5 ms·

I feel like this rapidly moves into Trolley Problem territory. 10,000+ people die in a year due to drunk driving, what's the acceptable death toll for rolling o

by ng12 8y ago

I feel like this rapidly moves into Trolley Problem territory. 10,000+ people die in a year due to drunk driving, what's the acceptable death toll for rolling out driverless cars as fast as possible?

lambda 8y ago
Because of the uncertainty behind something new, you should have strong evidence that driverless cars are substantially safer than cars with drivers. Additionally, if driverless cars are being tested on public roads, you need laws and practices that are robust enough to prevent the kind of accident that killed someone. They had only one operator in the car, who was responsible for monitoring the car and logging information about its performance. They had specifically disabled safety systems which would have caused it to brake sooner. Even if its expected that driverless cars will reduce fatalities due to drunk or distracted driving, you need to have sufficient evidence of the improvement, safety systems in place for test systems, and public discussion of the tradeoffs and what situations driverless cars might perform worse in than human drivers, because lots of people have behavioral expectations based on human drivers; things like making eye contact with a driver before crossing the street. So yes, it is good to be working on and testing driverless cars, but you need to have a serious sense of humility and safety about it, and not push it, because even if you are incredibly safe you can expect some backlash, but if you miss the mark at all and oversell the safety, you'll get a lot more backlash, with justification, much sooner, which could set the whole program back significantly.
- the8472 8y ago
  > you should have strong evidence that driverless cars are substantially safer than cars with drivers. No. Considering the other benefits of self-driving cars, such as increased mobility for the elderly, you can at most require that they are as safe as cars with drivers, not necessarily safer. If the benefits were great enough (reduced pollution indirectly saving lives? saved life-years due to more efficient traffic?) one could even accept that they could be a little less safe as far as direct fatalities are concerned.
  - miranda_rights 8y ago
    > If the benefits were great enough (reduced pollution indirectly saving lives? saved life-years due to more efficient traffic?) one could even accept that they could be a little less safe as far as direct fatalities are concerned. I disagree. I'm finding it hard to come up with an example that would make me think it's ok to have more people hit and killed by cars. I'm more than ready for self-driving cars to be a reality, but not if more people will die. "Saved life-years" is not good enough.
    - the8472 8y ago
      What does it mean to avoid a fatality other than saving that person's remaining life-years? > I'm finding it hard to come up with an example that would make me think it's ok to have more people hit and killed by cars. Entirely speculative: Heart failure is the leading cause of death in the US. If 100% self-driving traffic could let EMTs arrive faster at the scene that could save more lives than a slight increase from traffic accidents.
  - bumby 8y ago
    I understand your point but I doubt it would work well in reality when it comes to public opinion/backlash. Humans are disproportionately swayed by proximate causes, so deaths directly attributed to self driving cars are going to be more heavily weighted in the court of public opinion.
  - masonic 8y ago
    increased mobility for the elderly decreased, in fact. A self-driving vehicle can't go anywhere a cab or rideshare can't, while the latters have drivers who can assist loading/unloading elderly and their mobility aids.
    - glenra 8y ago
      Somebody with poor eyesight and slow reaction speed is an unsafe driver but doesn't need any particular help loading/unloading from a car. A driver who can help unload is a fine thing, but having somebody drive just so they can help unload seems like a ridiculous waste of resources. A car is ready whenever you need it; a car-and-driver might not be. And yes, you could schedule a car-and-driver to be available at a particular time...but you could similarly schedule somebody to be available to load/unload at the destination.
gamblor956 8y ago
Self-driving cars have killed at least 3 people this calendar year despite there being several millions fewer self-driving vehicles on the road. Scale up those proportionate rates, they would be killing far more than 10,000 people a year. So at this point, self-driving cars still have quite a bit of work to go to prove that they are actually safer than a drunk driver, let alone a sober driver.
- loeg 8y ago
  If you think drunk drivers are bad, sober drivers kill more than twice as many people! The actual number to beat is something like 37,000 annually[0]. And I suspect your "millions" factor has an extra zero or two when you account for (a) miles driven (self-driving cars are probably more highly utilized) and (b) driving conditions (city is more interesting for training data than highway, whereas probably more person-driving is done on highways). [0]: https://www.nhtsa.gov/press-releases/us-dot-announces-2017-roadway-fatalities-down https://www.nhtsa.gov/press-releases/us-dot-announces-2017-r...
  - masonic 8y ago
    Your source doesn't differentiate between sober drivers and drivers not tested for any/all impairing chemicals. Unknown sobriety != sober.
jdavis703 8y ago
Wouldn’t mandatory breathalyzer ignition locks solve this problem? This is a proven technology that shouldn’t kill anyone as it’s being rolled out. Move fast and break things is great for websites and apps, but when human lives are at stake it requires strict ethical considerations.
- loeg 8y ago
  Interlock breathalyzers have a host of problems that make indiscriminate application problematic. Additionally, judicial remedies can only be applied after someone has already been caught drunk driving. So even if we applied them to all prosecuted drunk drivers, it is not a complete prophylactic.
- rblatz 8y ago
  What about all the existing cars on the road? What about manual cars that can be push started? Who installs and certifies these, they need calibrated frequently, they are extremely simple to bypass for anyone with even basic vehicle knowledge.
  - jhbadger 8y ago
    If the law starts now, in 5-10 years the vast majority of old cars will be junked. Sure, there will still be a few old cars, but they will be an insignificant minority. It's not unlike how other rule introductions worked. Nobody banned old cars without seatbelts, they just required that all new cars have them.

