Self-Driving Cars: Whom to Kill

Self-driving cars are coming, but not, I think, from Google.

Google was asking its designers to think of a vital question: Whether, in an accident, if killing someone is inevitable, to kill the driver and other occupants or a bystander or several. This helps designers design to make that dilemma less likely.

This is a version of a philosopher’s Trolley Problem, and millions of public responses have been studied.

But, legally and financially, there’s no way a car company is going to design to prefer to kill the occupants rather than bystanders. If that conundrum ever came out of the design studios, and it will through legal discovery, when occupants are killed that way, their families and estates will sue.

(The hands-on company is Waymo, owned by Alphabet Inc., which also owns Google. They share ownership but are still separate legal entities, which eases the legal liability for Google.)

Documents showing who the designers thought about killing will damage Google’s defense. Defense witnesses could be engineers who could make good impressions when explaining their thinking, but engineers often leave employers for more interesting challenges elsewhere and those who left won’t be trusted enough by Google’s lawyers to be called to testify for the company. Required documentation will be available from Google but it won’t be as thorough on the engineering or plain in language as lay juries would want. Programmers everywhere are notorious for only minimally documenting their programs and that’s probably true of engineers (and other smart people) generally, because they’re busy with the substantive work and if someone wants a repair or upgrade maybe they’ll hire an old specialist to come back because no one else knows how to do the work.

Hi-tech designers and producers tend to replace their technology often. That tends to make old versions look bad. An accident will likely have been with a car designed at least a year before and thus using old tech. It will be hard to defend old tech as sound, especially in front of juries whose own computers and coworkers’ computers inexplicably crash and get hacked. Juries, tending to be disproportionately retired or civil service workers and tabloid readers, can be predicted to be uncomfortable with the newest technology of the day.

If bystanders are killed, that’s easier to understand as an unintentional outcome of an accident and damages, even if awarded, won’t be as costly and sales won’t suffer as much. The killing of bystanders is not new. The killing of occupants is not new either, but designing to prefer to kill occupants may be new. Some people risk their lives, such as soldiers (examples are legion) and mountain climbers (Mt. Everest is littered with human carcasses), but this would be altruism that even soldiers and mountain climbers probably would decline. Soldiers and mountain climbers do embrace safety in themselves, their companions, and their equipment. They’ll expect no less in their cars. If people were altruistically to sacrifice their own and their families’ lives, more of them would die under ambivalent conditions, that population would dwindle, and few people would mimic that model, making deadly altruism self-limiting. We have priorities for what’s important; staying alive is more important than having ice cream. By that scale, our own staying alive is more important than unknown bystanders staying alive. Plaintiffs’ lawyers won’t usually argue for the driver staying alive at the expense of bystanders; that’s too cruel. They’ll argue for family values, for the natural desire to preserve one’s own family who’s just innocently sitting in the car. Juries will agree with that.

Design is complicated and, on large projects, involves many designers. As designers weigh many considerations, probabilities are going to be more often influential on design than a stark dilemma. But the dilemma will be dramatic in front of a jury. A plaintiff’s lawyer will see to that. Lawyers can be dramatic to a jury. The drama can bring good money.

Courts will award huge damages, only partly lowered on appeal. Reputation, sales, and profits will suffer badly and long. There are plenty of other cars competing in the marketplace and some big companies have withdrawn from the national market because sales were too low.

The business depends on marketing to potential customers. That needs understanding the customers.

Most people probably don’t think of the dilemma of killing us or them, preferring to think, as I like to do, that we should drive more safely and not be so reckless as to kill anyone. Most people think the dilemma is a dangerous distraction in itself and the company should just design a safe car. People who think that way are the people who are the customers for cars.

Anyone making and selling the cars needs its customers to be confident in its products, but a decision to prefer to kill occupants is likely often to mean killing customers and their families. No business likes to kill its customers. Even gun manufacturers and makers of toxic acids don’t like to kill their customers. Google doesn’t want to kill its customers. But once accidents start happening, and, with statistically enough millions of miles of use, they’ll happen, the reputation of being willing to kill its customers will start to stick. Google will want no part of that. That’s why Google won’t make many cars or try to sell them. Google won’t have its nameplate on them.

However, the question will help contribute to developing good patents for self-driving technology. There’s much work ahead for Google; reportedly, there’s still a problem with snow hiding the edges of roads and there’s probably difficulty with a wide variety of back roads that are only slightly paved and hardly marked and not worth the cost of modernizing. There’s also plenty of informal traffic redirection with low-tech hand-wavings without GPS and it won’t do for a self-driving car to crack the legs of a nice person who was helping protect safety. An autonomous truck by another company is probably designed exclusively for use on modern superhighways and similar roads that are relatively easily managed by the technology. Google is edgier in what it is ambitiously trying to accomplish.

Google will build a large portfolio of good patents and trade secrets available for later patenting. Those patents will be licensed for good money to car makers and other manufacturers in allied industries that won’t ask that nerve-rattling question on the record but will know how to design and sell large numbers of fashionable cars.

The licensees won’t be burdened by Google’s asking that question, because the post-accident lawsuits will be against the manufacturers, not the inventors. Google’s thoughts will be judged irrelevant unless a lawsuit lies against Google, and, unless the invention in the patent was itself technologically defective in a way not evident even to an expert reading the patent (and no one should be using a patent they don’t understand at an expert level), that lawsuit against an inventor is unlikely.

Because Google will want to make its money from licensing its patents for autonomous driving, it won’t even make traditional cars that don’t use the technology. GM can make traditional cars and look good doing it, but if Google makes traditional cars, that will signal that it has little confidence in its own advanced tech, and Google knows advanced tech, so Google having little confidence in self-driving would lower the license prices for its self-driving inventions. So, other than small numbers of experimental cars for tightly-controlled experiments, it won’t be making Google cars for the public, or even for well-managed fleets. Making for the military could be an exception, but the military’s needs are for war, we’re usually nowhere near a theater of battle, and the military would likely want to buy cars from car makers who already have the experience and infrastructure to build in volume. The military may use Google’s tech, but someone else will make the cars with Google’s technology.