Our Need for Speed
Los Angeles Times Magazine, Aug 2004
Seat Belts, Air Bags and Anti-Lock Brakes Have Made Cars Safer, But There’s One Problem Engineers Can’t Solve: the Driver.
Also at Los Angeles Times
In 1981 the German Transportation Ministry conducted an experiment into the effectiveness of anti-lock braking systems, only to discover that no brakes can stop the hurtling imbecility of human nature.
The ministry fitted half of the 91-car Munich taxi fleet with anti-lock brakes, which at the time were considered a major leap in road safety. Then it monitored the fleet for three years using accelerometers, which measure G-forces, and undercover passengers. Every care was taken to avoid corruption of the data–the drivers were switched among the cars, for example, and they were not told which cars had anti-lock brakes nor how they would be monitored. The results were clear. Not only did drivers of ABS-equipped vehicles have slightly more accidents, on average, they also tended to brake harder, accelerate faster and create more traffic conflicts. In other words, the wonder brakes did nothing for road safety, but rather had the opposite effect: They encouraged drivers to speed up and be more reckless.
Granted, German cabbies represent only a small number of all road users, but nevertheless their behavior raises a number of questions about road safety, psychology and, ultimately, the way human beings are wired. Does this happen with all car safety features? Do we thwart them, albeit inadvertently? Maybe this is the tension of evolution–as science urges us to higher ground, our grubby animal nature tugs the other way. But why does the animal so often win?
According to Gerald Wilde, a Dutch psychologist now living in Ontario, Canada, the Munich experiment is best examined through the prism of risk. In his 1994 book “Target Risk”–since updated to “Target Risk 2″ in 2001 and with a third installment on the way–he argues that the cab drivers were exhibiting “risk homeostasis,” or our tendency as a species to maintain an optimum level of risk in our lives. The theory holds that when we feel safer because of anti-lock brakes, condoms or childproof bottle tops, to cite just a few examples, humans compensate with riskier behavior, such as driving faster, having sex with more strangers or being less vigilant in monitoring children’s access to medicine. The result is the same in many scenarios–the accident rate remains relatively unaffected despite the best efforts of scientists and legislators to reduce it. For Wilde, “safety features” are nothing of the sort.
The numbers are compelling. In the United States between 1923 and 1996, the number of annual road deaths per 100,000 residents has not shown the steady decline one might expect, but rather has fluctuated at about 23, on average. Though there has been a net decrease in accidents since the early 1970s, the graph is jagged, and the fluctuations bear no apparent correlation to the advent of seat belts, air bags or anti-lock brakes. This is not to say that cars today aren’t safer, because they are. Per mile driven, the fatality rate has dropped twelvefold in the same period. But today we drive more miles, crowd the freeways with more cars and drive faster–so many of us are still dying, per capita, as often as we did in the Jazz Age.
The logic seems intuitive at first: A perception of safety encourages risk-taking, just as a sense of danger encourages caution. In an uninsured, rusty jalopy with a busted windshield and dodgy brakes, a driver usually is on full alert, if only to watch for cops. But tuck that driver into a new Volvo, surrounded by state-of-the-art air bags, comforted by a small insurance deductible and reassured by the tug of the seat belts, and there’s a decent chance he’ll wind up in a hedge, or worse.
Studies show that a sense of safety lulls us into danger, hampering the effectiveness of pedestrian crosswalks and low-nicotine cigarettes. It also explains why the number of flooding deaths per capita in America hardly changed between 1906 and 1985, despite the construction of stronger levees in at-risk flood plains. Many people moved into the flood plains thinking that the new levees made them safe. When floods did occur, their impact was much greater.
The lull, however, is only one factor in keeping our risk level high. Another is our appetite for thrills and brinkmanship, as shown by the Munich taxi drivers. A study in 2000 by a graduate student at Western Oregon University showed that recently improved parachute rip cords did not reduce the number of skydiving accidents. Instead, a new spate of accidents emerged, caused by skydivers pulling the rip cord too late. The end result was that the accident level remained unchanged.
Perhaps this is the work of natural selection, the ruthless culling that permits survivors to label the dead as unfit. Or alternatively, risk homeostasis may be stalling our evolutionary ascent. No matter how “safe” our systems, we have the weight of human fallibility dragging us down.
Car safety engineers and legislators, who are the Victorian schoolmasters responsible, among other things, for the nagging seat-belt reminders in today’s cars, sometimes regard us as children who respond only to threats of punishment. And because there is no greater curative than the swish of a master’s cane, traffic humps now rib the streets like welts and we are beset by stinging caution signs.
But mere scolding hasn’t been effective. “You can’t shrink a balloon by squeezing it,” Wilde says. “You have to let the air out.” Armed with study after study, Wilde has set about slaughtering every sacred cow of the safety industry experts. Nothing is foolproof: seat belts, air bags, helmets and traffic lights at intersections. Stop signs, he says, are “liability protection.
