Mercedes Benz has worked out who to sacrifice in an accident. And: arguments for the sake of 23 Oct 16

Mercedes Benz has worked out who to sacrifice in an accident. And: arguments for the sake of 23 Oct 16

http://www.smh DOT

Not you, the driver or passengers

Ross1 | 14. Oktober 2016

Home Tech News
OCTOBER 14 2016 - 4:11PM
Mercedes-Benz driverless cars will hit pedestrians rather than endanger occupants

Josh Dye
Josh Dye

25 reading now



Future Mercedes-Benz driverless cars will be programmed to prioritise driver and passenger safety at the expense of pedestrians.

Comments by a senior Mercedes-Benz safety manager have again raised ethical questions about the future of driverless car technology.

MORE TECHNOLOGY VIDEOSPrevious slideNext slide
Ride in Mercedes' F 015 Driverless Car
Video duration
Rivals ready to pounce on Samsung's mobile tu ...
Video duration
Yahoo secretly scans emails for US
Video duration
How does a VPN work?
Netflix leads streaming video revolution (Video Thumbnail)
Video duration
Netflix, Stan, Presto and the video revolutio ...
Video duration
Man attacks French Apple store with steel bal ...
Video duration
China offers a glimpse into the future with a ... MORE VIDEOS
Ride in Mercedes' F 015 Driverless Car
A sneak peek inside the Mercedes F 015 concept car, the company's vision of a driverless future.
Debate was sparked again when Christoph von Hugo, manager of driver assistance systems and active safety, said at the Paris Motor Show: "If you know you can save at least one person, at least save that one. Save the one in the car.

"You could sacrifice the car. You could, but then the people you've saved initially, you don't know what happens to them after that in situations that are often very complex, so you save the ones you know you can save."

The Mercedes-Benz F015 driverless car is programmed to avoid collisions, but not at risk to the occupants.
The Mercedes-Benz F015 driverless car is programmed to avoid collisions, but not at risk to the occupants. Photo: Supplied
In other words, if a collision with a pedestrian was inevitable, the car would not deviate off the road and put the occupants at risk to avoid hitting the pedestrian.

Driverless cars have long raised ethical dilemmas because they take out the human element.


While the technology is programmed to save lives and improve road safety by removing human error, every stumble is scrutinised.

Earlier this year, Tesla was investigated after a driver was killed when his car, set to "Autopilot" mode, failed to detect a tractor pulling out in front. However, Tesla said it was the first known fatality in more than 200 million kilometres of Autopilot.

A glimpse into the future? The interior of the driverless Mercedes-Benz F015 concept car.
A glimpse into the future? The interior of the driverless Mercedes-Benz F015 concept car. Photo: Supplied
Mercedes-Benz is now developing completely driverless cars, using what it calls Level 4 and Level 5 technology.

The company's Australian-based senior manager for communications, David McCarthy, said there was plenty of time to test and perfect the technology.

A Google driverless car.
A Google driverless car. Photo: Supplied
"Driverless cars where the driver has no input - that's quite a long way off," he said. "We're talking five to 10 years."

But he said the technology would drastically reduce the number of collisions in the first instance.

A driverless car being trialled in the United States.
A driverless car being trialled in the United States. Photo: Supplied
He said in cases where a collision was unavoidable, driverless technology would ensure a better outcome.

"At worst, it will reduce the speed of impact. It is going to focus number one on avoiding that collision."

Mercedes-Benz is now trialling Drive Pilot, the company's semi-autonomous car technology, which is about Level 2, Mr McCarthy said.

"The car has 360-degree vision, using a combination of radar and cameras to get a picture of the environment it's operating in."

He discussed the hypothetical situation where a child runs onto the road.

"Your immediate reaction is to apply brakes and steering. Our systems do the same but take into account - are there cars in other lanes? If you take evasive action, you may well hit that car, which might push you in a different direction towards the pedestrian."

He said the technology could assess in a split second what would be the best possible way to avoid that collision.

