational game about fundamental earth science concepts. Drawing on the rich history of Philadelphia, our virtual learning environment was modeled after Charles Peale's early 19^th century Museum of Art and Science located in the Independence Hall, the largest U.S. natural history museum of its day. Our game followed the convention of education games and was designed specifically to support learners motivated by accomplishing goals as well as those motivated by exploration, two main player types identified in our prior work and existing literature.

Using Solving the Incognitum, we gathered data on how players interact with this virtual environment. By correlating players’ self-report survey data with how they actually play the game, our project advanced the existing understanding of how learners behave in educational games. Most notably, our study is among the first to provide empirical evidence that learners change their play styles within the same gameplay session and how this shift takes place in the context of our game.

To provide personalized support automatically requires the game to first identify the play style adopted by each individual learner. Existing techniques of player modeling assume static play styles. Thus, our project made important contribution to the scientific understanding of how to model the dynamic nature of play style, as observed in our study above. In particular, we developed a novel framework by segmenting each learner’s gameplay data into appropriate episodes and use sequential machine learning techniques to model them. Our results show that our framework outperforms the standard methods.

The final key intellectual contribution of this EAGER project is to explore how to scaffold individual learning by balancing learning autonomy and pedagogical process. We developed an extension of the Solving the Incognitum game where a trained human researcher alters the game in real time, based on each learner’s play style. We recruited 27 additional participants to play the game. Although its impact on learning is inconclusive, the adaptive version of the game led to higher learner satisfaction of the game. It also produced valuable information of how to design adaptive educational games.

Overall, this project directly led to four peer-reviewed publications at international conferences, two of which had the acceptance rates of around 30%. One of these publications was nominated for the Best Student Paper (overall acceptance rate of 28%). We also presented results from this project at two additional international conferences.

In addition to the above-mentioned intellectual outcome, this project fulfilled our intended goals for broader impact. Our collaboration with a Science teacher from George Washington Carver High School of Engineering and Science, part of the Philadelphia Public School District, allowed us to disseminate our educational game and research to a broader audience outside the traditional academic setting. 

We provided interdisciplinary research training and mentoring to multiple undergraduate students and graduate students from Digital Media, Computer Science, and Sociolo...

Please report errors in award information by writing to: awardsearch@nsf.gov.