Bad drivers should have their cars driven by robots (now with link)

A while ago I had one of those “Someone on the Internet is Wrong” arguments with the authors of an article arguing that we would need massively more evidence before we could conclude that autonomous cars are safer than those driven by humans. Rather than dig back to find those arguments again, I thought I’d <a href="http://I thought I'd link to this Bloomberg piece and, in particular the following passage”>link to this Bloomberg piece and, in particular the following passage

GM’s autonomous test cars were in 22 accidents in California last year, according to data from the state’s Department of Motor Vehicles … In a November interview, GM President Dan Ammann attributed the accidents to testing in a dense urban environment and noted the company’s cars weren’t at fault in any of the incidents.

Suppose that in any crash between autonomous cars and humans, each is equally likely to be at fault. What is the probability of seeing 22 crashes caused by humans and none by autonomous cars. Obviously, it’s the same as that of a fair coin showing 22 heads in a row, which is 2^-22 or about 1 in 10 million.

Of course, the drivers involved in the crashes aren’t likely to be a random sample of the population. As is standard in such things, the 80/20 rule applies: 20 per cent of drivers are responsible for 80 per cent of crashes and traffic infringements. THe 80/20 rule is derived from a Pareto distribution, and we can apply it a second time to say that 20 per cent of the remaining 80 per cent of drivers are responsible for 80 per cent of the remaining 20 per cent of crashes. That is, 36 per cent of drivers are responsible for 96 per cent of crashes. On that basis, it’s perfectly possible that the remaining 64 per cent of good drivers are as good as autonomous cars or even better.

It might also be argued that autonomous vehicles may fail in defensive driving, that is, in reducing harm in a crash caused by the failure of another driver.

Still, it seems pretty clear that autonomous cars are a lot better than the drivers responsible for most crashes and infringements. It isn’t that hard to identify a lot of these drivers before they kill themselves someone else, since prior driving record variables, particularly a driver’s prior traffic citation history, are the most consistent and powerful predictors of subsequent accident risk. Now that cars don’t need steering wheels or pedals any more, there’s no obvious reason to put people with bad driving records back in charge of them. Bad drivers should have their cars driven by robots.

40 thoughts on “Bad drivers should have their cars driven by robots (now with link)

  1. I got a 503 error on the last hyperlink.

    This is a fascinating subject to me. I’ll try to avoid the “someone on the internet is wrong” technical arguments but instead focus on the practical (and therefore ethical) limitations of navigation with autonomous systems. The fundamental ethical difficulty with requiring the use of autonomous vehicles is that they don’t handle “edge cases” well and therefore limit mobility in the real world (see the second link above). Putting aside statistical arguments about safety, even the strongest advocates of self-driving vehicles admit they don’t navigate well everywhere, such as over rough, unimproved roads or in very tight quarters (think of narrow streets prone to total blockage), and even residences in primitive locations. As a practical matter, this would mean that requiring the use of autonomous vehicles would also mean limiting the ability of those assigned to such vehicles to go wherever they want, or need to go. I personally do not believe this type of navigational problem will be solved anytime soon. Of course, that could be wrong (anticipating the aforementioned someone on the internet is wrong) but we couldn’t do it this week anyway.

    Are you willing to limit the mobility of individuals if they have a record of dangerous driving? It’s not as strict as requiring the use of rail or bus, but for many individuals it seems quite restrictive. Many would not notice any difference in their personal mobility but many others would be inconvenienced or even be forced to make major lifestyle changes. Would everyone else be safer yes, but at what cost of personal freedom?

    Basically I’m putting forward that while your idea may be suitable in certain geographical areas, it will not fit all individuals in need of robotic cars. I think it’s apparent that these edge cases will eventually be addressed technically but aren’t right now and it’s difficult to predict at what point they will be addressed completely, since it is a matter of ever diminishing returns on investment for the experts working on the matter.

