IN OCTOBER 1999, a local passenger train passed through a red light at
Ladbroke Grove junction and hit a packed express heading into London鈥檚 busy
Paddington station. Thirty-one people lost their lives. In the aftermath of the
crash, people were shocked to learn that British train drivers passed through
red lights on no fewer than 73 occasions that same month. It鈥檚 at times of
tragedy like these that we wish technology could step in whenever fallible
humans fail.
But are we kidding ourselves? Whether in our cars, at work or at home, many
of us just assume that technologies designed to take over during an emergency,
reduce our workload, or keep our hands free are a good thing. Our own research,
however, suggests automation should be approached with extreme care. Simplifying
a job or making it physically easier doesn鈥檛 always make things more
straightforward for our brains.
In recent years, researchers like ourselves have been looking at how
psychology can guide the design of technology, rather than just expecting people
to learn to cope with it. This is one element of the science of
ergonomics鈥攁nd it鈥檚 already thrown up a few surprises. We鈥檙e finding that
technologies as diverse as hands-free kits for phones in cars and automation
systems for railways and aircraft, which are meant to reduce accidents, may do
nothing of the kind. At best they may change the nature of accidents. At worst
they could actually cause them.
Advertisement
Car manufacturers are getting in on the automation game鈥攑erhaps not
surprising, given that 90 per cent of road traffic accidents are attributed to
human error. Volvo, General Motors, Nissan and Mercedes-Benz are just a few of
the companies working on collision-avoidance systems and intelligent
skid-control devices. BMW, Jaguar and Mercedes-Benz already sell cars that can
automatically adjust their speed to maintain their distance from the car in
front. In time, we鈥檙e also likely to see automatic steering. But are systems
like these, which take control from the driver, really the key to safe
motoring?
Tunnel vision
We all perform badly when we鈥檙e overloaded with information or things to do.
Stressed people develop a kind of tunnel vision鈥攖hey try to reduce their
workload by filtering out anything that isn鈥檛 in their central focus of
attention. It鈥檚 not surprising, then, that designers assume that giving drivers
less to do gives them less opportunity to get things wrong. But perhaps giving
people too little to do is just as dangerous.
In our driving simulator lab at Brunel University in Egham, Surrey, we can
recreate the experience of driving with anything from fully manual
controls鈥攕teering, acceleration, gear changes and braking鈥攖o fully
automated ones, where the driver simply monitors the car as it drives itself. We
measured the spare mental capacity people had when they were driving with
different levels of automation, by seeing how effective they were at completing
a series of visual puzzles at the same time. We also monitored drivers鈥� eye
movements to work out what they were paying attention to at any moment. Not
surprisingly, increased levels of automation reduced the mental demands of
driving鈥攑eople had much more time to pay attention to the puzzles.
You鈥檇 expect, then, that in the absence of the visual puzzle, the driver in
the more automated simulator would have the spare capacity to handle the
situation better if something out of the ordinary happened. Strangely, that
wasn鈥檛 the case. If the driver of a fully automated vehicle suddenly had to take
control for an emergency stop, they often didn鈥檛 react at all. Drivers were much
better at avoiding a crash if they had at least some level of manual control. It
seems that 鈥渦nderloading鈥� people鈥攇iving them no real psychological role to
play鈥攊s just as dangerous as overloading them.
We now think we know why. Traditionally, researchers have assumed that our
attention has a fixed capacity, and that we have to shift attention between
tasks more often to take in more information. But it turns out that the scope of
our attention can shrink and grow depending on what we鈥檙e doing鈥攁nd that
automation is one way to make it shrink.
In the automated car, drivers spent longer paying attention to the puzzles.
Yet their performance on the puzzles didn鈥檛 improve in proportion to the time
they spent on them. If they spent three times as long, they only solved twice as
many puzzles. Just as a computer closes down excess processing power if it鈥檚 not
needed, it seems that the mind sheds excess attention capacity when the task
gets simpler.
This attention shrinkage might explain why a sudden surprise is hard to cope
with in the automatic car. It demands more attention than you can supply, and
leaves you suddenly overloaded. Peter Hancock, a psychologist at the University
of Orlando, Florida, and an expert on human interaction with automated systems,
says: 鈥淚t is not the expected demands which the designer neglects. It鈥檚 only
when the child unexpectedly runs onto the road that the automation throws up its
metaphorical hands and gleefully returns control to the unready driver鈥攁nd
guess who will get the blame.鈥�
So it looks as if we鈥檙e faced with a paradox鈥攄esigners use automation
to relieve attention overload, because they think overload causes errors. Yet
automation can underload a driver, and this can cause problems too. One
controversial solution might be to give the driver extra tasks to do, to keep
their workload just high enough to stop their attention capacity from
shrinking鈥攊f only we knew what just the right workload was. Yet how do you
keep someone鈥檚 attention high, and still focused in the right place?
