The twofold aim of the ANR SKETCH project is to 1) collaboratively design an intelligent tutoring system (ITS) that can analyze learners’ actions in real time during the freehand production of a complex scientific drawing on a tablet, and 2) assess and optimize the effects of this system and the feedback it provides on learning. This project will be carried out jointly by two research teams in Rennes (France): the Psychology, Cognition, Behavior& Communication Laboratory (LP3C), and the IntuiDoc team at the Computer Science Laboratory (IRISA). It will focus on drawing activities intended to enhance learning about anatomy. Two of the partners in the project are paramedical colleges: IFPEK and IFPS. This will allow instructors and students to be involved in the project. The ITS will support the drawing activity on the tablet in a variety of ways (visual or verbal support, feedback, demonstrations, contextual aids). In the rst phase of the project, the instructor will decide which support to implement (static tutoring). In the second phase, however, the use of artificial intelligence techniques (pattern recognition, knowledge modeling) will allow for the provision of real time personalized feedback adapted to the difficulties encountered by individual learners (dynamic tutoring).
During an initial needs analysis phase, several actions (interviews and user tests) were carried out by LP3C with instructors and students from the project’s partner institutes. These instructors were then involved in the construction of the teaching materials on the tablet, as well as in the organization of the studies integrated into their courses. Questionnaires identifying students’ drawing-based learning strategies were also distributed in various medical and paramedical elds. A study to evaluate the user experience with the tool was carried out with psychology students. The results showed a high level of acceptability and situational interest. As part of a thesis in psychology, five studies were conducted. Students’ learning performance, self-assessment and perceptions were evaluated. The results show limited effects of drawing. Drawing seems to specifically promote the acquisition of visuospatial knowledge, as measured by subsequent drawing or diagram completion tasks, without providing overall benefits for comprehension. In contrast, the findings suggest a potentially negative effect on the processing of textual information. Furthermore, drawing influences the accuracy of self-assessment and the perceived usefulness of the strategy. Finally, the effects of drawing appear to be moderated by learners’ spatial abilities and also depend on the type of activity and the instructional content.
We implemented pattern recognition techniques based on two-dimensional grammatical composition rules to model document structure and enrich domain knowledge. To model the system’s tutorial knowledge, we rely on domain constraint-based modeling, i.e. a problem is represented by a set of constraints that must be satisfied for the exercise to be solved. These constraints are represented by a knowledge graph. We have implemented a dynamic ITS version («IntuiSketch«) in student mode, which analyzes students’ handwritten tracings relative to an exercise given as input (drawing of the anatomy of the lumbar spine). The system is able to analyze the student’s tracings and generate feedback on the shape, location and spatial constraints linking the various anatomical entities.
Teams:
- Eric Anquetil, Full Professor, Insa Rennes, IRISA Laboratory (ShaDoc team)
- Nathalie Girard, Senior Lecturer, University of Rennes, IRISA Laboratory (ShaDoc team)
- Islam Barchouch, PhD student, Insa Rennes, IRISA Laboratory (ShaDoc team)
- Omar Krichen, postdoctoral researcher, Insa Rennes, IRISA Laboratory (ShaDoc team)
- Bruno Hortollary, research engineer, Insa Rennes, IRISA Laboratory (ShaDoc team)
- In collaboration with the LP3C laboratory: Eric Jamet, Full Professor, Univ. Rennes 2, LP3C Laboratory.