However, in Philip K. Dick's 1952 short story A Present for Pat, robot taxi drivers display real feelings and emotions:

"Robots have no wives," the driver said. "They are nonsexual. Robots have no friends, either. They are incapable of emotional relationships."

"Can robots be fired?"

"Sometimes." The robot drew his cab up before Eric's modest six-room bungalow. "But consider. Robots are frequently melted down and new robots made from the remains. Recall Ibsen's Peer Gynt, the section concerning the Button Molder. The lines clearly anticipate in symbolic form the trauma of robots to come."

"Yeah." The door opened and Eric got out. "I guess we all have our problems."

"Robots have worse problems than anybody." The door shut and the cab zipped off, back down the hill.
(Read more about PKD's robot cab)

I'm not differentiating between an autonomous car and a car driven by an autonomous robot here, but the basic idea is the same.

In a recent study published in the Transportation Research Record, Noah Goodall points out that road vehicles are significantly different, in that (for example) trains only move in one dimension on a set track, and the only choices have to do with the speed of the vehicle. Cars on the road, on the other hand, have many more degrees of freedom, and there is an irreducible moral component to the choices that the autonomous driver must make.

These advanced automated vehicles will be able to make pre-crash decisions using sophisticated software and sensors that can accurately detect nearby vehicle trajectories and perform high speed avoidance maneuvers, thereby overcoming many of the limitations experienced by humans. If a crash is unavoidable, a computer can quickly calculate the best way to crash based on combination of safety, likelihood of outcome, and certainty in measurements, much faster and with greater precision than a human. The computer may decide that braking alone is not optimal, since at highway speeds it is often be more effective to combine braking with swerving, or even swerving and accelerating in an evasive maneuver.

One major disadvantage of automated vehicles during crashes is that unlike a human driver who can decide how to crash in real-time, an automated vehicle's decision of how to crash was defined by a programmer ahead of time. The automated vehicle can interpret the sensor data and make a decision, but the decision itself is a result of logic developed and coded months or years ago. This is not a problem in cases where a crash can be avoided—the vehicle selects the safest path and proceeds. However if injury cannot be avoided, the automated vehicle must decide how best to crash. This quickly becomes a moral decision.

Goodall then asks if it is possible to design an ethical robotic vehicle. Science fiction writers have long wondered about this possibility, and Goodall mentions Isaac Asimov's Three Laws of Robotics, rewriting them for this context:

• An automated vehicle may not injure a human being or, through inaction, allow a human being to come to harm.
• An automated vehicle must obey orders given it by human beings except where such orders would conflict with the First Law.
• An automated vehicle must protect its own existence as long as such protection does not conflict with the First or Second Law.

Many problems can be seen with this scenario, as Goodall points out. What if a vehicle decides that there is too much traffic, and Rule (1) requires that it give up and stay in the driveway? I could well imagine a Philip K. Dick-designed car that might just give up. Or a Douglas Adams-designed vehicle with a Genuine People Personality inclined toward depression:

"...I'll send the robot down to get them and bring them up here. Hey Marvin!”

In the corner, the robot's head swung up sharply, but then wobbled about imperceptibly. It pulled itself up to its feet as if it was about five pounds heavier that it actually was, and made what an outside observer would have thought was a heroic effort to cross the room. It stopped in front of Trillian and seemed to stare through her left shoulder.

“I think you ought to know I'm feeling very depressed,” it said. Its voice was low and hopeless...
(Read more about Marvin the Robot)

Goodall concludes his paper by setting forth a strategy for developing vehicles with the necessary moral behaviors, including self-expression, to explain its actions to human occupants and other drivers.

A three-phase strategy for developing and regulating moral behavior in automated vehicles was proposed, to be implemented as technology progresses. The first phase is a rationalistic moral system for automated vehicles that will take action to minimize the impact of a crash based on generally agreed upon principles, e.g. injuries are preferable to fatalities. The second phase introduces machine learning techniques to study human decisions across a range of real-world and simulated crash scenarios to develop similar values. The rules from the first approach remain in place as behavioral boundaries. The final phase requires an automated vehicle to express its decisions using natural language, so that its highly complex and potentially incomprehensible-to-humans logic may be understood and corrected.

Hopefully, autonomous vehicles will be able to achieve some measure of ethical clarity without having, as Dick puts it, "worse problems than anybody."