    As a future solution, I personally think we will probably all be better off when software is perfected to control personal transportation so in that sense I support your proposal as an eventual goal. I certainly have seen drivers that I thought were bad enough and inconsiderate enough of the safety of others that I could agree they should be forced to use self-driving vehicles even if it meant imposing limitations on mobility but thought to point out the ethical pitfalls of such a requirements. It’s bound to end up in litigation and legislative nuisance at some point.

    One other point: there are the car/driving enthusiasts. They are going to be a PITA on the subject of requiring the use of self-driving vehicles. I have to admit I pretty much don’t like those guys anyway and think they are much of the problem so I say make them do it first, but also think that will be a major lobbying group to contend with.

    For more on this subject from a technical standpoint, I refer you to the website of Rodney Brooks, an Australian expert on robotics and artificial intelligence (website not listed here to comply with site rules – was he one of the authors you argued with?), and go argue with him if you disagree on the technical considerations of autonomous navigation.

  2. I agree. I would give bad drivers some more choices though. They could also be driven by taxi, bus or train drivers. 😉

    The idea that a car licence (or motorcycle licence) is some kind of right, and some kind of rite of passage, belongs back in another era.

    One of the issues is policing compliance. The most dangerous drivers of all are those who ignore licence disqualifications, fines and even jail stints and then just keep on driving. How do we stop these recalcitrants? Perhaps each private car key should have encoded a licence key code issued linking one private licence to one private car (and a few variations on that to cater for families):- the code pad being in the dash of the car. Those disbarred from a licence could not in any way get a code to run their own car.

  3. The electronic senses, computing power, and communications connection required for automated driving will probably cost less than providing manual controls for a human driver. So for the type of person who buys new cars, requiring them to use a robo-car if they demonstrate they are not safe drivers will not a great loss for them.

    But once the technology has proved itself, we’d be fools not to require every new car to have it. So people who want to drive themselves could, but the car would take over in an emergency situation to prevent people being harmed. If a human can demonstrate they can drive better than a machine in an emergency then they could be permitted to drive all the time, but I think that would be difficult to do. We still have airline pilots because their careers are focused on flying safely while the very large majority of drivers are amateurs.

    But for people who do want to drive manually, a robo-car would be an excellent teaching tool. I’m just not sure where that particular skill would come in handy once the stock of non-robo cars are recycled, converted, or crashed.

  4. Popular attitudes to accident risk indicate that humans place a high value on personal control, and demand that a vehicle driven by another or a robot be far safer than one they drive themselves. The factor is at at least ten. Fortunately autonomous vehicle systems will easily be able to meet the bar. They never drink, lack sleep, get involved in arguments with spouses and children, ignore speed limits, jump lights, and get angry with the idiots driving other cars.

  5. “Still, it seems pretty clear that autonomous cars are a lot better than the drivers responsible for most crashes and infringements.”

    Erk, while it’s theoretically possible that autonomous car are better, I’d be leery of taking the word of an officer of a company that is trying to develop and market an autonomous car. Even assuming the best of intentions, the tests atm will be around the ‘best conditions’ for the cars, when it gets into the real world, we’ll find out how close these are to the average condition, and by how much the cars need the best conditions.

  6. James is right that to get public acceptance autonomous cars are going to have to be a lot safer than humans. Aviation is the precedent here – the computers have made commercial flying far safer than a few decades ago but those crashes that do occur for precisely that reason get far more publicity and generally arise from human deskilling when the computers are doing all the routine stuff (AF447 is the canonical example – the pilots did not recognise an ordinary stall because the computers do not normally allow one). This has led regulators around the world to be very reluctant to allow further automation.

    It is far easier to make a safe pilotless plane than a safe driverless cars because the airports and airspace is far more controlled than roads. That’s the reason I think very large scale infrastructure changes would be need to make true driverless cars, and that’s not going to happen. Though I do expect driver aids to get better and better.

  7. @Factory

    “I’d be leery of taking the word of an officer of a company”

    If you follow the links you get to the original accident reports. I didn’t check all of them, but i satisfied myself that Ammann’s claim was properly documented.

