Exploratory Learning Environments (ELE) are open-ended and flexible software, supporting interaction styles by students that include exogenous actions and trial-and-error. ELEs provide a rich educational environment for students and are becoming increasingly prevalent in schools and colleges, but challenge conventional plan recognition algorithms for inferring students' activities with the software. This paper presents a new algorithm for recognizing students' activities in ELEs that works on-line during the student's interaction with the software. Our approach, called CRADLE, reduces the amount of explanations that is maintained by the plan recognition in a way that is informed by how people execute plans. We provide an extensive empirical analysis of our approach using an ELE for chemistry education that is used in hundreds of colleges worldwide. Our empirical results show that CRADLE was able to output plans exponentially more quickly than the state-of-the-art without compromising correctness. This result was confirmed in a user study that included a domain expert who preferred the plans outputted by CRADLE to those outputted by the state-of-the-art approach for the majority of the logs presented.
|Number of pages||7|
|Journal||CEUR Workshop Proceedings|
|State||Published - 1 Jan 2014|
|Event||Workshops on Educational Data Mining, WSEDM 2014 - Co-located with 7th International Conference on Educational Data Mining, EDM 2014 - London, United Kingdom|
Duration: 4 Jul 2014 → 7 Jul 2014
ASJC Scopus subject areas
- Computer Science (all)