Safe at any speed -- until the ignition cuts out
October 30, 2014 7:32 AM   Subscribe

Countless articles have been written about General Motors and its massive recalls earlier this year. What hasn’t been fully told is how GM might have gotten away with multiple counts of consumercide were it not for the efforts of three men: a Georgia lawyer, a Mississippi mechanic, and a Florida engineer.
How an unsafe ignition lock helped kill tens of people and the people who found out the truth about it.
posted by MartinWisse (66 comments total) 22 users marked this as a favorite
 
consumercide

not sure why "homicide" doesn't cover it here.
posted by stupidsexyFlanders at 7:49 AM on October 30, 2014 [6 favorites]


it’s not out of the realm of possibilities that he and others at GM could face criminal charges.

Wow. Not "out of the realm of possibilities." It's like Bigfoot or UFOs.
posted by Marmaduke Hammerhead at 7:54 AM on October 30, 2014 [3 favorites]


When I first heard about this particular defecr, I was quite ready to place a large portion of the blame on drivers. One should not crash one's car merely because the engine suddenly stops running. Back when I was driving a crappy 80s Ford product, it happened to me all the freakin' time and was never any kind of serious danger beyond the usual risk of being stopped on the side of the road.

Then I got to the part about how the steering lock also becomes engaged and I was pretty much scared shitless. That is fucking insane. If I had to choose between a car that would accelerate on its own and one that would suddenly become impossible to steer, well, it's pretty obvious which I'd pick.
posted by wierdo at 8:04 AM on October 30, 2014 [13 favorites]


Then I got to the part about how the steering lock also becomes engaged and I was pretty much scared shitless. That is fucking insane.

No shit. Talk about a design flaw guaranteed to kill somebody.
posted by Pogo_Fuzzybutt at 8:11 AM on October 30, 2014 [2 favorites]


If you think sudden accelerations and locked steering is bad, just wait until google is driving every car.

As a software engineer who has watched his coworkers work, I'm scared of any machine with a computer inside.
posted by DU at 8:17 AM on October 30, 2014 [18 favorites]


Then I got to the part about how the steering lock also becomes engaged and I was pretty much scared shitless. That is fucking insane.

I think it was a similar situation with the Toyotas/Lexus' (Lexi?) stuck accelerator issue back in 2009.
It seems that in some cases, the drivers were not able to shift to Neutral for some reason - I think on the Lexus the shifter was controlled by an electronic switch.
posted by bitteroldman at 8:22 AM on October 30, 2014


These cars caused my best friend to have a horrifying accident -- when I first met her, when we were both first year second grade teachers in a not-super-great school, she was still (slowly) recovering from the accident -- she had spent the previous summer in a wheelchair and still had a lot of trouble climbing stairs and moving around.

When this information came out and she told me about it, she started to cry, because she said she'd just felt this enormous sense of guilt that she hadn't even realized she'd had until she felt the relief from it. She assumed the accident was her fault in some way; she remembered that the car wasn't steering right but she figured it had to be because of something she'd done. She also remembered that the airbag didn't deploy, but she wasn't sure why.

She's done an incredible, incredible job recovering and other than some pretty intense scars you'd never know she had an accident, but being in one of these cars took away a lot from her; it took years for her to recover, it caused her and her family pain, and it almost killed her. Thank God she turned out okay but it so easily could have gone the other way.
posted by Mrs. Pterodactyl at 8:23 AM on October 30, 2014 [17 favorites]


Where does it say that the steering lock engaged? The closest thing that I can find to this in the article is that power steering assist may have failed, which would only make turning the wheel harder at low speeds.
posted by indubitable at 8:26 AM on October 30, 2014




DU: "If you think sudden accelerations and locked steering is bad, just wait until google is driving every car.

As a software engineer who has watched his coworkers work, I'm scared of any machine with a computer inside.
"

This. The only reason I am afraid to fly is the friggin autopilot. I do not like trusting my life to some software coder. Sure, it has been tested and tested, but something unusual will happen and it will fail. I love to fly in a four seat Cessna. I have gone up about 5 times with a friend who I trust.
posted by 724A at 8:34 AM on October 30, 2014 [1 favorite]


This. The only reason I am afraid to fly is the friggin autopilot. I do not like trusting my life to some software coder. Sure, it has been tested and tested, but something unusual will happen and it will fail. I love to fly in a four seat Cessna. I have gone up about 5 times with a friend who I trust.

That's nice, but autopilots are not why people die in plane crashes. People die in plane crashes (statistically speaking) when they fly in small general aviation planes.
posted by atrazine at 8:40 AM on October 30, 2014 [48 favorites]


As a software engineer who has watched his coworkers work, I'm scared of any machine with a computer inside.

Liability and a culture of safety---that's what separates an "engineer" from a PE. Nobody really cares if a media player is kinda crashy (e.g. winamp, iTunes). Many customers, insurance companies and law-makers will care if a car is crashy.

These are well-understood problems in both aerospace and health devices. Software standards are very different there. I think the car industry is on the cusp of that: there's some awareness of the issue for automotive computers, less for the other systems in the car, like the media player in the dash. The two should just never, ever talk to each other.
posted by bonehead at 8:40 AM on October 30, 2014 [7 favorites]


This. The only reason I am afraid to fly is the friggin autopilot. I do not like trusting my life to some software coder. Sure, it has been tested and tested, but something unusual will happen and it will fail. I love to fly in a four seat Cessna. I have gone up about 5 times with a friend who I trust.

At least airplane software is intelligently designed. Nature on the other hand just kept flipping random bits and if nature screwed up we just died. You really want to leave your safety in an airplane to a human designed by evolution?!?
posted by Talez at 8:40 AM on October 30, 2014 [3 favorites]


but something unusual will happen and it will fail. I love to fly in a four seat Cessna. I have gone up about 5 times with a friend who I trust.