This is powerful stuff, but is it true? A cautious driver, for example, wouldn’t use anti-lock brakes or air bags as an excuse to start skidding around corners. “But the roads are not driven exclusively by little old ladies,” says Wilde. “So the little old lady has to take more risks than she would like to.”
Then what of seat belts? We’ve all seen the lurching dummies in the crash-test videos. How can seat belts not have reduced road deaths? “They probably would have if all else remained the same and people drove in the same way, but they don’t,” Wilde says. “An American study in 2002 showed that even though car occupant seat-belts usage has risen since 1985, it has not had the effect of reducing the traffic fatality rate per head of population in 14 years of data.”
Needless to say, Wilde’s theory of risk homeostasis hasn’t been embraced by those whose work he has, in effect, trashed. He has a steadfast opponent, for example, in Adrian Lund, chief operating officer of the Insurance Institute for Highway Safety in Virginia, where much of the crash testing is done. Although Wilde jokes that he is the Galileo of safety science–a humble researcher shaking the foundations of powerful institutions–Lund isn’t convinced. “It’s [an] interesting [theory], but I’ve looked into it, and it’s not true.”
In the 1980s, Lund conducted his own investigations into risk homeostasis and found that it applies only to anti-lock brakes, studded tires and other automotive features that give the driver active feedback. When you can feel the better brakes and surer handling, then some drivers instinctively exploit them. “But [the theory] doesn’t make any sense for passive safety features” such as air bags and seat belts, says Lund. “We know safety belts work. Why do you think fatalities have come down since 1975?”
Wilde isn’t swayed. He claims that many other things “like the economy, like the unemployment figures ” happened in the ’70s that show a better statistical correlation to the fatality figures. When the economy’s good, there is more movement of goods and people, and fatalities go up, he explains. When the economy is weak, fatalities go down.
Lund calls that argument “statistically fallacious. You just can’t take such a big macro view. Fatality figures vary according to lots of variables, so it’s only by doing carefully controlled studies that you can narrow your field to be meaningful.”
Rather than adjudicate a squabble over statistics, let’s put Lund’s opposition to risk homeostasis into a broader context. It’s not surprising that he should look with disfavor on a theory that regards the crash-test research he oversees as irrelevant to the advancement of road safety. Like world peace, road safety is an unrealistic but worthy goal.
For carmakers, the argument that safety features ultimately may not save lives does not alter the fact that safety features sell. The sport utility vehicle boom owes much to misperceptions about safety, as Keith Bradsher writes in “High And Mighty SUVs: The World’s Most Dangerous Vehicles and How They Got That Way.” Even though they roll over, rate poorly on collision avoidance and are vastly outperformed in crash tests by most minivans, SUVs impart a feeling of snug security. Drivers are fooled by all that height and weight, and often by the way the dealer serenaded them with talk of side-impact air bags, padded dashboards, whiplash-protection headrests and collapsible steering columns–a kind of safety-age lullaby rocking them gently into a sale.
Some car manufacturers have appeared open-minded about risk homeostasis–Volvo invited Wilde to speak in Brazil, and the Japanese auto industry, particularly Toyota, is “much more receptive,” or so Wilde has heard on the academic grapevine. But most have studiously ignored the theory, and others have been hostile. Several years ago, Wilde claims, a researcher from General Motors compared risk homeostasis to the theory of ether in physics–the long-held notion that an invisible, odorless medium filled all unoccupied space–which was finally discredited by Einstein’s theory of relativity
No brand has better reaped the benefit of that safe feeling than Volvo, which enjoys a reputation as the maker of sturdy but dull cars able to withstand the impact of meteors. Though Volvo now is fashioning a younger, funkier image, it remains the standard-bearer for the safety features that Wilde says have no effect. And the features keep coming. Volvo built a safety concept car in 2001 to showcase its latest gadgetry. It included blind-spot warning systems, rear-facing cameras, collision warning sensors, crisscross seat belts, a talking dashboard and a video screen–all the wonders of 21st century gizmology marshaled for our protection, or at least the impression thereof.
“It was too much information” says Ingrid Skogsmo, head of the Volvo Safety Center in Gothenburg, Sweden. “[People] didn’t know where to look, so we got rid of the screen and the voice warnings.”
Skogsmo encounters human limitations all too often in her work. “My job is to encourage correct driver behavior, but it can get a bit frustrating. Probably the best device is the seat belt, and people don’t use it. So we as an industry don’t get the optimum effectiveness of it.”
For Skogsmo and other safety engineers, the best solution might be to remove humans from driving altogether. The logical goal of the safety industry is to fully automate the driving task, and the writing is on the wall. Cars are full of little brains that attend to problems unseen, leaping unbidden to our rescue without our knowledge. Anti-lock brakes are one example, and the rollover protection you find in SUVs is another. Adaptive cruise control has arrived. We’re already surrendering the keys.
