Interesting article. Breaking down the three types of routes helped clarify some of the issues in my mind. It seems to me that a hybrid driver/driverless system might be more practical than a pure driverless system during the transition to all driverless cars. And be more acceptable to a large part of the public. It also eliminates a big part of the moral hazard during the driverless portion as the car would be unoccupied. This eliminates the “does the car protect the occupants over external persons?” question.
Car sharing where the car drives itself to a dedicated pickup point might be the answer. The biggest issue with driverless taxis is the unfathomable number of situations that can arise trying to find a passenger or their destination. If the driverless model was limited to specific routes, and arrive in specially designated parking spots like Zip cars have, then the driverless model becomes greatly simplified. This would be very important in high density areas like apartment complexes where finding an individual tenant can be frustrating even for a human taxi driver. In well laid out suburban areas cars can come to your door and return to a central parking area or move on to the next user as needed.
I do believe that driverless cars will be a part of the future. But I also believe that there will be a fairly long transition period. And there will always be situations where a human driver is the only option. And there will probably always be humans that desire control over their cars.