I know several people who have died in Cessna crashes attributed to "pilot error." They were all safe pilots who I would have trusted with my life. It's always strange to me that we trust software less than wetware.
posted by muddgirl at 8:42 AM on October 30, 2014 [5 favorites]


As a software engineer who has watched his coworkers work, I'm scared of any machine with a computer inside.

Sure, but a meat brain is just a crappy, buggy computer programmed by half a billion years of "Eh, close enough, ship it." So I'm more scared of things controlled by meat.

Though in this case the possibility of "Things controlled only by momentum and inertia" is, yeah, mebbe the thing to be mostest afraid of.
posted by ROU_Xenophobe at 8:43 AM on October 30, 2014 [14 favorites]


Except for sharks controlled only by momentum and inertia.
posted by ROU_Xenophobe at 8:43 AM on October 30, 2014 [3 favorites]


I'm scared of any machine with a computer inside.

This seems like a pretty silly attitude. Every car sold since the 80s has a computer inside, and a fair percentage of modern cars are "drive by wire" where there's no direct link between the throttle and the engine except through that computer. They work fine.

I certainly wouldn't want a car where the embedded systems' software was developed the same was as, say, typical business software, but that's because we do business software development cheaply and poorly, and accept a ridiculously high defect rate, not because there's anything intrinsically wrong with software. You could build business or desktop software to the same levels of reliability demanded by embedded systems in life-safety applications, but it would be stupendously expensive.

The GM situation is demonstrative: the computers and software in the car functioned as designed, but people died because of a few cents worth of spring steel. If the original defect had been in the code in one of the car's computer systems, it would have been harder for GM to perform the little switcheroo that they did.
posted by Kadin2048 at 8:47 AM on October 30, 2014 [7 favorites]


Sure, but a meat brain is just a crappy, buggy computer programmed by half a billion years of "Eh, close enough, ship it." So I'm more scared of things controlled by meat.

But these terrible meat computers are *really great* at predicting what other terrible meat computers are going to do. Not to mention the thousands of years of training for users (and legal precedence) with mechanical apparatus vs about 50 years for computer devices. That's on top of the fact that computers are much much MUCH more complicated than non-computers. So many, many failure modes, most of them never even seen or even conceived of.

When computer-controlled cars on the road, I build a brick wall at the edge of my property and leave it as infrequently as possible.
posted by DU at 8:48 AM on October 30, 2014 [1 favorite]


If that article were a work of fiction it would be a riveting page-turner; perhaps something John Grisham or Robin Cook could make into a novel. But I couldn't stop thinking about how this really happened and found myself having to stop from time to time so my anger didn't just boil over.

But one thing kind of bothered me:
Although attorneys are often justly lampooned...
Why is it that so many article portraying attorneys as heroes feel obliged to include some sort of disclaimer to the effect that "we know there are a lot of bad lawyers out there, but trust us, this guy is one of the good ones"? I am a physician and I assure you there are plenty of physicians out there who are greedy, incompetent, or just plain idiots (a number of the latter seem to go into politics). Yet no one ever says "we know Paul Broun thinks evolution is a lie from the pit of hell, but this doctor actually bothered to learn biology." It seems to me this only serves to reinforce negative stereotypes of attorneys in the public mind when in reality most lawyers are doing the best they can for their clients just as most doctors do the best they can for their patients.

He signed the email “Ray (tired of the switch from hell) DeGiorgio.”
"switch from hell" indeed
posted by TedW at 8:51 AM on October 30, 2014 [4 favorites]


It's always strange to me that we trust software less than wetware.

We've all had many experiences with machines being buggy and, as for wetware, we *are* the wetware so we instinctively trust it. It's been the gold standard for all of our species existence.
posted by Marmaduke Hammerhead at 8:51 AM on October 30, 2014


I certainly wouldn't want a car where the embedded systems' software was developed the same was as, say, typical business software, but that's because we do business software development cheaply and poorly, and accept a ridiculously high defect rate, not because there's anything intrinsically wrong with software.

I'm a software engineer at a place that does everything from realtime hardware control to scientific computing to "business software". It is pretty uniformly terrible and the people who wrote it are some of the best educated (in software and otherwise) in the world.

The terribleness is due to the size and complexity. They didn't make better code in the old days, they just made less of it. KISS means few failure modes/points.

Software, being (software) engineering on top of (electrical) engineering on top of (mechanical) engineering (and usually many more layers than that) is nearly always the opposite of KISS.

And I think it's pretty obvious I have no objection to sensor arrays in cars, that was just disingenuous.
posted by DU at 8:53 AM on October 30, 2014 [3 favorites]


But these terrible meat computers are *really great* at predicting what other terrible meat computers are going to do.

Almost every crash is that meat computers do something stupid and other meat computers aren't really great at predicting atypical things. Our meat brains are good at warning us based on experience.

Like I can get a feeling that some asshole is about to pull over abruptly into my lane based on his lane positioning and the car's attitude but it often doesn't happen. Conversely I was completely surprised by some asshole that pulled out without using his blinker and his car showed no attitude that he was about to pull out.
posted by Talez at 8:53 AM on October 30, 2014 [4 favorites]


Where does it say that the steering lock engaged? The closest thing that I can find to this in the article is that power steering assist may have failed

For the vast majority of people, this may functionally be the same thing. Power steering has allowed geometry that is much better dynamically but trying to turn the wheel (and backfeed the pump) without power assistance is extremely hard. The step difference between a wheel that is power assisted and is not (having experienced it a few times) makes the steering become (as good as) solid.

I don't think the key fault triggered the steering lock from my understanding of the issue (as in the mechanical lock that engages when the ignition barrel is fully turned off three clicks) but I did think it turned off the ignition (one click) which kills the engine and all steering assistance as a result, which would be - in every way that matters to most people - 'locking' of the steering.
posted by Brockles at 8:53 AM on October 30, 2014 [4 favorites]


But these terrible meat computers are *really great* at predicting what other terrible meat computers are going to do.

How many automobile crashes per year occur due to user error, and how many due to a mechanical or software failure?

That's on top of the fact that computers are much much MUCH more complicated than non-computers. So many, many failure modes, most of them never even seen or even conceived of.