But he said that, for now, the driver ultimately remains in control.

"The driver always has the ability to counter an action the car makes."

The feedback on the technology to date has been positive, Mr McCarthy said.

"Once people understand and experience the technology, they embrace it.

"Regulators and insurance companies are very interested in it because it will reduce incidents. It will absolutely make for safer roads," he said.

It is said that human error contributes to about 90 per cent of road accidents.

Earlier this year, US research showed that people did not think driverless cars should sacrifice a passenger to save only one pedestrian, but their approval rose with the number of lives that could be spared.

But while participants approved overall of cars that sacrificed passengers for the lives of 10 others, they were less likely to buy the cars themselves.

SeaP90D | 14. Oktober 2016

Seems fair to me, Tesla should do the same, except maybe give special dispensation if it detects a Tesla key fob in the pedestrian's pocket.

Ross1 | 14. Oktober 2016

Sorry about the messy article.
I thought it wasnt going to copy paste, and then it spewed out the lot, too late, I didnt edit it down. | 14. Oktober 2016

Let the driver decide in advance. Easy to have an option to specify who should be saved in a no-win situation:
- Preference for driver and passenger
- Preference for other vehicles
- Preference for objects and people in roadway

This leaves the manufacturer out of the decision.

kevin | 14. Oktober 2016

Ross, you can edit it now. Click Edit.

Linus | 14. Oktober 2016


You forgot one:

- Preference for kittens and puppies

Ross1 | 15. Oktober 2016

Yes, No, I can only edit the first OP, not the second post.
Thanks :)

Ross1 | 15. Oktober 2016

Yes, No, I can only edit the first OP, not the second post.
Thanks :)

Ross1 | 15. Oktober 2016

Who to sacrifice?
The Donald comes to mind

carlk | 15. Oktober 2016

Mercedes could sell special electronic tag for pedestrians to wear. The device can come at different level with different prices too. The car will make an effort to avoid hitting the pedestrian when the tag is detected. The higher priced tag it sees the greater effort it will be. The car will kill everyone in the car before hitting you if the $250K top model tag is detected.

Tropopause | 15. Oktober 2016

I think this question is pointless and only serves to distract from the advancement of autonomous cars. Of all the things autonomous cars can do, why are people asking about this do-or-die scenario? What did we do all these years of manually driven cars when faced with this dilemma? When is a car barreling out of control towards a crowd of people? Almost never! Because there is order in our world- cars drive on the streets in a known pattern and people walk on the sidewalks. Autonomous cars will be even more regimented about obeying traffic laws so the likelihood of this unlikely scenario is even less.

I think reporters just like to stir up controversy and scare the public with extreme scenarios that have no solutions. Maybe it's the insurance companies that are coming up with these scenarios to protect their market by prolonging the advancement of self-driving technology. ;)

SbMD | 15. Oktober 2016

@tropopause and others calling this an alarmist or invalid point are correct. This is a version of the "trolley dilemma" in ethics. This "choice" is not a choice at all for the reasons above, including:

1) human drivers cannot make an "ethical" decision in the same extremely unlikely scenario, hence this is not an autonomous car issue.
2) this assumes that there is only one remaining choice in a given "trolley" scenario: someone has to die. Not true except in the most extreme and unlikely case, in which case a human cannot make the right decision either.
3) the decision to be made is not to codify a way of making the most ethical decision but rather to avoid coding an unethical one: specifically, you do not instruct a person or car to take a life, but to take every measure to avoid taking any life. | 15. Oktober 2016

I have perhaps a more cynical view. It's all about the lawyers and who will get sued when the accident occurs.

- Is it the jaywalking pedestrian playing Pokemon Go?
- Is it the game manufacturer?
- Is it the driver who was not looking at the road?
- Is it the vehicle manufacturer for offering autonomous driving abilities?
- Is it the coder of the autonomous driving software?
- Is the insurance company that gave a discount for lower accident rating of autonomous driving?
- Is it the city for not putting up guardrails to prevent pedestrians from jaywalking?
- Is it the police for not ticketing jaywalkers at that location?

