Moral Crumple Zones: Cautionary Tales in Human-Robot Interaction (2019)
papers.ssrn.com
papers.ssrn.com
What if courts were lenient where the root cause is outside the defendant's control?
I think that's pretty much the GP's point...? You're responsible for your own actions, not the actions of the employer that put you in an impossible situation.
Start from the premise that most people in an employment situation should be considered as "acting under duress" (because if they get fired, they are at risk of homelessness, destitution, and ultimately starvation). Then consider an employer who demands more of the employee than they can safely give—what choices does the employee have? Violate the law in order to keep getting the money they need to survive, or choose not to violate the law and put themselves at very high risk of firing (and thus the above consequences).
I don't agree much with that. BUT! I am not American, nor do I live there. So YMMV.
> Violate the law in order to keep getting the money they need to survive, or choose not to violate the law and put themselves at very high risk of firing (and thus the above consequences).
I think that's the biggest issue. I was talking more from a "how it should be" kind of view, or how I perceive it from my central European background. You're reality might be wildly different, and I lack the experience to understand that I think.
My overarching point is: if you start to consider that the system is bad, how do you stop everyone else from doing the exact same thing, breaking laws to live? Where do you draw the line?
We start sending CEOs and members of the board to prison (preferably for life) for acts of a company's employees, contractors, and agents in service of the company. Rinse and repeat until companies get the message.
> My overarching point is: if you start to consider that the system is bad, how do you stop everyone else from doing the exact same thing, breaking laws to live? Where do you draw the line?
I think that there is a meaningful distinction between "I choose to break the law to survive" and "I choose not to refuse an order to break the law, because to do so risks losing the ability to survive." That said, I think that there's an argument to be made that a society where the laws are used to punish the poor for trying to do the bare minimum to survive is a society that throws all law into contempt.
Get a better union? This is why unions are similar employee protections are so important.
If employees can be forced into a corner and forced to act illegally or immorally by their employer, then you create a situation where employers that abuse their employees the most can out compete competitors that treat their employees well. Ultimately you just end up with a country of abusive employers, and disempowered abused employees.
> Start from the premise that most people in an employment situation should be considered as "acting under duress"
If that’s your starting point, then it does suggest a rather catastrophic failure of government.
That's not an "if." You just described much of today's America.
The point is that the law must give judges the options to do so, i.e. by defining a range of different punishments instead of insisting of some fixed minimum punishment.
It wouldn't really matter for sleep-deprived truck drivers falling asleep at the wheel and crashing.
It issues with allowing people like truck drivers to pass liability onto the system they operate within, is that it allows that system to abuse them further. If all truck drivers refused to work when unsafe, and also refused to work if they weren’t paid enough to only work when safe, then the market would naturally move towards increased pay (after all the good still need moving).
By creating a space where truck drivers are tempted to break the rules, and drive tired, you artificially reduce the value of their time, and effectively force all a drivers to drive tired into to remain competitive. Ultimately resulting in a downward spiral with every driver driving tired all the time.
In smart societies, drivers hours are carefully logged, and audited by state organisations. Drivers are responsible for ensuring the accuracy of their logs (using state specified hardware and methodologies), and are punished for driving unsafely, or attempting to hide unsafe driving. With a rigorous audit system, drivers know it question of when not if they caught driving unsafely, completely eradicating any temptation to bend the rules, and making it easy to dangerous and illegal instructions from their employer . Pair this with company level audits to identify companies that seem to be encouraging rules breaking, and punishing them directly, and you can get a system that ensures everyone is held responsible for their action. Nobody can say “they’re just following orders”, and simultaneously empowers everyone to refuse unsafe instructions.
If the driver's boss put a gun to their head and forced them to drive, it would clearly be the boss who was liable. Now replace the gun with the threat of insolvency and their kids living on the street. That's enough to push most people to bend the rules a little and take a few risks. If the boss isn't also held financially (or even criminally) liable then they have no reason (other than their own ethics) not to apply this kind of pressure.
> allowing people like truck drivers to pass liability onto the system they operate within, [...] allows that system to abuse them further.
If it allows the system to abuse them further, then the liability isn't being properly passed along. The auditing etc. that you suggest will only be implemented by trucking companies if it's in their interests to do so. Making the company (or even the company directors, personally) liable for breaches is the only way to ensure that kind of top-down compliance.
When trucks are stopped at weight station and inspected, the drivers log is also taken and inspected. Trucks are stopped on a regular basis and employers know to factor in stoppages (maybe 10-30min per stop) because they happen all the time.
The regulator moves between inspection points around the country to ensure they’re inspecting a representative sample of trucks.
Employers are not required to do anything except provide the mandatory equipment and ensure their drivers stop for inspections.
Someone else programmed the AI, and if it causes an accident, then the company that programmed it might be held liable instead of (or in addition to) the driver. It's similar to installing defective brakes, or a component whose failure causes sudden unintended acceleration (SUI) -- sure SUI is often operator error, but not always.
I think something which is new though is the question of who should take responsibility for the AI's actions? If a crime is committed via a proxy, then both parties will be charged. So it doesn't make sense to entirely just waive away the AI's responsibility in the crime just as it doesn't make sense to acquit the person/proxy who carried out the crime on behalf of someone else.
I think the answer is that the company(ies) which trained and/or operated the company behind the AI should also be held partially responsible for the crime. Otherwise we could end up with a situation whereby some companies are intentionally creating malicious bots which are trained to intentionally misinterpret benign-sounding language as orders of a more sinister nature in order to shield 'the customer' from criminal liability.
I'm much more worried about the consequences of humans being under-punished rather than over-punished.
Neglecting the role of AI companies seems like a sure way to bring about human extinction. We cannot allow all liability to be shifted to non-human entities. We need a better, more fine-grained way to attribute and punish shared and partial liability.
We should think in a more fine-grained way. For example, if someone throws out a plastic Coca Cola bottle on the ground, who is responsible for littering? Obviously the person who threw away the bottle is mostly responsible but isn't Coca Cola company also slightly responsible for having made their bottles out of plastic (which doesn't break down easily; thus worsening the impact of the littering)? Surely they could just have sold metal cans and glass bottles.
I thought that the concept discussed in the OP was slightly different (and possibly even more insidious) : The proxy is a human who carries out orders of (or covers for) an AI that is known to be unsafe.
I.e. an autonomous car manufacturer may acknowledge that their AI sometimes makes errors which can lead to accidents - so they equip their cars with a steering wheel and mandate that a human driver must always be present who can takeover from the AI in dangerous situations.
The problem is just that the human driver may not realistically be able to do that, because reaction times are too short, the AI releases control too late, etc.
Yet, if the AI malfunctions and the driver doesn't manage to prevent the accident, the company can comfortably blame the driver and keep the AI out of the discussion.