Automatic Detection of Peer Interactions in Multi-player Learning Games

Abstract

In Multi-player learning games (MPLG), learners interact with each other through game activities. This paper aims to establish the basis of automatic detection of peer interactions that could emerge from MPLG scenarios. The information provided by this detection could help learning game designers to construct scenarios fostering the interactions they need/desire. We present an algorithm extracting interaction features from MPLG scenarios. Because of the high number of possible sequences of activities in a scenario, we work only on a subset of the sequences which must preserve the information on the interactions emerging from the scenario. We evaluated this algorithm by comparing the calculated subset to the traced game paths obtained from 114 students. The results of this evaluation show that (1) the subset cover well the learners’ traced paths and that (2) the calculated features from the subset have the same types and are distributed similarly to the ones extracted from the traced paths.