I'm sure there are more :( So the question is important for everyone that doesn't want to get sued and to access who is the correct party to blame for such an accident.

Tiebreaker | 15. Oktober 2016

Paging Isaac Asimov.

SbMD | 15. Oktober 2016

@TT - yes, a good point. The legal underpinnings of autonomous driving still need to evolve and it should be better defined ahead of the first Level 4+ autonomous vehicles to hit the road.

For the purposes of discussion, and to the OP's point, there are already guiding principles for the trolley principle. Simply put, taking a life to save one's own (when not in self defense) is not a legally defensible position.

martin | 15. Oktober 2016

The basic problem is that an autonomous car has to handle irresponsible other drivers, crazy cyclists, little children, pedestrians and animals and not the other wa round.
We still have no autonomously flying aircraft that operate in a for more structured and supervised environment than our cars do and have a far more sophisticated "flight assist system" (normally called "autopilot"), why should cars be first to drive autonomously???
I just don't see autonomous cars on the horizon...

martin | 15. Oktober 2016

And I didn't even mention road and wheather conditions...

jvs11560 | 16. Oktober 2016

Insurance institute numbers are available for the Model Years 2012 - 2014. In the Large Luxury class, Tesla scored the worst on collision and bodily injury. The Model S was also much worse than average for Comprehensive too. Fatality-free models over the four model years were the Audi A4 four-wheel-drive, Honda Odyssey minivan and Subaru Legacy sedan. These are older model than the MS and have none of the MS safety features. Since I know of at least (9) fatalities in Model S Tesla's, how is the claim that this is the safest car still valid?

SbMD | 16. Oktober 2016

@jvs - these are claims data in terms of dollars and not normalized for cost, nor have the data to look at the cause or outcomes. Your assertions that safety is compromised in the MS cannot be backed by what you have posted.

Where is this fatality data you refrence?, exactly? Can you provide something more than anecdotal?

The NHTSA has more robust data on make, model, and year of cars including accidents, fatalities, and causes coded in their databases. This has been explored in past threads, and the data were overwhelmingly favorable for the MS. I haven't looked at that data in probably 6 months, but it was convincing.

jvs11560 | 16. Oktober 2016

@SbMD If you look at the iihs data, you'll see that the bodily injury number was the worst of all Full size luxury cars. Whether it's dollars, or injuries, it accounts for all luxury cars, using a standard set of data for all of the cars tested.

The fatality data is from news reports. 4 were in California, with the most recent, a young girl in the back seat on the 405. My post was just to point out that the Audi A4 four-wheel-drive, Honda Odyssey minivan and Subaru Legacy sedan had 4 model years (Each) with ZERO fatalities. Please bear in mind that there are many more of these vehicles sold, than Tesla's. Last year there were 127,000 Honda Odyssey's sold. Since each of these vehicles had ZERO fatalities, and score much better on all of the insurance loss data cahrts, how can Tesla make the claim that their cars are safer? It's a legitimate question, backed by real world data/facts.

negarholger | 16. Oktober 2016

Going back to the original question... how many car will a car maker sell if he states in an inevitable situation we will sacrifice you instead of the others? Answer : close to zero. There is no other "marketing" option then save the occupants of the car.

jvs11560 | 16. Oktober 2016

@Kleist If you read the article, it clearly states that if a collision with a pedestrian is unavoidable, the car will stay on the road to protect the occupants. Since a human being has peripheral vision, and a thought process, they might do a better job driving in this situation. Then, you could save both the pedestrian and the occupants. Win, Win! If one innocent pedestrian is killed by some AI in an automobile, you can rest assured that the company making that vehicle will pay out millions to the family of the deceased.

SbMD | 17. Oktober 2016

@jvs - thanks for the reply. The data you posted are based on cost of claims, not numbers, circumstances, nor fatalities, etc. Your interpretation is incorrect.