  8. There is no need to change the laws to get bad drivers out from behind the steering wheel. Existing laws could merely be enforced more. This is not difficult to do now that human eyes are no longer required to detect unsafe driving.

    But rather than bludgeon people with old laws applied 10 fold, I’d prefer a more sensible option. More frequent but lesser punishments. With perhaps a provision that any fines accumulated will be returned if a person voluntarily gives up self driving.

  9. With the accident record of self driving cars, the robocar is very unlikely to be at fault because they tend to follow the rules, but it is still possible a human driver may have been able to avoid some of those accidents. However, it is clear that the accident rate overall is low already and this is only going to improve with experience.

  10. Smith, most people who have their licenses taken away don’t drive. If you think the rate of unlicensed driving is too high then people could be photographed when their licenses are suspended and a camera with facial recognition software could recognize them on the roads and they could be arrested after the fact. Vans with dark tinted windows or people driving while wearing disguises may need some extra surveillance to determine who is driving them, but some people would consider this extra amount to be a good thing.

    A couple of years ago this suggestion would have been nuts, but the problem has been solved. In the lab at least, computers recognize faces slightly better than humans. This was achieved through software teaching itself.

  11. @Ronald

    “most people who have their licenses taken away don’t drive.”

    But the ones that do are likely to be the most dangerous, such as drug-addicted Craig Whitall who over Christmas killed a family in NSW (and thankfully himself), while driving a four wheel drive. He was not just unlicensed but had over 60 driving offences including multiple convictions and jail time. for driving while disqualified.

    Facial recognition software should stop people from driving before they drive. iPhones can now be unlocked with this software. The same should be true of car engines.

  12. Smith, a camera with a kill switch could be installed in every non-robo car, but it is far better to let drunken suspended drivers kill people through lousy illegal driving and use the money to instead reduce the incidence of malaria and dysentery, as more lives will be saved. Or if you are one of these people who are picky about dotted lines on maps, cameras/sensors that stop cars backing over children should save more lives, even though I think we are considerably better at not backing over children than Americans.

  13. “Bad drivers should have their cars driven by robots.”

    Or they could take an Uber. It’s probably cheaper, once you factor in rego, insurance, depreciation, maintenance and petrol/electricity.

  14. @Smith
    Smith, it’s dangerous to get your information from talkback radio and the Murdoch press. I can’t find any other sources in a quick search that make, or that fail to contradict, your claims.

    Craig Whittall certainly had a bad driving record. But he was not driving unlicensed. This 51 year old grandfather was driving on P plates, after getting a licence and having done his suspension and penalties for past offences.

    I don’t think the number of his past driving offences was anything like 60, but sources are unclear. It is, of course, hard to look up someone’s full criminal history without personal permissions.

    He is widely reported to have been heading home after attending a methadone clinic, the nearest to his home. Whether he was responsible for the accident, whether he was impaired in driving by drugs, or whether he made a driving error, will no doubt be assessed in the likely inquests.

  15. I heard most of the accidents with driver-less cars on test were small bumps from other cars due to the fact that the driver-less are more cautious than human drivers .They surprise people because they wont break rules. For example ,when merging or entering a busy lane it is normally necessary to barge a bit – driver-less dont do that .If driver-less cars are identifiable they will be easy to bully .

  16. How will driver-less cars react when teenagers start playing ‘chicken’ – jumping out in front of the car and forcing it to drive off the road? Will they then be programmed to not give priority to pedestrian lives to discourage this behaviour?

  17. @Smith I do not think you are representing Craig Whittall’s driving history correctly. From the article you cited “…reportedly had a long criminal history of more than 60 offences, was a known drug user and was on his P-plates… Whitall had been jailed for driving while disqualified and had only recently got his licence back when the crash occurred.”