Mechanical automobiles are much more complicated than bicycles, and there are so many more failure modes, most of them had never been seen or conceived of.

I have ethical issues with automated cars encouraging even more personal vehicle use in an age where mass transit is super-critical, but automobiles are already incredibly unsafe. Taking out the human factor removes a huge part of that. It will introduce other failure modes, sure, but if it can lead to net fewer deaths, that seems like a good thing to me.
posted by muddgirl at 8:58 AM on October 30, 2014 [6 favorites]


But these terrible meat computers are *really great* at predicting what other terrible meat computers are going to do

Are they?
Worldwide it was estimated in 2004 that 1.2 million people were killed (2.2% of all deaths) and 50 million more were injured in motor vehicle collisions
If there's a software bug it can be fixed, and then every car that uses that software will be fixed. If a person makes a mistake driving a car and crashes, that one person might learn "oh, don't brake too hard on ice", but the rest of the population will continue to make that same mistake, over and over again.
posted by EndsOfInvention at 9:01 AM on October 30, 2014 [5 favorites]


If there's a software bug it can be fixed, and then every car that uses that software will be fixed.

Your willingness to throw out all of history up until now (and spend thousands of lives in doing it) in the theory that THIS generation is ACTUALLY smart and can do it RIGHT is appalling. But not unexpected, since it's exactly what every generation does. This time around Software Will Save Us.
posted by DU at 9:08 AM on October 30, 2014


> When computer-controlled cars on the road, I build a brick wall at the edge of my property and leave it as infrequently as possible.

This is a lot like "I'm moving to Canada if Bush gets elected", and your whole premise here is a boring derail.
posted by Invisible Green Time-Lapse Peloton at 9:09 AM on October 30, 2014 [3 favorites]


Fascinating as this software discussion is, note that the problem here was a *physical* one, not a software problem: a lock that was too easy to accidently jiggle from on to off.
posted by MartinWisse at 9:12 AM on October 30, 2014 [19 favorites]


But these terrible meat computers are *really great* at predicting what other terrible meat computers are going to do. Not to mention the thousands of years of training for users (and legal precedence) with mechanical apparatus vs about 50 years for computer devices.

The current system is more flawed than you suggest here. Current human car drivers have almost no formal training at all, versus every other major form of transportation that employs professionally trained people and has some level of oversight. If a person is a terrible pilot who is completely incompetent in flying a plane, they will probably either never get to be a professional pilot or will be fired from any pilot job they have. Whereas if someone is consistently terrible at driving a car, either due to poor motor skills, poor judgment, or just general lack of skill, they can still receive and maintain a driver's license. And even people who would be great professional drivers if they recieved years of intensive training still don't have skills that are needed in life or death situations like knowing how to safely survive situations where a major component of the vehicle fails. Computers replacing high-performing humans at the top of their game concentrating on driving perfectly is one thing, but for the most part they will be replacing people who aren't very good at driving a car and aren't really paying attention to what they are doing.
posted by burnmp3s at 9:12 AM on October 30, 2014 [4 favorites]


The step difference between a wheel that is power assisted and is not (having experienced it a few times) makes the steering become (as good as) solid.

I used to drive a VW bus that had "strong-arm steering" in lieu of the power option. It wasn't steering you could adjust with your palm or a finger or two, like modern-day cars, but I wouldn't say it was locked up like solid ice, either.
posted by a lungful of dragon at 9:20 AM on October 30, 2014


"If you think sudden accelerations and locked steering is bad, just wait until google is driving every car.

Have you ever seen humans drive cars?
posted by octothorpe at 9:26 AM on October 30, 2014 [10 favorites]


I used to drive a VW bus that had "strong-arm steering" in lieu of the power option. It wasn't steering you could adjust with your palm or a finger or two, like modern-day cars, but I wouldn't say it was locked up like solid ice, either.

I've had electronic-assisted power steering go out on me, and it felt much, much stiffer than just regular old "strong-arm steering" or even a failure in a hydraulic-assist power steering column. Not completely locked-up, exactly, but I can imagine that many, many drivers would say that they were "unable to turn vehicle." Now, this is still different than intentionally locking the steering column while the car is still in gear but powered off - the article is pretty unclear on that.

...and after writing this, I should think about driving less-crappy cars.
posted by muddgirl at 9:34 AM on October 30, 2014


There is no way one can confuse a lack of power assist (at speed anyway) with an engaged steering lock. I've just recently been reminded of what no power steering is like on my SO's RAV4. When the body shop jackholes failed to put a sensor back on the power steering was at zero assist until I reset the ECU. Now it just acts like I'm driving 80.

Basically, when the car is stopped, it is in fact very hard to turn the wheel. When the car is moving more than a couple of miles an hour, you can hardly tell the difference. A neat trick of friction makes it so.

The article posted here doesn't explicitly state that the steering lock becomes engaged, but the NHTSA reports do. It is sort of implied when it states that in some cases it appears no evasive maneuvers were undertaken by the driver. A lack of assist would notnin any way prevent a driver from swerving. I'd link, but I'm on my phone right now and don't have it handy. It may be that the lock doesn't always engage, though, since that would require the switch jump all the way back to the off position.
posted by wierdo at 9:36 AM on October 30, 2014 [2 favorites]


I used to drive a VW bus that had "strong-arm steering" in lieu of the power option. It wasn't steering you could adjust with your palm or a finger or two, like modern-day cars, but I wouldn't say it was locked up like solid ice, either.

There's a huge difference, though, between steering that is designed without power-assist (i.e. "strong-arm") and power-assist steering that loses the power-assist. In the latter, you are fighting against all the mechanicals and belts and whatnot that are now idle and unpowered. It's much, much more difficult to steer.
posted by Thorzdad at 9:36 AM on October 30, 2014 [4 favorites]


I wouldn't say it was locked up like solid ice, either.

