Newton's Reactive Televisions work on exactly the same principle as its executive toy namesake. (see link) The big difference is that it is quite a bit bigger, and each of the balls contains two back-to-back, outwardly facing televisions.
These televisions are capable of displaying a variety of images,
which are linked to movement sensors in such a way as their content databases react to the degree of motion, timing, impact and amplitude of the hanging balls.
An example of this would be the selection of "water" as the image range. Each television would begin by depicting a scene of static water, such as might be found in a half filled fish bowl. Once the balls were displaced, and impacting each other, the image of the water would begin to slosh around, realistically, with the two outermost televisions displaying the greater disturbance. The ones in the middle would merely show a vibration indicative of the transmission of the momentum.
Human figures could also be selected. These would tumble about alarmingly along with various items of furniture in the outermost screens, desperately trying to brace themselves against the pendulous impacts, whilst in total contrast, the ones in the middle screens would stand around casually grooming themselves, unless two televisions were pulled out together to start the action.