AV "drivers" more likely to hit peds
 

Research article (paywalled, but abstract here: https://academic.oup.com/jcr/article...dFrom=fulltext) discusses AV cars and moral dilemma of an accident where either the driver or peds would take the effect, and driver has to determine which course of action to take. The finding was that in an AV the driver would be more accepting of taking out the ped, with the rationalization that it was the AV software "decision" that resulted in the ped adverse result, absolving the driver of moral responsibility. They found the effect existed up to groups of 5 peds, or if a ped was a child. I would assume if correct, the same moral reasoning would extend to cyclists.

scott s.
.

Ethical coding is becoming a thing.

When presented with the Trolley Problem, autonomous vehicles are more likely to hit what they were programmed to hit. The assumption that the currently on-board human would make a "better decision" in the moment is just that, an assumption. Algorithms don't have delayed reactions, they don't get tired or drunk or have road rage. The most dangerous part of any motor vehicle is the human driver.

Any algorithm would be complex. Choose property damage over causing serious injury to an individual.

Minor bumps and bruises for the person inside vs risk of serious injury or death to the pedestrian.

The dilemma only becomes a problem when the alternative is serious injury or death to the occupant. Head-on. Plunge over a cliff, barrel into a solid wall, etc.

Ideally the vehicles will be programmed to not over-drive their own capabilities.

If they can't program an AV to not hit pedestrians, they should perhaps consider and improve on this. Just make it deploy automatically.


Cheers

Algorithms shouldn't be controlling machines that can kill human beings.
If someone you care about had to die in a car accident, would you prefer an algorithm to take the responsibility, rather than a human being?
That's the moral dilemma we should be concerned with before considering anything else.

That horse has left the proverbial barn.

In most cases, the algorithms can be programmed to make the least bad decision, but first we would more or less have to agree what the least bad decision is, from a societal, not individual, point of view.

Tell this to Boeing.

First of all, algorithms don't make decisions. Free will would be a pre-requisite. Besides, they would be required to attain the age of majority just like human beings in order to be accountable for their decisions.

I’m fairly certain that algorithms will be optimized to prioritize driver and occupate safely including minor injuries, and if pedestrians get in the way that’s just too bad. (Thoughts and prayers)

There is no will now to protect cyclists and pedestrians if it inconveniences drivers even a little bit. Why would AVs change that mindset?

Given the current historical example, that automakers will give any thoughts at all to pedestrian or cyclists safety unless mandated by law is kind of a joke.

They should be programmed to self-destruct. If a hit can't be avoided then the wheels should be ejected so the car drops down to the road and grinds to a halt.

We have the Elaine Herzberg incident that was settled for an undisclosed amount of money. Potentially in the hundreds of thousands, or a million. And, almost killed off the Uber AV trials.

They were running a "Taxi" service, so the AV company (Uber) would have a lot of liability to whatever they run into, dangers to other vehicles on the road, and dangers to their occupants and the occupant's property.

If these systems are sold privately, we'll likely see liability where the owner will be liable for maintenance or personal choices they make. But, the manufacturer will be liable for algorithm design and algorithm choices.

Kill a bunch of cyclists and pedestrians, and the governments will pull the plug on the entire project. Millions, or Billions invested down the drain. Payouts for a few broken eggs vs payouts for killing someone... the company will make the choices following the pocketbook.

The problem with the Herzberg death was that she was picked up 6 seconds before the crash, plenty of time to slow down. But, by the time she was correctly identified as a pedestrian/bicycle, with her walking speed across the road, it would have necessitated an emergency stop. And, Uber decided not to force emergency stops for fear of false alarms. But, doing so cost Herzberg her life. Their Volvo might have even stopped if the manufacturers systems hadn't been overridden (can't have two AI systems fighting internally).

But we know this isn’t likely to be true. The government has stood by idly while car makers have stopped making sedans and switched to SUVs. SUV collisions with pedestrians are far more deadly than collisions with cars. Over the last 10 years this has contributed to an increase of almost 60% higher pedestrian deaths compared to 2009 resulting in 1000s of additional pedestrian deaths every year.

The government will mandate safety features to keep drivers safer, but have mandated very few features to keep cyclists or pedestrians safe because it’s not a priority.

They are forgetting that drivers become mere pedestrians the moment they get out of their cars.