As for the fatality info, your sourcing of this is not nearly as robust as looking at NHTSA data. Not doubting that there are auto fatalities, but you don't know how many fatalities per model car and the circumstances, both of which matter when assessing safety of different vehicles.

If you are going to make the claim that you are making, you need to have the right information. You do not.

jvs11560 | 17. Oktober 2016

@Sbmd To make it simple. The Honda, Audi, and Subaru models that I listed, have had ZERO fatalities for (4) years. Which means that each of those vehicles had no one expire driving all of those vehicles for millions of miles for four consecutive years. This is data that I read directly from the iihs web site. The nine deaths are also public information. If you do your own independent research, you can confirm all.

jordanrichard | 17. Oktober 2016

Just because there were no deaths in those cars, that doesn't mean they are safer. Everything is relative and needs to be put in perspective. When I was in my early 20's I had a Camaro and naturally bemoaned the cost of insurance. My father who was in the insurance business, albeit life insurance, said that if the majority of Camaro owners were older and more conservative drivers, the insurance rates for Camaros would be less.

My point is, you have to put any data point into proper context to get the real picture.

Run4Waffles | 17. Oktober 2016

@tedirelan - I would rather see someone in a Model S, THE safest car ever tested, than in a "Honda, Audi, and Subaru models that" you listed if they were to be in an accident.

Common sense alone tells you that something is wrong with the presentation of the data.

You can do the work to see the horrific accidents that have taken place around the world where people were able to walk away with their lives. You can also look at the Tesla website for stories and the owners, spouses and family thanking Tesla for saving their loved ones. And the owners staying that they were buying another Tesla.

Show me another vehicle that was airborne for 82' and the occupants were able to get out of the vehicle under their own power.

Sorry the thread was high jacked.

KubASS | 17. Oktober 2016

@jvs11560 - I don't think you're interpreting this data correctly.

"Bodily injury" Pays for injuries that you cause to occupants of another vehicle.
"Personal injury" and "Medical payments" refer to injuries of the driver and its passengers and are the best of all Full size luxury cars (and better than for the cars that you had said were safer that the Model S).
"Comprehensive" and "Collisions" unfortunately will be very high for the Model S - it's just a very expensive car to fix after an accident or insure against theft or non-crash-related damage.

jvs11560 | 17. Oktober 2016

@jordanrichard & Run4Waffles T

hen how do you explain the fact that bodily injury claims for all large luxury 4 door cars are the highest for Tesla? Actually, if you look up the following site:

If you click on "worst", you'll see that Tesla is on the worst list, of all category of cars, right next to the Nissan GT-R. Regarding the comment about, Camaros and younger drivers, How many kids can afford a $110K car? (Average price of a Tesla sold).

Google "Kaleb Whitby" and see how he managed to get away without a scratch. Sometimes, you just get lucky.

BTW, those kids in Germany were seriously injured. | 17. Oktober 2016

@jvs - just curious - what car do you drive? Seems like safety is important to you.

The two year old IIHS data (there is nothing newer) covers cars made in 2012 to early 2014. It doesn't say anything about fatalities that I could find. In fact the Personal Injury rating for Tesla is Green - Substantially better than others. Those that are white without a number indicates insufficient data, not that there are no claims or injuries. This occurs more frequently with models that have low sales volumes.

Here's the actual data:

rxlawdude | 17. Oktober 2016

Is jvs a short tool, competitor, or merely a troll? Argumentative and negative.
Results for collision, property damage liability and comprehensive represent overall losses, which reflect both the frequency of claims and the average loss payment per claim.

jvs claims about human safety can best be described as poppycock, given the nonsequitur data he proffers

Run4Waffles | 17. Oktober 2016

I didn't state that they weren't injured. They went 82 feet through the air. Seriously, that's your response to such an accident. WOW.


