If "everybody" still has to go to and from work between the hours of 8-10 AM and 3-6 PM, pretty much everyone's going to want their own car.
Extend this thought to its logical conclusion, and you realize that (self-driving) commuter busses and light rail can be programmed into the network as plain-old cars that happen to have high occupant capacities. You set that you're willing to take a bus (and one is en route), it'll stop for you. You set that you want a less noisy/drunk/homeless-filled experience, and you get a car instead.
You could also set a price threshold, where everyone in the car must already be willing to pay as much as you are to be picked up there and then. Low threshold—a taxi-like sedan comes. High threshold—a limo.
For the privacy factor, I think people will just get used to it as they have with current public transport, or tiers will be in operation (as you suggested) or someone will inevitably design a pod-based vehicle with seats facing port and starboard and a clever door system.
I do wonder if people will be uncomfortable with other passengers knowing exactly where they live and work? Perhaps designating pickup from the nearest corner would help with that, and pricing that encourages that (to save extra turns down side streets) will help too.
I wrote a bit more about all of these likely developments here for anyone else fascinated by the rise of automated vehicles: http://www.isaacforman.com.au/n/self-driving-cars/
Your phone could also learn your departure time, and automatically determine when there are cars available / pricing is optimal to tell you to leave.
I remember an enlightening graph of power grid demand over a year sorted by demand. (Which I can't find for the life of me.) The vast majority of the time is spent very near the average, but a very small amount of time needs almost twice the average, and another very small amount of time demands half the average. The problem is, the grid must be big enough to supply that 3-times-a-year double the average demand, which just sits there idly the rest of the year.
I imagine demand for transportation is similar though not as exaggerated.
Can everyone dynamically shift business routines due to "demand"?
There's no particular reason emergency service shift changes need to be synced to the common business day -- indeed, my intuition would be that accident timing patterns would make this a bad idea independent of traffic impacts -- but there's also no reason that employers with employees for which they have strong fixed-time needs couldn't subsidize (or, in the case of public necessity where the employer is the government, exempt) their employees traffic fees.
> Can everyone dynamically shift business routines due to "demand"?
Yes, they can -- but, of course, there is a cost (which varies from situation to situation) in doing so. Of course, there's also a cost -- particularly an externalized cost born by others -- of traffic. If you effectively internalize the external cost (i.e., make it born by the people making the decision to engage in the action), then behavior should reflect the true costs and benefits resulting from the action.
The costs associated with throttling/halting these sectors far outweighs the cost of overbuilt infrastructure. Electricity is already sold by the megawatt in "day ahead" commodity form so businesses have already priced this information into the market.
[1] http://en.wikipedia.org/wiki/Electric_energy_consumption#Ele... [2] http://www.eia.gov/todayinenergy/detail.cfm?id=10211
And of course they can be different sizes (eg 13 people, or 23 people). Or a local small car can take you to a feeder location where you get on a big bus and the process reverses at the other end.
Sure you can still pay for your own reserved car, but it is likely to be even more expensive than pay as you ride.
The big problem with existing solutions like public transport and car pools is they are cumbersome at either end and require precise in advance organisation for timing and capacity, plus being inconvenient if you don't fit their deployment patterns. The major benefit of self driving cars will be algorithms can maximize utilisation and reduce individual costs. Owning your own car will look like a really bad decision by then.
The change doesn't have to be MASSIVE. For example shifting by 15 minutes may be enough to better match capacity. The most expensive would be point to point direct service for just you. But the driverless vehicle could pick up several people for a lower cost to them, but taking more time.
ie the worst case wouldn't be substantially different than today with its limits and hard routing/timing on public transport. It will also be easier for competition to come in compared to today which will help.
It is still possible to provide social services. For example in Britain my over 65 parents get free bus passes. Lower income people can be subsidised in similar ways.