Manual steering cars have very different suspension geometry to cars designed to have power steering. There are advantageous geometries (like increased caster) for road holding and handling that would produce too heavy a steering weight for people to realistically turn without ending up with 8 turns lock to lock. A car/bus with manual steering is not in the least bit comparable to a vehicle designed for power steering. Even pushing the (inactive) power steering rack and back-feeding the stationary pump when it fails is not an insignificant steering load.

Drive a modern car in a deserted parking lot at 20mph, give yourself lots of room and flick the ignition off and try and steer it. Even when you expect it, it's a massive, massive increase in steering effort.
posted by Brockles at 9:36 AM on October 30, 2014 [8 favorites]


A lack of assist would notnin any way prevent a driver from swerving.

You're comparing a functioning steering rack and pump assembly that is (essentially) idling to one that is inactive. Not at all similar. Your steering rack was still functioning and was providing at the very least enough assist to overcome the system drag and load itself. It was not, in any way, replicating a failed system (where the pump isn't even turning). Try my parking lot suggestion if you are still unconvinced.
posted by Brockles at 9:39 AM on October 30, 2014


It may be that the lock doesn't always engage, though, since that would require the switch jump all the way back to the off position.

My experience has been that, when the ignition switch is on the Off position, the steering wheel has to be in just the right position for the lock to engage. If you turn your car off, quite often, there is still play in the wheel, and you can turn it somewhat until the lock engages.

It may be that, in the case of these cars, the lock doesn't engage immediately, until the driver happens to turn the wheel to the right position and the lock engages.
posted by Thorzdad at 9:41 AM on October 30, 2014 [2 favorites]


The article posted here doesn't explicitly state that the steering lock becomes engaged, but the NHTSA reports do. It is sort of implied when it states that in some cases it appears no evasive maneuvers were undertaken by the driver. A lack of assist would notnin any way prevent a driver from swerving. I'd link, but I'm on my phone right now and don't have it handy. It may be that the lock doesn't always engage, though, since that would require the switch jump all the way back to the off position.

The article strongly implies that the steering columns don't lock up, even in the off position, or at least that manufacturers believed it to be so:
They categorized the problem as one of “customer satisfaction,” not safety. The nation’s top auto regulator would later parrot this line of argument: “If a consumer can pull a car over to the side of the road and restart it,” David Friedman testified before Congress, safety is not a problem.
If the steering column locks in the off position, then the car couldn't be navigated to the side of the road except by happenstance.

Basically it's a confusing article in this regard.
posted by muddgirl at 9:42 AM on October 30, 2014


I should note that electronic power steering doesn't have as dramatic a difference when it goes out as hydraulic, but the Cobalts in question use EPS according to the article.

EPS is mechanically very much like non-assisted steering, but with an electric motor mounted on the steering column to help turn the wheel. Hydraulic systems are rather different, but still perfectly steerable at speed.
posted by wierdo at 9:43 AM on October 30, 2014


Brockles, I know from experience how the steering effort changes when hydraulic assist fails at highway speeds. I once threw a rod (it put a silver dollar sized hole in the engine block in the process) while driving down the Interstate at 75 in a car with hydraulic power steering. It's heavier, but could not be confused with a total lock.
posted by wierdo at 9:48 AM on October 30, 2014


Since reading this I've been trying to decide whether to passively-agressively forward it each time a colleague makes a change to a part or software without reving the part or version number (even replacing a complete product with an entirely new thing under the same name and part number! arg!) Beyond that, everyone needs to be clear on why something is fixed or changed and making sure that all follow up actions are taken based on how it interacts with other parts and who needs to know about the change. In this case, the fixed ignition switch should have been followed up by a full-on recall, and dealerships, mechanics and end customers should all have been notified. I'm amazed at this failure of GM's organization and culture, though yes, this kind of thing is not uncommon where less dangerous stuff is being developed (e.g. software).
posted by thefool at 9:55 AM on October 30, 2014 [2 favorites]


If you think sudden accelerations and locked steering is bad, just wait until google is driving every car.

As a software engineer who has watched his coworkers work, I'm scared of any machine with a computer inside.


Considering that only accident to have been recorded on a public roadway with a Google car was the fault of another driver, I'm okay with Google driving every car. Heck, I'd even be okay if they used my location data to target advertising. (It would be a step up from the Google ads that just show me things that I've already purchased.)

And as a person who has watched his coworkers drive, I've scared of any machine with a person inside.
posted by dances with hamsters at 9:56 AM on October 30, 2014 [5 favorites]


I had forgotten how easy electric PAS is, actually.

Weirdo: I think you're selling yourself short, there. I've had power steering fail in a number of vehicles over the years and I wouldn't have mistaken it for complete physical lock, but I don't think you are appreciating that it is not at all the average person's response to meet 'much more than expected resistance' with 'sudden and overwhelming force from the driver' like it sounds like you did (and I did). Besides failure on the highway is no big deal either. Halfway through a corner is another matter entirely (thank you, Triumph 2500S that I nearly put in a ditch).

The average driver is much less mechanically aware than you and I - they don't necessarily make the connection in an emergency that 'this is 5 times harder to turn' means 'throw your weight against it' in the split second it takes to recover the car. People (average people) tend to panic if they have a sudden and abrupt change of their control perception. Clearly, you and I don't so much.

I suspect the times that this failure happened on a highway on a relatively straight road are not ones we are hearing about. It's the ones where it happens and they need to either steer to regain/stay on the road or avoid something that involves a sizeable steering input is when it is reported they 'locked' the steering. It's most likely that you and I would call that 'loss of power assistance' but less knowledgable types report it as 'locked'.
posted by Brockles at 9:58 AM on October 30, 2014 [6 favorites]


Sure, but a meat brain is just a crappy, buggy computer programmed by half a billion years of "Eh, close enough, ship it." So I'm more scared of things controlled by meat.

Yeah, it's all spaghetti code with design patterns tending toward "throw it all at the wall and see what sticks," but take hope: they unit test the living fuck out of those things before they go to push out the next generation of hardware.
posted by Mayor West at 10:00 AM on October 30, 2014 [6 favorites]


In this case, if my reading comprehension is reasonable, the woman started hydroplaning when she tried to be the power in the steering. At that point, making corrections was even more difficult. It also shortened the time she had to make those corrections.

