4 seconds seems to be an odd choice if you are interested in how people react to an imminent collision - If I'm not mistaken, I think the length of time the traffic lights are at amber in the UK is 3 seconds or so, presumably going up as the speed limit increases, and in that time you are expected to recognize the change, decide what to do and brake to a stop before the line (or, you know, say fuck it and and foot to the floor. But I digress). This is in a predictable environment, where drivers really ought to be expecting to have to stop, but there are still many people who make awful decisions every day.
When someone/something appears in your path there is absolutely no way that you have anything like 4 seconds to make a weighted judgement and then (i.e. 4 seconds later) start to execute your plan. From personal experience I would wager that most people would go into some extreme tunnel-vision like instant reaction where they simply try to miss the thing the first grabs their attention as being something to miss, (which could be the ball or could be the child) so they hit the brakes and/or swerve.
Only once the instinctive manouvre is started then will they start to focus on what comes next in the list of things not to hit given the new direction they are travelling (oncoming vehicle, tree, pensioner, more children, etc.).
I don't really see where there is any ethical judgement being made by puny humans, and any machine capable of accurately recognising and categorizing all of these things in real time is also likely capable of driving entirely within its own abilities to miss everything. Whether or not people actually like being driven by machines that are that cautious around blind corners, junctions, etc. is another matter.
Edit: Ninja'd by several people above.