Kismet was a study by Dr. Rodney Brooks and his students of human interaction with robots at MIT. It mimics (or attempts to mimic) human facial expression, and uses a set of unstructured voice like sounds to sound like it is communicating. Because it has a mouth, nose, ears and eyes - people feel like it is a real presence in the room with them.
Using an vision system previously seen in Cog, kismet has two sets of camera “eyes”, one which is less focused and has a wider angle, and another foveal camera which can give a more directed, focused and detailed image of something. Because of the way the two camera move (again derived from Cog) - a human is able to make eye contact with the robot, or follow its line fo sight - and understand what it is “seeing”. Kismet will become interested in bright, moving objects, but will primarily look for what appears to be skin - because of the techniques used - in a way that transcends skin pigment. If you interact with it, it will interact back. If you look towards something interesting - Kismet will follow your line of sight and become interested too. If ignored - it will sulk, pine and readily attend to the next skin coloured object.
Kismet is a fascinating robot indeed - and it has left quite a few people feeling confused because of how closely its expression mimics ours. The only time it seems lifeless is when either switched off or going through calibration routines during startup - otherwise, it is very animated and lifelike in its reactions.