It’s behind a paywall, but the WSJ columnist looks into a question touched on repeatedly in this space and connects it to the unpredictability with which juries may credit expert testimony, as an Oklahoma jury recently did in Toyota litigation:
Toyota had been vigorously fighting hundreds of complaints that its cars are prone to unintended acceleration. Now it’s moving toward a global settlement as a consequence of a single Oklahoma lawsuit that appears to establish that Toyota can’t prevail if it can’t prove a negative—that its software didn’t go haywire in some untraceable and unreplicable manner. …
The Bookout jury was apparently impressed by the testimony of software expert Michael Barr. He said a single “bit flip” (the smallest instance of data corruption) could cause uncontrolled acceleration when the driver had been using cruise control, stopped using cruise control, then resumed using cruise control to let the car accelerate back to its selected speed. …
The connection to Ms. Bookout’s crash, which didn’t involve cruise control and took place on an exit ramp? None, except Mr. Barr claimed that “software failure is consistent with the description of the [Bookout] accident” and “more likely than not” a factor.
Jenkins notes, as have others, that if some mysterious and unreplicable bug is causing Toyotas to accelerate suddenly while disabling the brakes, it seems to differentially appear in cars being driven by elderly drivers, which are greatly overrepresented in the crash statistics.
More: Kyle Graham on whether vaccine liability limits make a plausible precedent for limits on liability for driverless cars.
8 Comments
Look, car makers get sued when people crash and get hurt, even if the car makers aren’t at fault at all. So, with driverless cars, all that matters is the injury rate. If the injury rate in driverless cars is significantly better than in driven cars, driverless cars will cost the automakers less in legal fees and payouts.
This seems a dangerous precedent well beyond driverless cars. And it could be very bad for the government that uses technology for prosecutions. Imagine, “the [government] can’t prevail if it can’t prove a negative—that its [breathalyzer] software didn’t go haywire in some untraceable and unreplicable manner…”
The connection to Ms. Bookout’s crash, which didn’t involve cruise control and took place on an exit ramp? None, except Mr. Barr claimed that “software failure is consistent with the description of the [Bookout] accident” and “more likely than not” a factor.
How can the courts allow such testimony? This is tantamount to stating that because I am an “expert” anything I say must be valid. If the courts understood anything about technology they would have thrown out his testimony because he has not shown how it could happen. Mere speculation is meaningless. He might as well have stated that a Gremlin got into the software and made bad things happen. What the courts should demand is that this so called expert show that he can reproduce the crash by a “mere bit flip”. If he can’t do that then his testimony is meaningless.
Did Mr. Barr explain how check digits are used during data transmission to ensure that the bits still add up, and if any corruption is found, the data is retransmitted? Probably over the head of the judege and jury, so he ignored it.
Or was he implying that the corrupted bit is in the read-only memory of a chip, and that at a crucial point during operation of the vehicle, the corrupted bit was read incorrectly, and caused several safety protocols to fail, all at once? That no one, even through exhaustive efforts of Toyota and Nasa engineers and the promise of a generous reward, has found this “magic” corrupt bit?
As a longtime software engineer, I’m offended when I hear this sort of nonsense spewed as factual.
Mr. Barr’s testimony sounds like it should appear in a chapter of a Harry Potter book on magic spells.
With the increasing instrumentation of vehicles for diagnostic purposes, I am surprised that “black box” data is not available from which it could be determined whether the driver pressed the brake pedal or the accelerator. If it is true as the engineers seem generally to believe that that in the great majority of cases the problem was that the driver pressed the accelerator when he or she intended to brake, such data would stop most suits of this sort at an early stage.
While I would concede that SUA could occur, (and that’s what I have seen demonstrated on some older lowtech cars and believe could happen on computer controlled cars) I don’t think that the manufacturer can be held liable beyond running into someone/thing immediatley.
I.E. I give it a little gas, it goes to full throttle and it surges into the back of another car.
How can you believe that the car lost all functionality? Even for direct connections like the brakes, shift lever and key cylinder that may only be power assisted?
These are pure theory, though grounded in reality for some cases. And, that’s what gives these theories legs: people are so blinded by COULD it happen that they lose sight of if it DID happen. That these pure conjecture ideas are taken for science is terrible.
@Bill Poser
My understanding is that black box data was available for some of the cases and that all showed pedal misapplication.
I feel badly that our Japanese friends are victims of American idiocy. But their reaction to Fukushima shows they have their own limits on rationality. No body was killed or harmed by radiation. The contamination yielded radiation lodes similar to background at Denver, Colorado.
Mr. Holman’s article appears to quote Mr. Barr’s expert testimony both incorrectly and out of context and the comments don’t make the discussion here anything but more confused. Barr’s full testimony in that case is available for those who care to read:
http://www.safetyresearch.net/Library/Bookout_v_Toyota_Barr_REDACTED.pdf