    I would suggest that though we know at minimum he has had his license suspended and been convicted of driving while disqualified, the remaining 59+ offences may have nothing to do with driving. We just do not know.

    Speaking only personally, I look forward to the introduction of driverless cars, because I am lazy!

  18. If this was based on the PLOS One paper of Favaro et al (Sep 2017), then those authors calculate that the AVs experienced accidents 10 times more frequently (given distance travelled) than human drivers, even though they were largely low speed rear-end, and so scored as “not at fault”.

  19. @Loco Jack

    The headline says “serial traffic offender”. “Serial” means “repeatedly committing the same offence and typically following a characteristic, predictable behaviour pattern”.

  20. You would think that the worst 20% of drivers would be priced off the road by their insurance companies pretty quickly. Why does that not happen?

  21. @Tom Davies

    They drive unlicenced and unregistered. The only way to stop them is a technological solution. This could utilise a system requiring a private driver’s license to be electronically keyed to a given vehicle which they legally own or lease. That one person can drive that one vehicle. There would be ways to set this up and to permit family members to all be able to drive a family car. It could be combined with face recognition technology in the car: a system of multiple checks.

  22. Smith is at it again. ‘Serial’ means repeated – but none of the other things he said. And this was a headline, not reporting, and not based in any way on known offences by Whittall.

    Now, who was at fault in the accident? The police say someone crossed onto the wrong side of the road…but they won’t say who did. Maybe Whittall was drug-impaired…but no one has yet said so and whether a maintenance methadone user is an impaired driver seems to me likely to need expert evidence.

    Iconoclast chimes in to say some drivers drive both unlicensed and unregistered. Indeed. But Whittall was driving licensed and registered.

    This is a witch hunt against the underclass. Get them off the road forever: when unpaid fines for whatever offences suspend their licences, and when they have to drive because of where they live and are convicted then of driving unlicensed, don’t let them ever get a licence again. But why? Driving unlicensed is not, inherently, unsafe driving. And even unsafe drivers can learn better, so someone who has had to qualify for a licence again after a period of suspension should be allowed to do so. Maybe, for those with a dangerous record, raise the bar for qualifying again. More than that is just Telegraph/Tory speak.

    Now, if we want driver standards substantially raised, we would have to face up to a system in which much more substantial numbers of people would not be or be able to be drivers. What is our transport alternative for them, other than treating them as untermenschen?

  23. @ChrisH

    I have advocated better public transport as have others on this blog. There’s a difference between the underclass and the criminal underclass. Remove inequality and then if some still turn out recalcitrant or criminal socially they still have to be restrained in some ways from damaging others.

  24. @Ikonoclast
    So answer my point. People who lose their licences may not do so for dangerous road behaviour. Poeple who have lost their licences for whatever reason may later get licences again.

    Should they be prevented from ever driving again, because they once drove while disqualified?

    Should they be prevented from ever driving again, if they ever did something dangerous? (Like, say, speeding, for which so many who are not underclass lose so many points every year?)

    ‘Better public transport’ is not an answer to how transport works if most people become permanent non-drivers. It is no better an answer if it is the underclass who are singled out for non-driver treatment for offences which leave the comfortable protecting their points but paying their fines and still driving.

  25. @ChrisH

    “Should they be prevented from ever driving again, because they once drove while disqualified?”

    Yes, on a sliding scale, unless they can demonstrate extenuating emergency circumstances requiring their driving on a public road. Being caught three times drives disqualified without extenuating circumstances should certainly result in a life ban from driving if alcohol or other function impairing drugs are involved even once. An automobile can be a lethal weapon.

    “Should they be prevented from ever driving again, if they ever did something dangerous?”

    It depends how dangerous obviously. If they deliberately tried to run someone over, yes they should be banned for life. If they repeatedly drive under the influence they should be banned for life.

    Don’t forget people who die from these recidivist people. The person who dies gets a death sentence. Their surviving extended family get a life sentence of grief. Don’t cry me a river for seriously dangerous drivers. It has nothing to do with the underclass issue. That is classic misdirection.