This is a story about many things, but the corruption, the flagrant hiding of the engineering mistakes and the change of part without changing the part number is criminal. I think the federal government should hold GM and the engineers involved criminally liable.
posted by 724A at 10:07 AM on October 30, 2014 [2 favorites]


muddgirl: ...and after writing this, I should think about driving less-crappy cars.

It may just reduce to "don't drive GM cars". Corvettes of the same era, hardly what most would refer to as "crap", have similarly frustrating issues with the steering column lock that GM never fixed effectively. Now in this case, it would (as far as I know) only leave you stranded at the last place that you parked the car, but it is astonishingly slipshod for what is usually touted as their flagship car.
posted by indubitable at 10:09 AM on October 30, 2014 [1 favorite]


Your willingness to throw out all of history up until now (and spend thousands of lives in doing it) in the theory that THIS generation is ACTUALLY smart and can do it RIGHT is appalling. But not unexpected, since it's exactly what every generation does. This time around Software Will Save Us.

You are showing a terrible ignorance of the history of improvements in automotive safety. Cars get safer just about every single year despite the idiots driving them (and making them). It's not really debatable because the evidence is available and public.

And I am saying this as a former automotive insurance analyst who gave up driving about 20 years ago because I didn't want to hurt anybody else for the sake of my convenience.
posted by srboisvert at 10:16 AM on October 30, 2014 [9 favorites]


If my power steering were to off, and I were going 65 mph, I would be really worried to turn my wheel hard enough. Think about it - at speed you really don't want to over correct. In fact, knowing to keep your movements appropriately sized is one of the hall marks of a good driver. So, suddenly having to apply what would otherwise be enough force to spin the car out? That's effectively locked. There's NO WAY I would apply that much pressure to my steering wheel. Absolutely not. If I knew it was coming, maybe maybe I could overcome my reflexes and turn hard enough.
posted by stoneweaver at 10:39 AM on October 30, 2014 [3 favorites]


Assigning blame in this case is a bit of a philosophical issue. The car didn't suddenly accelerate or steer in the wrong direction; instead it failed to operate, admittedly at a random and unpredictable time. Mere unreliability is apparently not something GM was concerned with.

If a driver desperately needed to get to the hospital, and his car suddenly failed, and he was able to safely pull over to the side of the road but then he died, would the manufacturer be at fault? Most people would say no -- the car was unreliable but was not the cause of death.

So then how is it different if the car fails and then the driver is hit by another car? Shouldn't other drivers on the road be prepared for anything, including random stopped cars?

I am guessing that GM itself did not prioritize the issue because they felt it was a problem with mechanical unreliability, and not a true safety issue (like an airbag failure). If the car suddenly turning off is a safety problem, then isn't the entire drivetrain for the car a safety-critical component, since any problem with it cause the car to stop?
posted by miyabo at 11:07 AM on October 30, 2014


not a true safety issue (like an airbag failure).

It is exactly like that, because it IS an airbag failure. The issue is that the car turns itself off in an impact of big bump and so the airbags become disabled when they are needed the most. So.. precisely an airbag failure. The cause of it being the faulty ignition switch.

If it turned itself off randomly, it's not *that* big a deal, but the main safety concern (and cause of death) is the ignition disabling all the safety and control aspects in the car during an accident by turning them off.
posted by Brockles at 11:15 AM on October 30, 2014 [7 favorites]


Isn't the entire drivetrain for the car a safety-critical component, since any problem with it cause the car to stop?

Yes. Unexpected stops are a safety issue. The whole "maneuver to the side of the road" bullshit is just that. A convenient sounding blame shifting exercise allowing GM to continue to do nothing to address the problem. An unexpected stop on the train tracks or on an icy mountain road (or a dozen different scenarios ) cause a great risk of harm or injury.

The defective switch also had the unexpected effect of switching off the airbags and other safety features which led to a number of the deaths noted in the article.
posted by mygoditsbob at 11:20 AM on October 30, 2014 [3 favorites]


It's not reasonable to expect average drivers to, in a split second, diagnose the fact that they've lost their entire electrical system and adjust their driving behaviors to accommodate for missing power steering and anti-lock brakes. Most people have never driven a car without anti-lock or anti-skid brakes.
posted by muddgirl at 11:21 AM on October 30, 2014 [7 favorites]


For anyone who is interested, the CBC's Fifth Estate is covering this same issue tonight (not sure if the two reports are related, though hazarding a guess based on the timing, I'd say they are). One relevant point to note stated on the National last night is that unlike the US, Transport Canada does not in fact have the power to order automotive recalls, so they have to ask the manufacturer politely. *sheesh*
posted by northtwilight at 11:25 AM on October 30, 2014


The only reason I am afraid to fly is the friggin autopilot.

I know this is a little derail, but I want to point out that an autopilot failure would not have nearly the same consequences as one of these ignition switch faults. You would be much, much safer if your autopilot suddenly died, for several reasons:

-They are designed to be easily cut out of the loop if they go haywire. There are multiple ways to disable an autopilot; standard on/off switches completely shut out power to the autopilot, disengage switches turn functionality off, and you even have a bank of circuit breakers right in front of you that you can pull if it's really that dire. Also, the control servos are designed that you can manually override them (I believe they're on slip rings), so that if all else fails you can literally wrestle control away from the autopilot. On a Cessna, it's not hard to do and is part of your preflight checklist (turn on autopilot, crank the yolk left or right, make sure you can overpower the servos by hand, turn off autopilot).
-There's a lot less to hit in the air. If the autopilot dies, it's not going to send the plane careening off into a concrete barrier. You honestly probably wouldn't notice anything had gone wrong for several moments.
-In-flight failures of anything short of wings falling off or engine catching fire are not immediate we-are-dead-in-two seconds kinds of emergencies. An autopilot failure is not anywhere near that kind of catastrophe. You simply cut it out of the loop, take stock of the situation, and fly by hand.

