Investigating Models Of Social Development Using A Humanoid Robot PDF Print E-mail
Written by Rizki Noor Hidayat Wijayaź   

Brian Scassellati: MIT Artificial Intelligence Lab, Cambridge, MA, 02139, USA:


The evaluation of models of social and behavioral development is dificult in natural settings; ethical concerns, diculties in implementing experimental procedures, and diculties in isolating hypothesized variables make experimental evidence dicult or impossible to obtain. We propose the use of human-like robots as a testbed for the evaluation of models of human social development. Robotic implementation of human social models allows for unique opportunities to evaluate those models. In this paper, we review some of the implications of this proposal by examining a case study of an on-going project to implement an existing model of one aspect human social development, the development of joint attention behaviors.

Introduction Research on humanoid robotics has been motivated by a variety of di

erent goals. Some research groups have focused on the construction of machines with humanlike form and motion to meet anticipated commercial needs as a flexible factory worker, a domestic assistant, or to operate in areas that are dangerous to humans (Hirai, Hirose, Haikawa & Takenaka 1998, Kawamura, Wilkes, Pack, Bishay & Barile 1996).

Other research has focused on the construction of humanoid robots to examine issues of human-robot interaction and cooperation (Takanishi, Hirano & Sato 1998, Morita, Shibuya & Sugano 1998), or to examine issues of sensory-motor integration and architectural techniques from articial intelligence (Kanehiro, Mizuuchi, Koyasako, Kakiuchi, Inaba & Inoue 1998). The majority of these research efforts have focused on the challenging engineering issues of building intelligent and adaptive systems. We have proposed that humanoid robotics research can also investigate scientic questions about the nature of human intelligence (Brooks, Breazeal (Ferrell), Irie, Kemp, Marjanovic, Scassellati &Williamson 1998). We believe that humanoid robots can serve as a unique tool to investigators in the cognitive sciences.

Robotic implementations of cognitive, behavioral, and developmental models provide a test-bed for evaluating the predictive power and validity of those models. An implemented robotic model allows for more accurate testing and validation of these models through controlled, repeatable experiments. Slight experimental variations can be used to isolate and evaluate single factors (whether environmental or internal) independent of many of the confounds that afect normal behavioral observations. Experiments can also be repeated with nearly identical conditions to allow for easy validation. Further, internal model structures can be manipulated to observe the quantitative and qualitative efects on behavior.

A robotic model can also be subjected to controversial testing that is potentially hazardous, costly, or unethical to conduct on humans; the oundary conditions" of the models can be explored by testing alternative learning and environmental conditions. Finally, a robotic model can be used to suggest and evaluate potential intervention strategies before applying them to human subjects.

In this paper, we discuss the potential biological and engineering questions that can be examined by implementing models of human social development on a humanoid robot. Our group has implemented biological models at many di

erent abstraction levels, including interaction models of infant-caretaker interactions (Breazeal & Scassellati 1998, Breazeal (Ferrell) 1998), behavioral models of the development of infant reaching (Marjanovic, Scassellati & Williamson 1996), and neural models of spinal motor neurons (Williamson 1996, Williamson 1998). In this paper, we present an on-going implementation of one behavioral model of social development which focuses on the recognition and production of joint attention behaviors (Scassellati 1996, Scassellati 1998c).

[download: 5,35 MB]