Research 2
Developmental robotics
One use of robotics is to test theories of cognitive development – if you can implement your theoretical model, it is a lot more credible. Moreover, you can create “existence proofs” to rule out claims that something “can’t be done”. BEV was the most recent example of this, testing the hypothesis of John Watson (1975) that multimodal contingencies are a primary engine for perceptual learning, and showing that faces can be quickly learned without any prior face bias.
Baoo?
Yeee!
Hello little girl!
Yes you are so cute!
Bggh.
In 2001, I helped create the MESA project (Modeling the Emergence of Shared Attention) with Gedeon Deák, Jochen Triesch, and Javier Movellan. The project combines observational techniques of infant-caregiver interaction, computational learning models, and robotic models to provide a learning account of typical and atypical development of shared attention. Here is our founding paper. Below you can see motion-capture data of mother-infant interactions we collected at ATR (Kyoto) in Hiroshi Ishiguro’s lab in the Intelligent Robotics and Communications laboratory.