What does kill people with autopilots more often than not is pilot error. Most autopilots in small aircraft are fairly dumb; they probably know what direction you're going (and what direction you want to go) and whether you're climbing or descending. Unless you have a fully integrated cockpit, it doesn't know your position, your speed, your altitude, or anything else that might give warning that something bad is about to happen. Cessna pilots die because they let the autopilot fly into a hillside or ask it to do things that the plane is not capable of complying with (excessive climb rates that stall the wing, for example - autopilot doesn't know any better and can't recover from the stall for you). However, these limitations are very clearly stated in the operating manuals and the training you (should) receive will teach you what you can and can't do.

The autopilots in our club planes don't even have software; they're analog computers from the 70s. I guess they're immune to failure!
posted by backseatpilot at 11:28 AM on October 30, 2014 [9 favorites]


DU: If you think sudden accelerations and locked steering is bad, just wait until google is driving every car.

As a software engineer who has watched his coworkers work, I'm scared of any machine with a computer inside.
Yeah, this is a pretty common, massively unlikely luddist philosophy.

Which is more likely to give you wrong change, a retail clerk or an ATM?
Who is more likely to fall, a human on stairs or a human on an escalator?
Which is more likely to be accurate, a gas station attendant's directions or a GPS?
What is more likely to cause landing gear to improperly function, a pilot or an automatic safety system?

Bonus round: to the nearest order of magnitude, how much more likely is a human train operator to crash a train than a fully automated system?

Since driving a vehicle without killing onself or others is one of the things humans are truly awful at, on a scale with "not making a lifelong habit of eating things that encourage heart disease", I'll take the fully-vetted software.
posted by IAmBroom at 1:28 PM on October 30, 2014 [8 favorites]


the problem here was a *physical* one, not a software problem

That's a distinction without a whole lot of difference in terms of engineering safety. It's better to say that it was a design problem, that wasn't caught by proper testing.

They system as a whole has to be safe. The distinction between a physical issue and a software one isn't one of kind, just expertise and specialist. A comprehensive approach has to be used to deal with either and in the whole, finally. Hardware and software in modern systems are not fully separable.
posted by bonehead at 2:22 PM on October 30, 2014 [3 favorites]


As a software engineer who has watched his coworkers work, I'm scared of any machine with a computer inside.

As someone with a Ph.D. in psychology and neuroscience, including specific training in human judgement and decision making theory - as well as being just a plain old person who has seen human beings - I'm scared of any machine with a human being at the controls.

(And my day job is software engineer, so, there's that, too.)
posted by dmd at 2:28 PM on October 30, 2014 [3 favorites]


that wasn't caught by proper testing.

Well, it's more damning than that. It was caught in testing, but the manufacturer (via their engineering staff) decided it wasn't a problem.

It seems, from the memos that have become public, as though there was a perceived tradeoff between key-turn torque and wear leading to failure of the electrical components of the switch. For some reason—maybe related to previous GM problems with ignition switches leading to cars stuck in parking lots, who knows—an engineer decided not to increase the key-turn torque and risk failure of the switch.

In so doing, what they did was lessen the incidence of a benign failure (car stuck in driveway / parking lot because switch has worn out), but increase the incidence of a much nastier failure (key turns off while driving).

At this point I suspect a lot of argument is going to hinge on who at GM knew about that tradeoff and the consequences of not strengthening the spring in the beginning, which was then made so much worse by the lackluster (to put it mildly) response once problem reports started coming in from the field, and then capped off by the sneaky attempt at updating the part without admitting that the first iteration was defective.
posted by Kadin2048 at 3:08 PM on October 30, 2014 [3 favorites]


> I certainly wouldn't want a car where the embedded systems' software was developed the same was as, say, typical business software, but that's because we do business software development cheaply and poorly

The automotive software industry certainly has standards and practices for safety and reliability, but you'd have to follow them. EDN:
Toyota's own internal standards make use of only 11 MISRA-C rules, and five of those were violated in the actual code. MISRA-C:1998, in effect when the code was originally written, has 93 required and 34 advisory rules. Toyota nailed six of them....

The Camry ETCS code was found to have 11,000 global variables. Barr described the code as “spaghetti.” Using the Cyclomatic Complexity metric, 67 functions were rated untestable (meaning they scored more than 50). The throttle angle function scored more than 100 (unmaintainable).

"Task X"..., an arguably outsize task handling everything from cruise-control to diagnostics to failsafes to the core function of converting pedal position to throttle angle.
More: 1 2

The unreliability of human operators doesn't absolve engineering management from the responsibility to give the fallible the best possible chance of arriving home alive.
posted by morganw at 5:28 PM on October 30, 2014 [5 favorites]


On the power steering cutting out thing, this varies SO MUCH from car to car. All cars have different steering ratios, various brands and models have different EPS or hydraulic systems. I absolutely have driven cars that with the engine off but the steering unlocked were very close to impossible to turn.

Hell, my friends toyota van that loves to stall all the time was hard enough to crank with the engine off that i'd say a small to average woman would basically not be able to do it. I'm a big, tall guy and i was like both arms muscles flexed to the max to turn sharp enough quickly enough one of the times that thing conked out in traffic as someone was pulling out in front of me.

So yea, for all the people saying "meh, i tried it and it wasn't that hard" it might be way easier on your car because of the steering ratio and geometry, and it also just varies from car to car. In my old 80s tercel(which was basically a go cart with a roof) it made pretty much no difference to the point that if it broke i wouldn't have ever fixed it. In that van, and my gigantic plymouth you better be bluto. And i've been in some smaller cars like my dads subaru that are shockingly hard to turn for their size and seeming heft.

The impossible-to-turn thing just doesn't seem that implausible here to me. Especially since with a direct drive EPS system like i know minis have, and it wouldn't surprise me if these cars had, you're cranking the motor and freaking gear train of that motor... which has added resistance in the motor from being connected to an electrical system.