  26. @Ikonoclast


    Underclass my arse.

    I have checked Thomas Paine’s Rights of Man, and there is no right to a driver’s licence for habitually dangerous drivers.

  27. So Ikonoclast thinks someone who drives while disqualified should, yes, be prevented from ever driving again…on a sliding scale. So, no.

    So Ikonoclast thinks at a sufficiently dangerous driving record someone should be banned for life. That could be just one thing, if it is very and deliberately dangerous; or it could be an accumulation of things, making them a ‘seriously dangerous driver’ forever.

    What has this to do with Whittall? He had a lot of, unspecified, offences. Well, he was an opiate addict on methadone. There’s no list of offences showing he had many, or even any, dangerous driving offences. He had been disqualified as a driver, served his time for that, and requalified for a licence.

    How does an accident, where someone went onto the wrong side of the road, go to his account only when the police refuse to say who was on the wrong side of the road?

    How does the accident repeat a history of dangerous driving, when we know of no such history and we don’t know who or what caused the accident?

    Perhaps Smith should start with another Thomas Paine work: Common Sense.

  28. @ChrisH

    A bigger issue is being discussed here. You are the one who tried to derail it by talking about Whittall. No poster before you mentioned Whittall by name. It was you who introduced the red herring and then you complain when others won’t take the red herring as bait.

    You ask “What has this to do with Whittall?” Well, nothing. His legal case, if any, is still in the future. This is about decided cases. Events that have occurred. Real evidence that it is in for many cases. A driver’s licence is privilege not a right.

  29. I responded to Smith’s post. Which instanced Whittall. By name (misspelt).

    Smith instanced Whittall to make a wider case. He argued that unlicensed drivers are the most dangerous and he, like you, went on to make a version of the ‘never allow to drive again’ case.

    Apparently pointing out that the claims against Whittall are unfounded, and that someone who has been caught driving unregistered might reasonably do their suspension, serve their penalty, and then be able to qualify for a licence and drive again, is to ‘derail’ your claims.

    So it is.

    Apparently pointing out that a much higher driving qualification means a much smaller pool of drivers and a much changed transport system is to ‘derail’ your claims.

    So it is.

  30. @ChrisH

    Perhaps it might be useful to consider instead a case where the claims of habitual carelessness have been substantiated? Like this one in Aotearoa where a truck driver has now run down and killed two cyclists. Both, of course, accidental “never to be repeated” incidents.

    How many people should he be allowed to kill before he loses his driving privileges?

  31. This is the same problem of anecdote swamping discussion.

    The truck driver had an accident thirteen years ago. He was dealt with for his driving offence. He was suspended and returned to driving. I have no more idea than you how many kilometres he has done since, with what safety record.

    Now, was he at fault in this latest collision? Will he lose his licence, be imprisoned, or what will the consequences be? You don’t know and neither do I. Yet you ask ‘how many people should he be allowed to kill’, as if his responsibility for this latest death were established. You say that his habitual carelessness has been substantiated. Yet you have no basis to say so.

    I distrust the ready answer ‘ban drivers forever for being involved in a fatal accident’. I think people can be kept off the road for a while, required to requalify to be on the road, and perhaps be subject to limitations, depending on circumstances and depending on their real driving faults and correction of those faults.

    But if a ban’s what you want, and regardless of fault or circumstance, then we have to discuss the consequences. Those consequences are keeping many banned people from their only transport option, and perhaps from their employment, or providing a very different transport system in which driving is not required by most people or for most heavy transport.

    The original post posited the very strongest version of autonomous vehicle: the one in which the vehicle does everything and no passenger can override (else there is no solution to the problem of the dangerous driver, who may still intervene dangerously). We haven’t got that available yet and may never have it (for social as much as technical reasons). If we had it, we would still need to have a transport system that made this available to all or pretty much all the population.

