This video gives me a bunch of ideas based on Isolation and perception training. When the mask is worn and the video feed is turned on, then it could be delayed for a second or two. The result would be a slight delay in response time if someone was talking to the rubber doll/robot illustrating processing time within the internal circuitry.
If sound was isolated then it could be turned on depending on what the handler wants. You could make your doll/robot earn her upgrades abilities as she progresses. Combine that with a controllable delay of visual feed and you’ll have some powerful tools at your disposal for perception and mind control.