The thought of dealing with that scares the shit out of me. And i've driven a 4000+lb 60s land yacht car at speed when the engine cut and the power steering died(because my ignition switch was fucked! hah! oh god why did i think it was a good idea to fix that with lemon juice). This, to me, seems like it would be worse.

bitteroldman: I think it was a similar situation with the Toyotas/Lexus' (Lexi?) stuck accelerator issue back in 2009.
It seems that in some cases, the drivers were not able to shift to Neutral for some reason - I think on the Lexus the shifter was controlled by an electronic switch.


Ugh i can't find the link, but wasn't this total horseshit? it was some quack in florida who desperately wanted attention, and was convinced shifting to neutral would "flip the car".

I remember, at the time, a bunch of info coming out about that guys sordid past with making up bullshit... but it all just got sucked down the drain because everyone was going "OMG UNINTENDED ACCELERATION".

Other cases of it might have been real, but that original one stunk to high heaven.
posted by emptythought at 5:03 AM on October 31, 2014


That whole unintended acceleration thing makes me very sad for the state of mind of people in a panic. It also makes me really angry for the perpetuated myth around it. There is not one single car on the market (potentially, although not necessarily, including supercars) where the brakes won't overpower the engine if you stand on them hard enough. There is not one car that I know of that you cannot just select neutral. Not one car on the market where you cannot just turn the ignition off at speed. Every single one of those incidents was entirely preventable by the driver. Every. Single. One.

Even the tiny proportion where the pedal genuinely stuck open (most were not). Assuming the throttle stuck on them, then standing on the brake pedal as hard as you can WILL slow the car. It will also give you time to put the car in neutral, at which time the car will slow perfectly fine. You can also turn the engine off. It is impossible that the brakes 'burned out' before the car was slowed or stopped unless you are pressing it ineffectually for a prolonged period - far longer than the period required to slow the car if you have maximum braking force applied. I've tried it in a couple of cars (BMW 330i and Jaguar XJ8 and several rental type cars) and, at 40mph or so and more, just stomped on both pedals and the car.... stops. Not immediately, but it immediately slows and does stop. It's noisy and smells a bit and feels horrible, but it stops if you jam your foot on the brake pedal as hard as you can. But much as a lot of ABS systems are never even triggered in a lot of accidents, a scary amount of the population just ... don't press the pedal hard enough when it matters because it's so out of whack with how they normally use it. Which is why the technology was created and brought to market to detect a sudden brake application which produces an automatic full brake force in some models. It is enough of an issue that manufacturers developed a system to counteract it. Just in the same way as a sudden increase in steering effort becomes 'locked steering' and putting the car in neutral will 'cause it to flip' - things that are outside their normal use case just blows people's heads up in a panic situation and they become paralysed or unable to think logically. Which is why these unintended acceleration issues (and possibly so many accidents form the ignition key thing) actually happen. Human error in an 'out of normal operation' incident.

I think maybe 5% of them actually had a physically stuck throttle mechanism and the others were the mat over the throttle pedal (in Toyota's case). Scary and dangerous, but only panic and mental blocks caused the vast majority of accidents, not the issue itself. Most of the cars involved weren't even that powerful(so weren't accelerating all that hard) and I'd wager that the a lot were people that went for the brake, hit the throttle instead and froze - utterly convinced they had the brake pedal to the floor and the car was still not slowing ergo they were utterly out of control. They pressed harder rather than moved their foot to check they were on the right pedal. There should really be only about two-three seconds at the most where the car was accelerating against intended control before the issue was remedied - full brake, neutral or engine off. Anyone that was driving within 3 seconds of continued speed or accel actually losing control was likely driving too close to the car in front or went straight into panic mode. 3 seconds is enough time to process 'the car is still accelerating' if you aren't panicking, but a large amount of people went straight past 'what's happening' and their brain said 'I'm completely out of control'. In which case I'd expect some relatively minor accidents with people driving into the back of people and possibly being hurt, but no-one killed. These 'twenty minute wild, out of control rides' are 100% mental failure on the driver's case. Anything more than 3-5 (being generous) seconds is human error/mental failure/panic.

This happens to a relatively small proportion of the population, which is why a lot of people here are not really grasping the possibility of total panic and 'steer harder' and 'brake harder' or 'turn the car off' are perfectly sensible and instantaneous conclusions that it MUST be a bigger issue in some way. But so many people are driving that this proportion of people have a lot of cars between them. The extremely low bar for the driving test in the US and the extremely low challenge of driving in the US with their huge roads and massive road space (funny how these only happen for the most part in the US isn't it?) means that people just don't give driving enough thought. When it is 10% challenging most of the time, it is too much of a mental leap to give it 80-100% of their mental focus before they go to panic mode. I know of just two widely reported incidents in the UK at the same time - both of which were old, frail people with very powerful automatic cars that turned out to have had 'pedal confusion'.

A large number of the driving public panic in the US. Not most, not even a majority, but that very small minority is still a huge number of people, which is why so many of these cases occur. I feel it is a direct result of making driving too easy and the bar to a driving license too low in the US (driving around cones in a parking lot in Georgia? Seriously?), producing people not really suited to dealing with something as dangerous as a car doing something unexpected. This is why such things as the unintended accel and this ignition key has such wide ranging consequences.
posted by Brockles at 6:14 AM on October 31, 2014 [5 favorites]


the others were the mat over the throttle pedal (in Toyota's case).

We had three Toyotas through through that time period. At least two of them had this problem (a Matrix and a Yaris). I had it happen to me once on the road. It was rather scary. Dealing with a fight or flight adrenaline rush and a car that's not behaving isn't fantastic when in traffic. I was able to unstick the pedal without a major incident, but it was pretty unsettling.

