Hacker News Reader

Comment by hervature | Hacker News Reader

Completely agree that what is lacking here is the regulations. I have tried to argue (often unsuccessfully) that VW did not commit fraud, that the numbers reported by the tests were 100% accurate. The problem being that the way the car is tested is not even remotely similar to how people drive. The clear fix to me is to simply give normal people free leases for a short period of time and generate something similar to http://www.fuelly.com/ . The only way to cheat that is to convince people to drive more efficiently, seems like a win.

View on HN

How is embedding software that changes the way the engine behaves during tests versus during actual usage not fraud?

The numbers may have been accurate _for an engine with the 'limit emissions' mode enabled_ but it's still fraudulent if you do not make that explicitly clear.

If you use the car the way tests use the car, the performance will be the same. The trick is you can't complete your daily driving needs by driving the way the test does.

The intent was to cheat the tests by secretly tailoring the behavior to the tests, even though they knew that real-world driving would be different and that the tests are supposed to indicate real-world emissions.

Intent matters. You can’t set out to cheat a test, succeed, and then say that it’s legal because you succeeded, any more than you can steal from a store by hiding an item under your coat and then say that it’s legal because you disguised it well enough to fool the guard.

The only way to drive in a way that would have replicated the results of the VAG tests would be to not touch the steering wheel. At all.

For VAG this wasn't a case of the test not being quite the same as regular driving, it was VAG deliberately detecting a tests and altering the way the engine worked to avoid failing.

If you turned the steering wheel the test would fail, you can't drive without turning the steering wheel.

It would also turn off if the back wheels were spinning.

That's the whole point: you cannot use the car the same way as during the tests, even if you exactly match speed, acceleration, distance, friction, temperature and any other variable. The software that limits emissions is disabled during "normal" operation, ie. driving.

I will attempt to make an analogy that tries to shine light on the angle I'm coming from. When we interview for software engineers, many companies place much of the emphasis on algorithms as opposed to other hard-to-test skills. Now, you have a candidate that aces any algorithm question put in front of them yet completely sucks at software engineering. Did that person commit fraud? It seems, to me, that the test is what is broken.

We all clearly understand the angle you're coming from, but you haven't been able to make the case it isn't fraud. Because it is textbook fraud.

If you think the test is bullshit, then you make that case separately. You don't lie about how you're passing it and then when caught say 'but it's the test's fault!'.

Emissions tests are different from fuel efficiency tests, and in fact are often at odds with each other. A more efficient car doesn’t necessarily have better emissions, and there’s no easy way for a typical driver to measure their emissions.

NOx emissions specifically are a problem because the engines are run very efficiently at high burning temperatures with complete combustion.

Thank you for the extra insight, I may have had the wrong conclusion but I think the logic still follows. From this system I outlined, the manufacturers would be forced to alter designs to get the desired emissions during normal driving conditions. This is the whole spirit of the regulation right?

My point is that the system you outline doesn’t work, because ordinary drivers can’t check their own emissions. It would be fine for fuel consumption since that’s easily measured, but emissions requires special equipment and careful testing.

The VW stuff wasn't simply about fuel efficiency it was about emissions output. I think this was the video where a guy basically reverse engineered the system. (Starts around 40m in) https://media.ccc.de/v/32c3-7331-the_exhaust_emissions_scand...

I'd encourage you to watch that, and still see if you say that it wasn't fraud.

VM gamed the test in order to produce results that were not achievable by consumers, with the intent to deceive people.

I will watch that video later and get back to you. I understand what they did. I'm not saying that what they did was right, simply that it wasn't fraud. My point being that if the test is already deceptive for true vehicle emissions (laboratory settings, very particular configurations, etc.) then the test itself is the problem. If we flip the script and say that a car fails the emissions testing but is in reality the cleanest car, is it not "fraudulent" to label this car as dirty?

Intent matters. Your example of flipping the script would only be comparable if the manufacturer engineered the car to fail the test.

VW built their car to produce a specific test result to get buyers to buy the car. It fits the description at the link below all too perfectly.

https://legal-dictionary.thefreedictionary.com/Fraud

Without knowing the general driving population's driving habits, it's difficult to estimate how effective driving will improve fuel efficiency and/or reduce pollution. Sure there are known bad driving habits like sudden accerlation/sudden stops will increase fuel consumption, but what about road sign design, speed limits, traffic light designs and Electronic Toll Collection systems? Those are all factors of fuel efficiency.

Also, the government regulation on emission is going to get tighter and tighter due to incentives to push for EV adoption. So arguing emission test method is a moot point.

Regulations are obeyed and enforced by humans. There’s a general expectation that common sense is applied, which is hard to define but “you know it when you see it”.

“The engine will fundamentally perform differently when you’re testing it because it will detect the test” is not common sense and is not a reasonable interpretation of how to comply with the rules.

I agree that a better method would be to attack sensors to a random (and ideally, concealed from OEM) sample of cars and charge automakers per unit of detected emissions. That would force them to optimize for real world driving, which is much harder to game.

Reply on news.ycombinator.com