During the course of Project H2020 MaTHiSiS Managing Affective-learning THrough Intelligent atoms and Smart InteractionS, research was oriented towards understanding how to custom-tailor learning for students with special needs based on an analysis of their emotional states.
Annaleda Mazzucato, who directed testing in Italy, Spain and the United Kingdom for the Fondazione Mondo Digitale, explains the research methodology used to conduct research in collaboration with 15 public and private schools, 3 centres specialising in autistic spectrum disorders and 2 organisations.
Each of the involved students completed 12 sessions with the MaTHiSiS Platform. Based on the profile of the user, including competences, knowledge, scholastic curriculum, learning objectives and abilities, the platform provides teachers with a series of didactic activities that can be completed on a computer or tablet, or interacting with a robot (Nao or TurtleBot). During the testing phase, the platform collected a wide range of information on interaction with the learner: answer times, device movements and quality of answers, as well as the posture and look of the learner to determine his/her emotional state in real time.
The testing was conducted with an ABAB Model to confront two identical conditions that only vary in a given parameter, in this case the role of the interpretation of the emotional state during the presentation of didactic material. In condition “A” (experimental), the didactic session was managed via the interpretation of the learner’s emotional states, while in condition “B” (control), the emotional interpretation was deactivated and the didactic materials were presented progressively.
From the point of view of the final user, who experiences both conditions, the experience is similar (same device, same setting, same equipment and same learning materials). This allowed the researchers to control some disturbance variables.
Data analysis included the collection of teacher answers and an individual and focus group questionnaire and cross-examining it with the quantitative data provided directly by the platform. This allowed the researchers to verify the relation between emotional states identified by the system and the didactic session, the student results and performance in general.