James Moore, a professor at the Daniel J. Epstein department of industrial and systems engineering at USC, envisions a future on the freeways worthy of “Minority Report” author Philip K. Dick–streams of cars tucked in close together, driving themselves at high speed. Without the sluggish constraints of human reflex times, stopping distances can shrink, speeds can rise and traffic flow can be maximized. The whole enterprise starts humming with efficiency.
“In 1997, there was a demonstration project in Southern California,” Moore says. “We had automated platoons driving in the carpool lanes. So there were people in the cars, but they weren’t controlling them. And [the cars] executed all kinds of maneuvers like passing, overtaking, speeding up, slowing down.” Moore talks about cars with “internal intelligence” that speak to each other, that make decisions on our behalf. But safety does not necessarily follow from automation. Accidents are rarer, but more severe. As Moore acknowledges, “the stakes of failure [in an automated highway] are much higher. The cost of a bug in this system is much larger.”
“Automation is not the panacea some people think it is,” warns professor Najmedin Meshkati, also of USC, who has investigated safety as it applies to air traffic control, refineries, nuclear power plants and other technological systems. He describes human operators as having “boundaries of loss of control”–boundaries that we like to explore because we’re such curious creatures. “It’s like holding your hand out in a darkened room,” he says. “You develop an understanding of your boundaries, a mental model, and you adapt to those boundaries. It’s the same with technological systems. We naturally seek out our limits.”
Automation has tampered with these limits. By relieving us of the driving task, car safety engineers are eroding our skills as drivers. The visceral thrill of stick-shifting a high-speed, gas-burning chariot of roaring metal has been replaced by an anemic purr and the mild push of a pedal and a tweak of power steering. Cars have become passive transportation cockpits, leaving us with nothing to do. So we seek out distractions, obsessing about fripperies such as cupholders, cellphones, iPods, the GPS and the CD changer. As a result, drivers are not as sharp or engaged as they once were. We’re devolving into flabby house cats, or as Meshkati would have it, “de-skilled human operators. [They’re] a serious problem for automation.” If something happens that requires a human override and the operator is de-skilled, “then that’s potentially very dangerous,” Meskati says.The irony is delicious: The greatest danger that an automated system faces is a human operator whose skills have been diminished by automation. This is how risk homeostasis works. To use Wilde’s metaphor, automation is merely “squeezing the balloon.”
Though the idea that human beings are doomed to crash no matter what seems pretty gloomy, the glass may be half full, depending on your perspective. Risk homeostasis needn’t take a pessimistic view of human nature.
There is something reassuring, for example, in its assertion that human beings will always reclaim their primacy when it comes to technological systems. For those who are alarmed by the whip-crack pace of modern technology, particularly its preference for machines over people, it’s a comfort to be reminded that we can, by the sheer creativity of our ineptitude, keep at bay any soulless robo-future that safety scientists have in mind. Risk homeostasis strikes a blow for the inventiveness of human nature over computerized homogeneity. It promises that we’re up to the challenge of new technology. For every new safety feature, our inner chimp has the antidote; for every foolproof innovation, we can muster a better fool. The effect is almost Newtonian.
Perhaps the most optimistic outcome of risk homeostasis theory is that it offers a chance to reexamine the orthodoxy that continues to install seat belts in our cars on the one hand and fine us for not wearing them on the other. Rather than focus on individual features and transgressions, risk homeostasis offers a more holistic approach to road safety. By embracing human nature with all its quirks and faults, it purports to treat the patient rather than the illness.
Wilde suggests that we shift the prevailing philosophy of punishment for reckless behavior into one of rewards and incentives, that we bet on greed rather than fear as a motivating force of change. “There are several studies from all over the world that show that employers who offer their drivers incentives [to keep accident-free] have a lot of success,” he says. “Incentives increase the potential value of the future. They encourage you to dial down your level of acceptable risk. It’s a very positive philosophy.”
In one example, professional truck and van drivers in Germany in 1957 were offered a financial bonus for every six months they went without being at fault in an accident. After a year, the accident rate dropped by almost one-third, and it kept sliding for the next two decades to a stable level of 14% of what it had been prior to the incentive program. The cost of running the program was estimated at $35,000 a year, and during the more than four decades it has been in place, “these costs are reported to be far outweighed by the reduction in insurance fees resulting from the much-improved safety record.” In Iceland, the accident rate is minuscule because drivers there get free car insurance after 10 years if they remain accident-free.
Although he’d like to see such measures tried in the U.S., Wilde doesn’t foresee a day when the carrot supplants the stick. “It’ll never happen in America,” he says. “Who’s going to profit, apart from drivers?”