Inside the Minds of Autonomous Car Programmers

Would the jaywalking pedestrian be charged with the death of the driver that swerved to avoid hitting them? Who would be at fault in this death?

Driver? Aren't these autonomous?And programmed to save the driver at all costs.

More $300.00/hour lawyers fees will be the outcome.

No I'm saying in this circumstance without the autonomous control that a driver swerved to avoid hitting a pedestrian crossing the road illegally, and was killed because they hit a tree or something...Who's to blame for the death? The illegally crossing pedestrian should be held accountable...If the driver instead hit the pedestrian killing him, the pedestrian should still be to blame, because he was in the wrong....as long as the driver is breaking no laws and is in his lane, those entering that path illegally should be at fault...Now if the pedestrian is in a crosswalk, that's a different story....or if the driver is in a parking lot, he should always yield to a pedestrian....or in a school zone or other specially marked area...

Lawyers have gotten rich arguing both sides of your rationale.

You are also assuming that the, at fault" pedestrian stays around to be charged.

"The illegally crossing pedestrian should be held accountable..."

I don't disagree with you, but let me muddy the water a bit.

Suppose that illegally crossing pedestrian is a child. Perhaps they were chasing a ball or responding to a parents call to dinner. Is the child responsible? Or should the parents of the child be responsible (not for calling them to dinner, but for not teaching them to not cross the street illegally)?

Fight mud with mud....

If instead of the child running into the street and getting hit and injured or causing the occupants injury as a result of an attempt to avoid hitting the child, suppose the child rolled a large round rock (about the same mass as the child) down a hill into the road resulting in occupants injury either from impact or attempts to avoid...?

Wouldn't the parents be responsible? Why would the parents become less responsible if the mass is flesh rather than stone?

In either case, a preset algorithm acting on limited information will from time to time make the wrong choice, it is inevitable.

I remember Asimov's robot psychologist, Dr. Susan Calvin, somewhere addressing this problem in the "I, Robot" collection of short stories that every positronic brain that confronted this impossible choice was useless after dealing with such a dilema, regardless of the actual outcomes.

Sometimes 'better' is the enemy of 'good', but 'perfect' is a far more reliable enemy. Huge benefits would be eschewed if the requirement for implementation becomes 'zero wrong decisions'.

In the US yesterday around 100 people lost their lives in violent traffic accidents....which is about the same as the day before, and the day before that, and...

Far and away, the largest culprit is one or more of the decision engines operating a vehicle involved, making one or more 'wrong decisions'. If the number of deaths could be reduced by an order of magnitude by completely replacing all the carbon based decision engines with silicon based decision engines, so long as incremental replacement did not result in an increase death rate, shouldn't implementation be considred based on a reduced death rate being an improvement even if the remaining deaths were all due to wrong decisions by the silicon based decision engines?

Well, I plan to die in my sleep like my grandpappy and not like his screaming passengers.

Aside from my grandpappy , I don't know that it's a proven fact that machine intelligence is better at accessing a traffic situation and is a better driver than a human. It is always assumed that driverless cars will save many lives.

That's a gimme due to all the impaired and distracted drivers....If the autonomous driving system is as good as the average experienced healthy sober driver, it will save lives....

..."In 2015, 391,000 injuries were caused in distracted driving related accidents. In that same year, distracted driving was cited as a major factor in 3,477 traffic deaths. 9 people in the U.S. are killed each day as a result of crashes involving a distracted driver, according to the Department of Motor Vehicles.Apr 5, 2018"...

1. "Distracted driving accounts for approximately 25% of all motor vehicle crash fatalities."

...."In 2016, 10,497 people died in alcohol-impaired driving crashes, accounting for 28% of all traffic-related deaths in the United States. Of the 1,233 traffic deaths among children ages 0 to 14 years in 2016, 214 (17%) involved an alcohol-impaired driver.Jun 16, 2017"...

That's 53% of traffic related deaths...

https://www.teensafe.com/distracted-driving/100-distracted-driving-facts-and-statistics-2018/

https://www.cdc.gov/motorvehiclesafety/impaired_driving/impaired-drv_factsheet.html

So 75% of fatalities are caused by non-distracted drivers, and 72% by sober drivers.

These dangerous drivers should be got off the roads!

Top 25 causes of accidents....

1. Distracted Driving

2. Speeding

3. Drunk Driving

4. Reckless Driving

5. Rain

6. Running Red Lights

7. Running Stop Signs

8. Teenage Drivers

9. Night Driving

10. Design Defects

11. Unsafe Lane Changes

12. Wrong-Way Driving

13. Improper Turns

14. Tailgating

15. Driving Under the Influence of Drugs

16. Ice

17. Snow

18. Road Rage

19. Potholes

20. Drowsy Driving

21. Tire Blowouts

22. Fog

23. Deadly Curves

24. Animal Crossings

25. Street Racing

Honorable Mention

• High Winds
• Loose Objects in Cars
• Police Car Chases

https://seriousaccidents.com/legal-advice/top-causes-of-car-accidents/

OK, but there's a lot of overlap in many of those categories, as Rixter said. And there are so many of them the wonder is that we can ever venture on to the road without having an accident!

They're not easy to define - you can be a bit or a lot distracted, you can be 3 mph over the limit, or 2x the limit.

Surprisingly, the list doesn't mention vehicle in dangerous condition.

Well poor maintenance I'm thinking mostly would be tire blowout, slick tires in the rain, brake failure might be noticed by running stop sign or red light or possibly tailgating , with a few dozen other rare things that happen, like exhaust leak in car causing drowsy driving and others....I think we should concentrate on the big ones....

"Distracted driving accounts for approximately 25% of all motor vehicle crash fatalities."

...."In 2016, 10,497 people died in alcohol-impaired driving crashes, accounting for 28% of all traffic-related deaths in the United States.

That's 53% of traffic related deaths...

53% minus those both drunk and distracted...Just sayin'

No, it's listed cause is either one or the other.... If you have 10 traffic deaths and 4 are listed as distracted and 6 as impaired, you still have 10 deaths...

<Ahem>, this linked article is heavily biased with personal opinion and hypothetical examples designed to elicit an emotional response instead of actual facts. The answer is not so simple, both from an engineering and ethical point of view.

For example, most people (in a split second decision) would probably react by braking and swerving out of the way of an oncoming vehicle rather than just hitting it, regardless of the circumstances.

Who's to say this programmed response is better or worse than not avoiding a head on crash potentially killing all in both vehicles than potentially someone on the side walk or other lane?

Any way you look at it the car makers open themselves up to lawsuits no matter what they do, regardless of how safe using their autonomous vehicles make the roads.

This is the classic ethical dilemma known as the trolley problem. Wikipedia even has a couple of paragraphs on how this problem affects autonomous vehicle programs.

IMHO The crux of the problem is this. How does any decision maker choose between limited options when each option has dire consequences for somebody? Quantifying which option will become less tragic from limited and certainly incomplete knowledge is an impossible problem. It becomes even more impossible when the decision-making process is rigidly precoded.

I'd be happy if they just fixed the airbags....We're not going to achieve perfection, that doesn't mean we shouldn't try....just keep pushing in that general direction and over time things get better.....

The assertion of causality in the OP seems to be at odds with standard interpretations of response differences seen in various Trolley problems iterations, that is, being less immediately/physically involved in the process tends to make it more likely people will act to alter an impending event to minimize total tragedy.