Brian Scassellati, B.S., Brain and Cognitive Science, MIT, 1995 B.S., M.Eng., Ph.D., Computer Science, MIT, 1995, 1995, 2001 Joined Yale Faculty 2001.

Brian Scassellati's picture
Professor Computer Science and Mechanical Engineering and Materials Science
AKW 501, 51 Prospect St, New Haven, CT 06511-8937

Brian Scassellati builds human-like robots to investigate models of human development. Many scientists have used computer simulations to model complex interactions of neural function, sensory processing, and motor control. Scassellati believes that robotics offers a new kind of tool in investigating human behavior. By building robotic systems based on models of human intelligence, the full complexity of real world environments can be used in evaluating these models. This research integrates work from artificial intelligence, machine vision, classical robotics, psychology, cognitive science, and human-machine interfaces.

One of the major focuses of Scassellati’s work has been on how children develop social skills and an understanding of other people. Children gradually acquire many skills that allow them to learn from their interactions with adults, such as responding to pointing gestures, recognizing what someone else is looking at, and representing that other people have beliefs, goals, and desires that differ from those of the child. These abilities have often been called a “theory of mind” and are believed to be critical for language acquisition, for self-recognition, and in the development of imaginative play. Computational models of these skills are being developed, implemented, and tested on robotic platforms currently under construction. His previous robots have performed tasks such as imitating human arm gestures, distinguishing animate from inanimate stimuli based on self-propelled motion criteria, and learning to reach for visual targets.

By building machines with social skills, we can not only begin to model human development but also construct machines with more natural interfaces. Machines that can engage in the same social dynamics that people use with each other will be easier to use, will be usable in a wider range of situations, and will not require the user to learn arbitrary commands.

One of Scassellati’s current projects is to develop robotic systems that aid in the diagnosis of autism by autonomously classifying social interactions. He is also currently building a new robotic system that will help teach social skills to autistic children. Additional projects focus on how children build a representation of objects as entities that exist through space and time, how animate stimuli are distinguished from inanimate objects, how sensation and motor control coordination are achieved in simple arm movements, and how children and animals distinguish themselves from others.

Representative Publications:

  • “Distinguishing Animate from Inanimate Visual Stimuli,” International Joint Conference on Artificial Intelligence (IJCAI), 2001.
  • “Challenges in Building Robots that Imitate People,” with C. Breazeal, K. Dautenhahn and C. Nehaniv, eds., Imitation in Animals and Artifacts, MIT Press, 2001.
  • “Theory of Mind for a Humanoid Robot,” Autonomous Robots, 12, 13-24, 2002.
  • “A Context-Dependent Attention System for a Social Robot,” with C. Breazeal, International Joint Conference on Artificial Intelligence (IJCAI), 1999.