The system logs the sequence of these cues, and then compares sequences from different speakers to see which combinations routinely lead to a listener nodding, and which do not.

ECONOMIST: How to teach computers to make appropriate gestures