    There are people who fail to learn. We find this out when they show they failed to learn. A system where any error (of this or that kind) denies any opportunity ever to relearn and ever to requalify is like our crime factories, ever larger even as crime continues its substantial long decline. Done time? Banned forever from an ever-increasing list of jobs; and in practice put permanently to the bottom of the list of applicants from those jobs for which there is no formal ban.

    I am a long-term commuting cyclist myself. I hate being monstered by other vehicles, and by heavy vehicles. I know that overall other vehicles are very likely to be the cause of any accident involving a cyclist. But for any particular accident the other vehicle may not be at fault. For any particular accident the fault may not be great (however serious the consequences were). For any particular accident the driver may face serious consequences and may have to requalify to drive. Automatic exclusion from driving, forever, based on a particular consequence (a death) or a particular offence (driving disqualified), regardless of circumstances of the event, is likely to be unjust and likely to apply unreasonably.

    How many straw men from particular driving incidents make a good argument? I think none do so.

  32. @ChrisH

    Yes, you’re right, we can’t use a single anecdote to generalise so the outcome of that case doesn’t matter. What we can do is take your “overall other vehicles are very likely to be the cause of any accident” mis-statement and look at the actual data. It would help if we looked at all crashes rather than only accidents, because it’s very rare to be able to prove intent or wilful negligence in a motor vehicle crash (and few law enforcement agencies are willing to try*). But those statistics show, per the OP, that a few very bad drivers cause a disproportionate amount of harm. Those drivers are also often identifiable by the trail of minor offenses they accumulate before they get caught doing/do something dramatic.

    I think that it’s quite reasonable to say to someone “maybe you are just unlucky, but there’s a limit. Sorry about that”. If that means that someone can’t drive who is genuinely blameless, they can join the long list of other people whose lives have been screwed up by government policy. We do, IMO correctly, make laws on a utilitarian “greatest good” basis a lot of the time, and practicality also supports that approach. For example, we don’t require proof for criminal conviction, just “beyond reasonable doubt”.

    Specifically in the context of us paying 1000+ lives a year to have cars on the road, I think it’s perfectly reasonable to look into ways to reduce that cost. If we can come up with a strategy that trades off a little bit of liberty for a great many lives why would we not do that? It’s no worse than continuing to run Don Dale despite knowing that even if it worked as designed, it wouldn’t work very well.

    * “mens rea” is the original thoughtcrime in that respect.

  33. Not being worse than Don Dale, or any other part of our crime factory system, is a pretty low bar.

    Facts are just what’s needed. The crash stats are a good place to start.

    But the number of minor offences, and what offences qualify, is a problem. What offences show unfitness? What are the stats, too, on people returning to requalify as drivers after a period of suspension or other inactivity? These are worth looking at and thinking about.

    ‘Ban them forever on suspicion’, or for a single bad-outcome event for which they aren’t shown to be responsible…that’s like the Minister for Immigration (Ruddock) who decided it was right and legal to deport a victim of wife-bashing for failing the character test: she was associated with a wife-beater.

  34. Simply requiring drivers to re-qualify after having their license “suspended” would probably solve a lot of the problems, because I suspect many would fail. More would likely not even try, but either way, further offending would hopefully face a much more skeptical system. “you can’t pass a driving test, yet you kept driving”… at some point it will become criminal rather than merely naughty.

  35. I mention Don Dale to show that your “this must be exemplary” is not the standard used by Australia, or Australians, when evaluating proposed laws or adjustments to the legal system. Even “must be morally defensible” is more of an aspirational goal.

    I agree that ideally we should aim for laws that are just, reasoned and moral, but sadly such laws are very difficult to write and even harder to pass through parliament. Aotearoa’s nuclear free legislation, for example, was very unpopular even with countries that agree with the ban on developing and using nuclear weapons. Obviously Australia didn’t oppose the ban because it was unjust, immoral or unreasonable… which is my point.

  36. then they came for the rapists, and I was not a rapist so I did not speak out? #metoo

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s