by ng12 8y ago

I feel like this rapidly moves into Trolley Problem territory. 10,000+ people die in a year due to drunk driving, what's the acceptable death toll for rolling out driverless cars as fast as possible?

lambda 8y ago

Because of the uncertainty behind something new, you should have strong evidence that driverless cars are substantially safer than cars with drivers. Additionally, if driverless cars are being tested on public roads, you need laws and practices that are robust enough to prevent the kind of accident that killed someone. They had only one operator in the car, who was responsible for monitoring the car and logging information about its performance. They had specifically disabled safety systems which would have caused it to brake sooner. Even if its expected that driverless cars will reduce fatalities due to drunk or distracted driving, you need to have sufficient evidence of the improvement, safety systems in place for test systems, and public discussion of the tradeoffs and what situations driverless cars might perform worse in than human drivers, because lots of people have behavioral expectations based on human drivers; things like making eye contact with a driver before crossing the street. So yes, it is good to be working on and testing driverless cars, but you need to have a serious sense of humility and safety about it, and not push it, because even if you are incredibly safe you can expect some backlash, but if you miss the mark at all and oversell the safety, you'll get a lot more backlash, with justification, much sooner, which could set the whole program back significantly.

the8472 8y ago
> you should have strong evidence that driverless cars are substantially safer than cars with drivers. No. Considering the other benefits of self-driving cars, such as increased mobility for the elderly, you can at most require that they are as safe as cars with drivers, not necessarily safer. If the benefits were great enough (reduced pollution indirectly saving lives? saved life-years due to more efficient traffic?) one could even accept that they could be a little less safe as far as direct fatalities are concerned.
- miranda_rights 8y ago
  > If the benefits were great enough (reduced pollution indirectly saving lives? saved life-years due to more efficient traffic?) one could even accept that they could be a little less safe as far as direct fatalities are concerned. I disagree. I'm finding it hard to come up with an example that would make me think it's ok to have more people hit and killed by cars. I'm more than ready for self-driving cars to be a reality, but not if more people will die. "Saved life-years" is not good enough.
  - the8472 8y ago
    What does it mean to avoid a fatality other than saving that person's remaining life-years? > I'm finding it hard to come up with an example that would make me think it's ok to have more people hit and killed by cars. Entirely speculative: Heart failure is the leading cause of death in the US. If 100% self-driving traffic could let EMTs arrive faster at the scene that could save more lives than a slight increase from traffic accidents.
- bumby 8y ago
  I understand your point but I doubt it would work well in reality when it comes to public opinion/backlash. Humans are disproportionately swayed by proximate causes, so deaths directly attributed to self driving cars are going to be more heavily weighted in the court of public opinion.
- masonic 8y ago
  increased mobility for the elderly decreased, in fact. A self-driving vehicle can't go anywhere a cab or rideshare can't, while the latters have drivers who can assist loading/unloading elderly and their mobility aids.
  - glenra 8y ago
    Somebody with poor eyesight and slow reaction speed is an unsafe driver but doesn't need any particular help loading/unloading from a car. A driver who can help unload is a fine thing, but having somebody drive just so they can help unload seems like a ridiculous waste of resources. A car is ready whenever you need it; a car-and-driver might not be. And yes, you could schedule a car-and-driver to be available at a particular time...but you could similarly schedule somebody to be available to load/unload at the destination.

gamblor956 8y ago

Self-driving cars have killed at least 3 people this calendar year despite there being several millions fewer self-driving vehicles on the road. Scale up those proportionate rates, they would be killing far more than 10,000 people a year. So at this point, self-driving cars still have quite a bit of work to go to prove that they are actually safer than a drunk driver, let alone a sober driver.

loeg 8y ago
If you think drunk drivers are bad, sober drivers kill more than twice as many people! The actual number to beat is something like 37,000 annually[0]. And I suspect your "millions" factor has an extra zero or two when you account for (a) miles driven (self-driving cars are probably more highly utilized) and (b) driving conditions (city is more interesting for training data than highway, whereas probably more person-driving is done on highways). [0]: https://www.nhtsa.gov/press-releases/us-dot-announces-2017-roadway-fatalities-down https://www.nhtsa.gov/press-releases/us-dot-announces-2017-r...
- masonic 8y ago
  Your source doesn't differentiate between sober drivers and drivers not tested for any/all impairing chemicals. Unknown sobriety != sober.

jdavis703 8y ago

Wouldn’t mandatory breathalyzer ignition locks solve this problem? This is a proven technology that shouldn’t kill anyone as it’s being rolled out. Move fast and break things is great for websites and apps, but when human lives are at stake it requires strict ethical considerations.

loeg 8y ago
Interlock breathalyzers have a host of problems that make indiscriminate application problematic. Additionally, judicial remedies can only be applied after someone has already been caught drunk driving. So even if we applied them to all prosecuted drunk drivers, it is not a complete prophylactic.
rblatz 8y ago
What about all the existing cars on the road? What about manual cars that can be push started? Who installs and certifies these, they need calibrated frequently, they are extremely simple to bypass for anyone with even basic vehicle knowledge.
- jhbadger 8y ago
  If the law starts now, in 5-10 years the vast majority of old cars will be junked. Sure, there will still be a few old cars, but they will be an insignificant minority. It's not unlike how other rule introductions worked. Nobody banned old cars without seatbelts, they just required that all new cars have them.