Clearly you have to be careful about what extra tasks you do while you鈥檙e
driving. As in-car Internet services and mobile-phone-based tracking systems
develop apace, we find it extremely worrying that Charles Spence from the
University of Oxford and Jon Driver from University College London have found
that the risk of having an accident while driving and phoning at the same time
is about the same as drinking and driving. We know from our own research that
talking on the phone interferes with driving performance more than talking to a
passenger does. Many people have assumed the problem is having the hands
occupied鈥攁nd the obvious technological solution is hands-free sets. But
Spence and Driver have found that using a hands-free mobile phone is no less
risky than a hand-held.
So should we be following Portugal鈥檚 lead in banning all mobile phones from
cars? Or is there a better solution? If it鈥檚 not the physical distraction of
holding the phone that鈥檚 the problem, then perhaps we should look at the
psychology.
Another long-standing theory of attention says that we have separate
鈥渃hannels鈥� for hearing and for seeing. If that鈥檚 true, the sounds of a phone
conversation would reach the brain and be dealt with by a totally different
system to the view of the road. Spence and Driver don鈥檛 believe our attention
system can be so compartmentalised, or there鈥檇 be no conflict between talking on
the phone and driving. Instead, they think that our hearing and seeing channels
are linked in a number of ways, and that one of these links depends on where the
signal comes from.
Spence and Driver asked volunteers to watch a moving road scene on a computer
monitor while repeating a stream of recorded speech that was played to them.
They looked at how well people could pick up information from both sources
depending on where the sound originates鈥攆rom a headset, the passenger
seat, or in front of the driver. They discovered that people took in most
information if the sights and sounds came from the same place.
In other words, phoning and driving at the same time should be easier and
safer if the voice seems to come from the road you鈥檙e watching. This makes good
sense. From the moment we鈥檙e born we instinctively turn our attention towards
sounds. And we nearly always look at the person we鈥檙e talking to. Spence and
Driver suggest the phone鈥檚 loudspeaker should be in front of the driver鈥攊n
effect, you鈥檇 have a talking windscreen.
While clever use of technology might help focus our attention on the road,
can it also keep our eyes open? Sleepy drivers are blamed for as many as 30 per
cent of road accidents. So perhaps it鈥檚 not surprising that there are plenty of
automatic fatigue-detection technologies in the works that spray the driver with
cold water or change the radio station if they spot the telltale signs that he
or she is dozing off鈥攍ane drifting, a lack of steering movements or even
drooping eyelids
(New 杏吧原创, 21 July, p 24).
They sound like a great idea, but are they monitoring completely the wrong thing?
Michael Russo, a lieutenant colonel in the US Army and neuroscientist at the
Walter Reed Army Institute of Research in Maryland, found that when tired truck
drivers crashed in a driving simulator, only a few of them were actually asleep.
Instead, when he and his colleagues investigated how well the tired drivers were
paying attention to their surroundings, they found what amounted to a form of
tunnel vision. 鈥淭hese people are awake, conscious and driving, but not paying
attention to their peripheral visual field,鈥� says Russo.
He says that the frontal and parietal areas of the brain, which are vital for
visual attention, spatial awareness and decision making, become particularly
sluggish when you鈥檙e tired. He even likens it to the metabolic equivalent of a
mild stroke. The message is that tired drivers are more likely to crash because
they fail to notice pedestrians, other vehicles, changing road conditions or
road signs鈥攏ot because they鈥檙e asleep. He suggests that a low-tech
stimulant like a cup of coffee might be more effective than any technological
fix for detecting sleepiness.
Fit to drive
Whether or not the detection systems are monitoring the right thing, there
are other concerns about them. The Australian National Road Transport Commission
has already voiced its concern about automatic fatigue-detection systems,
warning that drivers may ignore their own judgement if an automatic system tells
them they鈥檙e fit to drive.
Similar concerns about over-reliance on technology have also been voiced
about the automatic train protection or ATP system, whose introduction was
called for after the Paddington crash. It has a simple task: to apply the brakes
if a train passes a danger signal. At any other time, it shouldn鈥檛 interfere
with the driver鈥檚 job. Yet Swedish researchers have found that ATP can actually
influence the behaviour of train drivers, to the point where it could counteract
the safety gain.
One problem with automatic safety systems is that people form expectations
about what they are designed to do, and can even learn to depend on them.
Instead of driving normally, train drivers might push the ATP system to its
limits, skirting the boundaries of safe performance. Lisbeth Harms of the
Swedish Road and Transport Research Institute is worried that equipment intended
to reduce driver error might not be appropriately designed. Speaking at an
ergonomics conference in Stratford-upon-Avon in 1996, Harms concluded that while
ATP has the potential to reduce Paddington-type incidents, it might not
eradicate them. It could even spawn new types of errors. Drivers might end up
relying on ATP rather than the signals themselves鈥攖he last line of defence
becoming the first point of control.
One thing鈥檚 for sure鈥攁utomation doesn鈥檛 automatically make things
safer. While we don鈥檛 wish to hold back the tide of technological progress, we
must be cautious about how we use it. Automation systems should still engage
drivers in the business of driving, rather than trying to make them redundant.
If we can鈥檛 work out how to do this, perhaps we should put automation systems on
hold until we can.
With so much technology at their disposal, designers might say this defeats
the object. It essentially takes us back to square one, with the driver firmly
behind the wheel. But until we really know our own minds, that鈥檚 exactly where
the driver should be.