Here’s a question that is a question in theater and film as well: how much “character” (characteristic behavior) is enough?

An action film is almost all action (except for an occasional “character” actor for flavor). A Chekhov scene is almost all character. Most dramatic works fall somewhere in between.

Are there diminishing returns from adding character to robotic performances?

It’s possible that adding too much (too many levels of? too many variations on?) behavior can make the performers look merely random.

Which raises the corollary questions: what is the objective of adding behavior? What is the objective of robotic performance? Are we trying to achieve a “human-like” performance? Is that possible?

Here’s one partial answer: we judge the intelligence of a system by the complexity AND specificity (appropriateness to input) of its responses. Intelligent systems are usually more interesting than dumb ones. So complexity should not overwhelm specificity … both qualities need to be balanced in order to achieve an interesting performance.

Here’s another thought: someone who responds inappropriately on a consistent basis is often judged to be schizophrenic. Schizophrenia can be interesting for a while, but ultimately it’s boring … we lose the cause-and-effect chain of narrative. So a schizophrenic robot performer, too full of tics and quirks to respond to the action of a scene, would be a curiosity unable to maintain the continuing involvement of the audience.