Multimodal learning analytics for studying creative problem-solving with modular robotics

Creative problem solving is a complex process that is being studied through a diversity of tasks. CreaCube is an open ill-defined task whereby the player is required to engage in a creative problem-solving activity (Romero et al., 2017) that we try to analyze with computational models (Alexandre, 2020a). To represent the behavior of the subject through the activity evolution, we analyze the activity through a coding schema considering the observables for this tangible problem solving activity. Observables correspond to different states of the artifact of behaviors of the participant at a given time. In this study we introduce the learning analytic strategy corresponding to a temporal sequence of observables. Through this sequence we aim to infer the participants’ internal state based on a different sequence of observables.Through this study we aim to advance in the learning analytics strategy of a tangible problem-solving task with educational robotics. Our main goal is to be able to both collect data more easily, avoiding as much as possible the manual analysis of the video recording of the activity, and propose enriched observables. To this end, we refined the observables’ model as detailed in (Mercier et al., 2021) and added new observables and decomposed existing ones into more specific ones based on the learner and task model. We distinguished observables with automatable data collection and those which require manual identification. In the end we discuss the relevance of this new version of CreaCube by discussing to what extent it offers additions to actual data analysis and ongoing research on this subject.