At the moment, a driver is responsible for the consequences of an accident. If a child/animal runs into the road, the driver can 'choose' to hit the child/animal or swerve into on-coming traffic.
I don't see any problem with cars being programmed to hit anything they can't avoid by braking. I certainly don't think hitting an oncoming car is an option we need to preserve.
If there is an option to use the other side of the road without hitting anything, then I guess driverless could use that more reliably than a human would.
Additionally, in a mixed scenario (some driven, some driverless), driven cars can/aill become more reckless, cut in, ignore priority as they know the driverless cars will avoid them.
I think this is a valid concern. Perhaps it will be mitigated if the driverless cars immediately email video footage of such driving to the police and the other drivers insurer, every time it happens.
The scenario I've previously imagined is idiots nonchalantly striding into the road in full expectation that cars will emergency brake for them. I have experienced this once as a human driver, they were looking right at me so knew what they were doing. Once I'd come to the stop they walked very slowly, laughing to themselves. (I want to be clear here: these people weren't just wanting to cross the road. They were deliberately playing a game with drivers.) I imagine the idiots would have far more confidence in computer drivers, and it would happen a lot more often. I guess even pedestrians can/should be prosecuted by the police if they do this. (It it's not currently an offense, maybe it will need to be.)