Applying Military Research to a Very Different Problem
Many years ago, in the era before development of stand-off air-to-ground weapons, an obscure Soviet research paper appeared in English translation examining the growing challenge of attacking a well-defended ground target from the air.
For the Russian researchers, the ground attack problem was binary. To survive in the teeth of increasingly dense and lethal anti-aircraft artillery, Triple-A in the jargon, attacking aircraft would have to eschew standard and thus predictable flight profiles, instead continuously "jinking" on their run-in to the target.
But the physical disorientation associated with those non-standard flight maneuvers perforce would enormously complicate pilots' ability to aim their weapons and release them at the correct moment.
The solution, the researchers believed, might lie in automating one or the other of those two tasks, either controlling the aircraft during its final attack run, or the targeting and release of its weapons.
Accordingly, like good Marxist empiricists, the researchers set out to learn by experiment which of those two activities ó flying or targeting ó would be more susceptible to successful automation.
The results were surprising. The researchers had anticipated that targeting the weapons would be easier to automate and more effective than flying the airplane. After all, attack aircraft of the time already used computer-assisted sights to aim and release their bombs and rockets.
Instead, actual experiment proved the opposite. The most difficult part of the attack sequence wasn't aiming at the target or timing release, but rather acquiring the target in the first place and retaining acquisition throughout the attack run.
Difficult to do even on a normal flight profile, it became almost impossible while the pilot was throwing the aircraft all over the sky to avoid Triple-A.
What ended up happening was that, desperate to acquire and lock up their targets, pilots tended to abandon jinking at the very moment when the aircraft became most vulnerable. The result was predictable.
Instead, the researchers concluded, the better solution was to turn management of the aircraft during the attack run-in over to the computer, freeing the pilot to acquire, target and release his weapons without worrying about flying the airplane.
Of course, the development of precision stand-off weapons rendered the issue moot. Apart from aircraft like the elderly A-10 that were built specifically for close air support, today's attack aircraft normally stay well out of the reach of anti-aircraft artillery, though they remain vulnerable to ground-to-air missiles.
But the study was more interesting for what it revealed about human-machine symbiosis. And the issue to which it might still have something valuable to contribute, incongruous though it may seem, is to the current debate over self-driving automobiles.
Today that debate centers almost entirely on the ability of computer-controlled vehicles to negotiate the challenges of America's roadways. Just this week, for example, the Washington Post reported that a driverless shuttle in Las Vegas suffered a collision only two hours after being put in service.
The article was careful to point out that the shuttle was the victim of the collision, not its perpetrator, backed into by a delivery truck in circumstances in which the shuttle couldn't back up. But the episode has only fueled the concern of those doubting the safety of turning driving over to a computer.
As was true of the Soviet study of attack aircraft automation, however, the question of automobile automation is binary. The issue isn't whether having a computer drive is dangerous. It's whether leaving the driving to humans who may or may not be skilled or sober, and who very likely will be talking, texting, and otherwise distracting themselves, isn't more dangerous.
As a recent RAND study noted, "More than 90 percent of crashes are caused by human error, such as speeding, miscalculating other drivers' behaviors, or driving impaired."
That doesn't even begin to address the operator distraction produced by the phones, tablets, nav systems, dashcams, and other electronic gizmos with which today's automobiles increasingly are loaded.
Efforts through exhortation and fines to discourage the use of those gadgets have thus far proved futile. For starters, prohibitions on their use are nearly impossible to enforce until after they've caused an accident.
So, unless we're prepared by law to mandate the installation of some sort of "kill" switch that automatically shuts down every electronic gadget the moment the vehicle is put in gear ó about as likely as snow in July ó maybe we should take a lesson from that long-ago study, and recognize that it's both easier and safer to have the computer drive the car, while allowing the passengers to phone, text, play and otherwise distract themselves to their hearts' content.
Doubtless it would please the Soviet researchers to learn that, albeit differently from what they intended, Marxist empiricism paid off after all.