Using sequence mining to reveal the efficiency in scientific reasoning during STEM learning with a game-based learning environment

• Published in: Learning and instruction 2018-04, Vol.54, p.93-103
• Main Authors: Taub, Michelle, Azevedo, Roger, Bradbury, Amanda E., Millar, Garrett C., Lester, James
• Format: Article
• Language: English
• Subjects: Game-based learning; Log files; Metacognition; Process data; Scientific reasoning; Self-regulated learning; Sequence mining
• ISSN: 0959-4752; 1873-3263
• DOI: 10.1016/j.learninstruc.2017.08.005

The goal of this study was to assess how metacognitive monitoring and scientific reasoning impacted the efficiency of game completion during learning with Crystal Island, a game-based learning environment that fosters self-regulated learning and scientific reasoning by having participants solve the mystery of what illness impacted inhabitants of the island. We conducted sequential pattern mining and differential sequence mining on 64 undergraduate participants’ hypothesis testing behavior. Patterns were coded based on the relevancy of what items were being tested for, and the items themselves. Results revealed that participants who were more efficient at solving the mystery tested significantly fewer partially-relevant and irrelevant items than less efficient participants. Additionally, more efficient participants had fewer sequences of testing items overall, and significantly lower instance support values of the PartiallyRelevant--Relevant to Relevant--Relevant and PartiallyRelevant--PartiallyRelevant to Relevant--Partially Relevant sequences compared to less efficient participants. These findings have implications for designing adaptive GBLEs that scaffold participants based on in-game behaviors. •Use of sequence mining to examine self-regulated learning with a game-based learning environment.•Examined hypothesis testing behavior based on the efficiency at solving the game.•More efficient students tested fewer partially-relevant and irrelevant food items.•More efficient students had fewer sequences with fewer partially-relevant items.