Red Sage ca us | 17. Oktober 2016

We lose more puppies that way.

jvs11560 | 17. Oktober 2016

@TeslaTap My daily driver is a 2016 Chevrolet Suburban. I also own a 2016 Escalade ESV, and 4 other cars including a BMW X6 50xdrive. The Suburban just hit it's one year anniversary and it has 33,000 miles. For the record, I had a CTS-V and wanted something different. I went to a local mall and drove the MS. I decided now wasn't the time. Every Tesla owner I know has at least one other car. Most love it! Even though I get beat up her, I may well buy an MS in the future.

Years ago, I was in a very bad head on collision. I was driving a 280ZX. I woke up in the hospital with my head spilt open, my front tooth chipped, knees bleeding, and a broken noise. I also witnessed a bad accident where the children in the backseat of a car were crushed. I am a commercial pilot, but I do not fly for a living. I have a good grasp on physics and I would like everyone to have fun, and live to tell about it.

Magicstix | 17. Oktober 2016

It only makes sense that the car's highest priority should be its own occupants. Sure, it can make an effort to protect people outside of the car when it can, but at the end of the day the people who are inside the car and have entrusted themselves to its safety systems are the ones who should get the most protection.

The only truly "fair" way to do it would be to have the car pick a random number and decide who dies, but it just makes far more sense for the car to protect its master at all costs when it has to make that kind of decision. | 17. Oktober 2016

@jvs nothing like a bad accident to make you consider safety in every purchase. A healthy attitude.

I didn't notice it before, but in the insurance chart for the Luxury car category (where they put the Model S), The Model S is the best rated of any vehicle in both the Personal Injury and Medical Payments categories. Backs up my thinking that the Model S is a very safe car. It is expensive if you have a collision, but I'd rather pay for damage than be personally damaged.

jvs11560 | 17. Oktober 2016

I read the data wrong.

SbMD | 17. Oktober 2016

@jvs - Your data and interpretation are incorrect. You wrote:

"The fatality data is from news reports. 4 were in California, with the most recent, a young girl in the back seat on the 405. My post was just to point out that the Audi A4 four-wheel-drive, Honda Odyssey minivan and Subaru Legacy sedan had 4 model years (Each) with ZERO fatalities. "

This struck me as very unusual. I looked through the NHTSA FARS data started with the Audi A4. Here they detail many fatalities in the Audi A4 in just one calendar year in the FARS data. This dataset is far more robust than IIHS. Your claim was so off in relation to NHTSA data that I stopped there. There was no point.

Your interpretation of the IIHS data are also incorrect, and looking through their site I cannot find any information there to back your claims about safety.

You should be more careful with your data sourcing and interpretation when making your claims.

SbMD | 17. Oktober 2016

Furthermore, the NHTSA data shows that MS fatalities are so few, it is in stark contrast to the models you list above and others.

The example you give of the child who died, if I recall, was the MS being struck by a drunk driver speeding in a large SUV which rear-ended the car, plowing so deep into the car that no passenger car would have protected a passenger in such an event.

Again, your claims are far off the mark.

SbMD | 17. Oktober 2016 | 17. Oktober 2016

But what if the pedestrian owns a Mercedes?

bubslug67 | 18. Oktober 2016


I've looked into the NHTSA FARS database on fatal accident data. It is missing a number of driver fatalities in the Tesla Model S, namely the Malibu Canyon CA, Jenner Hwy 1 CA, La Brea Ave CA, and Houston TX events. I'm not sure why as my understanding of the database is that it is to analyze cause and effect of all auto-related fatalities, be they from police chases, alcohol, excessive speed, suicide, mechanical failure or whatever. I didn't check to see if the Williston FL event was entered but I expect that won't be available until the end of 2016 in any case.

Whether you agree they should be included in a comparison of fatality rates is another issue but be aware in any case that this database is missing some Tesla fatal accidents.

AmpedRealtor | 18. Oktober 2016

This is very simple. When you, as a human driver, take evasive maneuvers in an accident scenario who are you trying to save? I would bet, 10 out of 10 times, you are trying to save yourself and your passengers. Nothing about that situation should change if there were a pedestrian in the scenario. The reason we buy ostensibly "safe" cars and give credence to safety ratings is because we want to be as well protected as possible. A pedestrian presumably has the ability to see a car coming and might be able to get out of harm's way.

