I think a neural network could do it. You just train it on a bunch of videos with different shutter speeds, and then you ask it to convert a given video from one speed to another.
I'm sure it would quickly learn to add/remove motion blur on moving things as appropriate.
But in addition to determining motion blur, shutter speed also massively affects which areas of the images are above/below the brightness range the sensor is capable of picking up.
I'm sure it would quickly learn to add/remove motion blur on moving things as appropriate.