I have a fair bit of sympathy for the folks who did panic entirely. It wasn't immediately obvious what was happening. That Toyota pedal design was pretty stupid. A lot of their customers lived in places with snow and salted roads, where thick winter mats are very common.
posted by bonehead at 10:56 AM on October 31, 2014 [1 favorite]


The extremely low bar for the driving test in the US and the extremely low challenge of driving in the US with their huge roads and massive road space

I know a couple of people with their private pilot's license, and my mind keeps returning to that point in their training where their instructor unexpectedly turns off the engine during flight and the student must go through the procedure of scouting a suitable ad-hoc landing area on the ground. Nothing even remotely like this is practiced in the US driver's licensing process. People are simply not introduced to handling unexpected situations or have any idea how their car behaves at the limit. The first time I ever activated ABS was in competition on a closed course. Ditto for spinning out. How many people find the interest or motivation to explore that on their own?
posted by indubitable at 3:05 PM on October 31, 2014 [2 favorites]


I guess there's an advantage to growing up driving beat up old cars with bald tires in a rural area, after all. You get exposed to most of the unexpected things a car can do pretty early on. About the only thing I feel like I missed was literally having a wheel fall off. That and actually crashing a car, but that was dumb luck on my part more than skill. Plenty of my friends ended up getting to know various trees and ditches more intimately than they would have preferred.

The first time I spun a car I froze and did nothing to stop it. It was, again, pure dumb luck I didn't end up sliding off into the creek. The second time I had some idea what to expect and didn't actually spin the car. It took a while for me to learn that maybe I ought to not drive past the limits of my tires so I wouldn't have to correct, but I was able to do it in relatively low stakes situations.

Seems like those who grow up in cities ought to get that experience in driver's ed, which probably should be mandatory. (and paid for by the government since cars are almost a necessity even for poor people in most of the country and they definitely can't afford private driver's ed classes)
posted by wierdo at 8:57 AM on November 1, 2014


> There is not one single car on the market (potentially, although not necessarily, including supercars) where the brakes won't overpower the engine if you stand on them hard enough.

One claim is that the brake booster in the Camry & Prius was vacuum powered & manifold vacuum would be low at wide-open-throttle, so "stand on them hard enough" would be quite hard.

> There is not one car that I know of that you cannot just select neutral.

The Prius transmission is pretty weird & the gear "selector" mostly non-mechanical. Supposedly, you can't get a Prius with a dead engine & battery out of Park without getting underneath to reach the pawl.

> Not one car on the market where you cannot just turn the ignition off at speed.

Start button Toyotas required a several-second press for off to avoid inadvertent shut-off. 3 seconds is a long time at a panicked 100+ MPH.

Sounds to me like combination of adverse factors*: 1. Different causes for unintended acceleration lead to prolonging of the period before the floor mats, sticky pedal and bad firmware were all fixed. 2. Possibly harder to brake. 3. possibly impossible to engage neutral. 4. turn off protocol is novel. 5. humans are fallible: a. turning off a moving car is never practiced b. pressing the brakes (hard) while the engine is screaming is never done, c. panic due to seemingly out-of-control car.

If human factors engineering is prioritized and accident response by manufacturers treated more seriously and drivers are trained better, people would have a better chance.

The GM ignition switch problem seems simpler, but one thing that got me was that supposedly without the engine running, "While Brooke’s lap belt glued her waist to the seat, her shoulder harness went slack the instant the engine shut off." What!?

*Air France 447 is my "favorite" case of this. Clogged pitot tubes, dodgy UI and human panic all "conspired" to fly an operable plane all the way into the ocean.
posted by morganw at 11:00 AM on November 1, 2014 [1 favorite]


so "stand on them hard enough" would be quite hard.

There is more than enough vacuum at WOT to stop the car with. I think it is far more to do with 'but that's as hard as I've ever needed to press them' limitations than genuine limitations of boost assistance. It's only at WOT *and* max RPM that you have minimum vacuum which, again, can only happen once they have got up to speed. In a 'not very fast' car like these ones in the examples there is no real chance of loss of brake boost assistance for 15-30 seconds after the incident started (depending on initial speed compared to Vmax). Either way, someone was not pressing the brake very hard for many times more than the time needed to stop the car before brake boost vacuum ran out.

3 seconds is a long time at a panicked 100+ MPH.

None of these incidents started at 100+mph and the kind of cars involved took many times longer than that to get to that kind of speed from their initial starting speed. You'd be at no more than 20mph faster than you started off at (even assuming the throttle 'stuck' at 100%) and still would have had time to do that three seconds press a couple of times over. It's not like the cars instantly went to VMax.

The Prius transmission is pretty weird & the gear "selector" mostly non-mechanical.

There was nothing stopping the cars selecting neutral, though. I've driven a fair few Prius' (and moved more than a few with no electrical power, as it happens) and there is nothing weird to stop you putting it in neutral. The Park detente in a Prius is a physical mechanical lock that is held out by power. The transmission is a hydraulic pressure system that is kept engaged by continued pressure via a solenoid. Pressure drops? Transmission is in neutral - this is what would happen in the event of power shut off or the lever being moved out of drive. There is nothing stopping neutral being selected (other than your option 5!) in any car.

While Brooke’s lap belt glued her waist to the seat, her shoulder harness went slack the instant the engine shut off.

That sounds like rubbish to me, unless they are somehow meaning through some clumsy prose that the seat belt tensioning mechanism (that triggers in the event of an accident and the airbags triggering) that pulls the belt tight from the shoulder didn't activate? I mean, nothing is *holding* the seat belt tight except a spring on the reel, and usually a clutch mechanism prevents it loosening quickly (that thing that stops you getting your wallet out at the drive through...) but neither are dependent on power at all. If they meant that the 'slack' would have been taken up if she had crashed with the ignition in, then sure. I mean it would have tightened the belt as the airbags fired. But it wasn't slack as such, because of the existing mechanical systems and power has nothing to do with it other than the tensioner during the accident (on some cars).
posted by Brockles at 3:11 PM on November 1, 2014 [2 favorites]


« Older You say H-O-R-S-E, I say S-K-A-T-E   |   Tim Cook - I'm Proud to be Gay Newer »


This thread has been archived and is closed to new comments