Wednesday, September 16, 2015

Empathic robotic tutors

Our robot is an empathic tutor that was developed as a part of the European Union funded project called EMOTE. This is a project with several partners in Europe comprising of teams from UK, Portugal, Sweden and Germany. Our empathic tutor is an autonomous agent who will present tasks to children's aged 10-15 and as they do the task help them out. It is empathic because it can sense the emotional state of the student(s) and adapt its responses accordingly. We have developed two scenarios: a treasure hunt and an urban planning game.

In one scenario, the task for the student is to carry out a treasure hunt. The map is laid out on a huge touch table. They are given tasks such as "find the museum about 300 meters North of your current location". Students answer these tasks by touching the appropriate feature on the map. This answer is diagnosed as correct, incorrect or partially correct. The tutor responds with an appropriate feedback. If they get it wrong, it tries to help them by suggesting appropriate tools to use, prompting them, giving them hints, etc. These pedagogical tactics are adapted based on the skills of the learner. As the learners progresses and his/her skill levels increase, less help is provided.

In the other scenario, the tutor interacts with two students. They play a game called Enercities. It is an urban planning game where the team builds a city. In order to build a city, they will have to balance between constructing features pertaining to the economy, environment and citizen's wellbeing. The role of the tutor is to play the game with the students as a team and reinforce the cause and effect relationships between concepts in sustainable development.

In both these scenarios, I was involved in design and implementation of the central decision making module called the Interaction Manager. IM is the module that decides what to do based on the state of the interaction. We designed and built an IM engine that can be loaded with IM scripts. Scripts tell the engine how the conversation should proceed. It is written in XML. Based on our consultations with educationalists and psychologists, we implemented scripts for both the scenarios.

The system was tested in several schools across Europe and currently we are analysing the results. Overall the students seem to have enjoyed interacting with the robot. 

No comments:

Post a Comment