I don't believe this decision should be up to the individual via a setting. Settings can change themselves, as Tesla has shown us in almost every software update. Others can change your setting without you knowing about it. And are you really in the best position to evaluate the best option? As a driver, what if you changed the setting to protect the pedestrian and sacrifice the occupants. Wouldn't you have to disclose that decision to anyone who rides in your car? What if they feel differently? What if there is an accident where the car followed its user setting to sacrifice occupants instead of the pedestrian. Your passenger is critically injured and paralyzed for life. Guess who gets sued, and guess who is ultimately responsible? Is that a responsibility you, as the driver, are willing to assume?

I don't think any passenger anywhere on the planet would knowingly ride in a vehicle that would sacrifice the occupants to save a pedestrian. That labels every car with such logic as "unsafe" by default.

AmpedRealtor | 18. Oktober 2016

Then there is the question of how such a setting would affect your insurance rates. Would one setting get lower rates than another, and how would the insurance company enforce such a setting? This gets complicated very quickly.

SbMD | 18. Oktober 2016

@bubslug67 - yes, there is a lag between data acquisition and availability in the database. I believe it is 9 months. Also, you have to dig into the reports and look at the undefined categories as well because the fatalities with Tesla are low and they are a relatively new brand. Even doing that turned up a few fatalities only.

Either way, @jvs' assertions are unsupported.

Dig deeper, and looking at accidents without fatalities will be interesting; if there is enough detail recorded to see if the safety features of the MS and MX reduce injury severity as well as fatality.

And with respect to IIHS, they may not have tested the MS, but this is interesting and supports the impressive build of the car:

As for the OP's point, I've already posted earlier about it. | 18. Oktober 2016

@AR - I was sort of joking about having a setting although it would remove the manufacturer from some of the responsibility. Very doubtful any setting like I proposed would ever be offered.

Ross1 | 18. Oktober 2016

When it comes down to the wire, it will be about what the lawyers say to set.

But no reason there could not be settings for various owner or driver own ethics.

@Amped: Agree. Why would you get in a car or bus which will offer you up as a sacrifice?

What policy do the autonomous buses have, which already exist?

Ross1 | 18. Oktober 2016

If you were not too happy with your significant other you could tweak the settings remotely with your iPhone. Another first world problem. Perhaps there should not be variable settings...

bubslug67 | 18. Oktober 2016


"Even doing that turned up a few fatalities only."

I can't see "Tesla" as a make listed the NHTSA FARS query system. I had downloaded the raw data files and looked into them using a spreadsheet, so I know the codes for the accidents which involve Teslas because I looked into file named "VINDecode.dbf" . I note in the "vehicle.dbf" file Tesla vehicles were coded as Make = 29 (Other Domestic) or Make = 98 (Other Make), or Make = 69 (other import) in the "vehicle.dbf" file. How are you determining which accidents have Teslas where the fatality was an occupant of the Tesla?

jvs11560 | 18. Oktober 2016

The fact remains that there are vehicles on the road that have had ZERO fatalities for (4) years running. They are subject to the same conditions (Drunk drivers, Careless drivers, and inattentive drivers) as any other car. It is also relevant that these cars (Audi, Subaru, and Honda) are sold in much higher quantities than the MS. Can anyone dispute that?

bubslug67 | 18. Oktober 2016


"And with respect to IIHS, they may not have tested the MS, but this is interesting and supports the impressive build of the car:"

I don't know if you're aware of the European NCAP testing but the 2014 Model S was rated the worst of all "Large & Executive" Sedans for adult passenger crash worthiness for all models tested since 2012 (against 21 other cars). Not that it's actually that bad since large cars typically test better than small, but it certainly it was no better than other Euro-spec large sedans.

Also while your linked article notes the IIHS rating should have been out for the Model S by now, it's not yet in the IIHS ratings database, nor can I find a link to a news release on those test results.