神経科学的知見を用いた教授ロボットの設計手法に関する研究

2015 Fiscal Year Annual Research Report

• Project/Area Number: 15F15007
• Research Institution: The University of Tokyo
• Principal Investigator: 開 一夫 東京大学, 総合文化研究科, 教授 (30323455)
• Co-Investigator(Kenkyū-buntansha): JOUEN ANNE-LISE 東京大学, 総合文化研究科, 外国人特別研究員
• Project Period (FY): 2015-07-29 – 2017-03-31
• Keywords: 認知科学 / 社会脳科学

Outline of Annual Research Achievements

I have spent the first months of my post-doctoral project to define more precisely my area of research and to design experiments to answer it. I chose to focus on language acquisition because of the ability to quantify progress and the huge neuroscientific literature concerning the evaluation of language through neuroscientific tools, but also because of my own knowledge in language comprehension acquired during my thesis. I have also decided to work on foreign language acquisition because of the scientific interest to evaluate the acquisition of a new language and the real benefits of using robots as teachers of a foreign language. Despite the fact that language is broadly studied through neuroscientific technics, the neuroscientific evaluation of robots as foreign language teacher is an innovating and challenging topic of interest.
For that purpose, I am developing several experimental protocols using the robot Nao as a French teacher, for experiments with both adults and children. Particularly, I have created two experimental protocols. 1) The “Nao story-teller” experiment (the robot is telling stories in French and participants learn naturally to discriminate French language by listening to him). 2) The second experiment concerns the acquisition of vocabulary (French words) during an interaction of joint attention with the robot Nao or its avatar (pointing objects on computer screens and naming the objects). Developing the robot behaviors and the set-up of the experiments have been time-consuming but will be useful for the other experiments I plan to do.

Current Status of Research Progress

2: Research has progressed on the whole more than it was originally planned.

Reason

Currently, I am finishing to develop the two experiments previously mentioned.
The first experiment, the “Nao storyteller”, is designed for children. It is inspired by a developmental experiment created by Kuhl et al. (2003, 2011) where they demonstrated that after a short period of exposure to a foreign language during sessions of story-telling by a person, 9-months old babies were able to discriminate phonetically this foreign language. Children of control-groups exposed to audio stimuli or television videos of story-telling did not show the same ability of discrimination, demonstrating that social interactions are important for language learning. With the aim of reproducing this experiment with a robot, I created behavior of story-telling for the robot Nao. At the same time that the robots tells stories in French, pictures of the stories are displayed on a computer screen. The robots demonstrates “simulated” social behaviors that creates basic social interactions (gazing at the public, pointing objects on the computer screen, using the name of participants to draw children’s attention, moving and mimicking the actions of the stories). The purpose of this experiment is to know if these basic social skills of the robot are sufficient to teach language. The protocol is ready to be tested (several stories and robot behaviors have been designed).
The second experiment I am designing, learning French trough joint attention with a robot or avatar, will be tested initially with adults using a NIRS system to obtain information of brain activity during learning.

Strategy for Future Research Activity

A behavioral pilot version of the Nao story-teller experiment will be conducted soon (May) in a pre-school with 3-5 years old children (10-15 children). During 1 week, the robot will be telling stories in French to a group of children (15 min/day). Each day, children will be evaluate individually to test their ability to discriminate words in French through a preferential-looking paradigm (the words used for testing will be words repeated frequently during the story of the day). If children can discriminate french words, it will be a first clue that robots have basic social skills sufficient to teach language.
Depending of the results of this first experiment, a more complex protocol will be organized in the lab using neuroimaging recording (EEG) and behavioral testing. To reproduce completely Kuhl’s experiment, the children will be younger (9 months) and the exposure to French more intensive (10 sessions of 30 minutes telling stories). To evaluate the effect of robot’s social behaviors, two groups will be used: in the experimental group, children will be exposed to an interactive robot (gazing, pointing, naming) and in the control group, the robot will not be interactive (just telling the story, no move). TDue to the complexity of organization of this protocol, it will be conducted later (from august-December). But during its organization, the NIRS “Learn French through joint attention” experiment will be conducted with adults in order to obtain brain data faster and easier to analyze. The NIRS experiment will be conducted and analyzed during summer.