Abstract

One important challenge in mathematics education is teaching modeling skills. We analyze the logs from a game-based learning system used in a massively multiplayer online tournament. Students had to detect an input-output pattern across 20 rounds. For each round, they received an input and had 2 minutes to predict the output by selecting a binary option (2 points if correct, -1 otherwise), or writing a model (4 points if model prediction was correct, -4 otherwise), or refraining (1 point). Thousands of 3rd to 10th grade students from hundreds of schools simultaneously played together on the web. We identified different types of players using cluster analysis. From 5th grade onwards, we found a cluster of students that wrote models with correct predictions. Half of the 7th to 10th grade students that detected patterns were able to express them with models. The analysis also shows diffusion within the teams of modeling strategies for simple patterns.