Ouverture de session
Oublié votre mot de passe?
Formation >> SMS >> Articles Imprimer cette page   Envoyer cette page à un ami     A   A   A 
Article 3 - Inattentional blindness: let's not blame the victim just yet
 
Occupational Health & Safety
Ja/F'02
Page: 23-29
Byline: Marc Green
Source: The Canadian Index (Business)
Volume: VOL. 18, NO. 1
Full Text: Yes

Many accidents that seem like ``our own fault'' are caused by the simple impossibility of paying close attention to everything, all the time. (And the sooner we understand this, the sooner we can design tools, tasks and jobs to reduce the risk of our own inevitable lapses.)

A driver pulls out onto the road right in front of you. A roofer at a construction site steps backward into a hole for a skylight. A control room technician at a pulp and paper mill fails to notice the needle moving into the red on the pressure gauge right in front of her nose. A cut-off saw operator at a sawmill reaches under the guard to remove a broken piece of wood even though the saw is running.

What causes all these errors and why are they so difficult to prevent? The reasons are often complex, but they generally arise from a simple fact: Humans have a finite mental capacity and are unable to process all the sights, sounds and other input that floods the senses at any given moment. To avoid being overwhelmed with useless information, humans have evolved a mechanism to filter away the irrelevant input, allowing only important information to reach consciousness. We call this mechanism attention.

Because of attention, seeing requires more than simply pointing the eyes in the right direction. Attention acts as a gatekeeper that decides what is worth noting and what should be ignored. It allows us to focus our limited mental resources efficiently to the task at hand. William James, the founder of behavioural psychology a hundred years ago, summarized the benefits of attention by saying, “Only those items which I notice shape my mind -- without selective interest, experience is utter chaos.”

While attention enables us to interact intelligently with the world, it can also cause accidents and injuries. Attentional filtering has a cost -- people are “inattentionally blind'' to most of the world at any given moment. If attention mistakenly filters the wrong sensory input, then we may be blind to important information. Time and again, accidents occur because someone looks but fails to see something that, in retrospect, should have been plainly visible: train crews miss flashing stop signals, drivers fail to see other cars, workers fail to notice warnings.

People who make such errors are often viewed as stupid, careless or incompetent. In fact, occasional look-but-fail-to-see errors are very difficult -- perhaps impossible -- for anyone to avoid 100 per cent of the time. The person making the error does not know about the inattentional blindness because, ironically, he or she is, by definition, unaware of being unaware. Training and instructions to ``pay attention'' have little effect because most of our perceptual processing occurs automatically and without our conscious knowledge.

Onlookers may find satisfaction in blaming people for errors due to lack of attention, but this does nothing to explain the error's cause or to suggest ways of preventing future accidents. It is more productive to determine how inattentional blindness errors occur in the first place.

The starting point is analysis of attention's fundamental properties: limited capacity and selectiveness. Selection occurs in two distinct modes. One is called “endogenous'' (from the inside) selection in which the viewer consciously controls attention and scans the scene. The second is automatic “exogenous'' (from the outside) control, which occurs when some conspicuous object automatically grabs attention and gets noticed. The factors that create conspicuousness are critical for getting people to notice unexpected information.

Attention also has limited capacity, so the more attention devoted to one input, the less that is available to others. It is important to understand how the individual allocates attention across tasks and learns to perform with minimal attention in automatic behaviour.

Selection modes attention is often conceptualized as a spotlight that focuses on a selected part of the visual field. The viewer consciously perceives the illuminated portion and is blind to the rest. In order to view the entire scene, attention must move focus from place to place. During normal viewing, attention moves with the eyes, attempting to bring areas of interest to the central part of the visual field. The viewer then focuses the image on the highest-resolution part of the eye and sees most clearly. The spotlight's movement is guided in two ways. Viewers can consciously (endogenously) reposition their attention. People may be searching, as when seeking a specific tool or person, or may be merely exploring, as when window-shopping. In either case, the eyes usually move in a series of quick, sharp jerks called “saccades.'' The eye rests for only about a quarter second between saccades. If there is nothing to warrant attention in that brief time, the eye moves on to the next sample. The trajectory that the eye takes is called its “scan path''.

At other times, attention moves without volition. Something simply pops out of the scene, grabbing attention and directing the viewer to look at it. These things usually involve motion, or bright colour, or something that is conspicuously different, unusual or out of place. This exogenous attention is critical for getting people to notice warnings and dangers that they do not know about in advance or have forgotten. The attentional spotlight has variable intensity, so the viewer may examine a large area with a low degree of attention, or a small area with a high degree of attention. The beam is wide when attention is not heavily taxed and there is some spare capacity. Conversely, the beam can become very intense and narrow when the task is difficult, visibility is poor or stress is high. The resulting condition is called “tunnel vision'' because ability to see objects in the peripheral field declines.

However, tunnel vision and the spotlight metaphor cannot explain all look-but-fail-to-see errors. People frequently fail to notice objects that are located directly in their line of sight. There are many cases of drivers, for example, looking directly at another vehicle and failing to see it. In some cases, attention is focused on objects rather than on locations. Attention selects near or far as well as left, right, up and down. When focusing on nearby objects, for example, a viewer may fail to notice objects at a distance, even though they lie directly in the line of sight.

  A good example comes from aviation, which has developed heads-up displays (HUD) for pilots. HUDs project information up in front of the pilot's eyes or onto a helmet-mounted display. The idea is that pilots need not look down at their instruments on the console and can keep their eyes looking straight ahead out the windshield. Simulator studies, however, find that HUDs can cause runway incursion accidents, where one plane collides with another on the runway. While it might seem unlikely that a pilot would fail to see something as big as a jumbo jet sitting straight ahead, this is exactly what happens. HUDs, and their attendant dangers, are not limited to aircraft. In some companies, maintenance and repair staff use head-mounted displays that project information onto a screen located a few inches from the eye. This removes the need for technical manuals and frees the hands.

HUDs will also soon become common in automobiles. Drivers will be able to view the speedometer and other gauges without looking down, but the potential for ”roadway incursion'' accidents is huge. At other times, a particular object captures attention to such an extent that capacity is entirely absorbed and everything else goes unnoticed. The resulting inattentional blindness has caused dramatic accidents. For example, an airliner crashed because the crew became so engrossed with a blinking console light that they failed to see the ground approaching through the windshield or to hear the danger alarm.

Finally, other cases of inattentional blindness are more puzzling because the viewer fails to see the obvious even without tight focus on another object. In one accident, for example, a forklift operator simply failed to see a large power pole, even though it lay directly ahead in unobstructed view. In these accidents, the usual culprit is “violated expectation''. It may be true that seeing is believing, but it is just as true that believing is seeing. 

Conspicuousness Safety professionals have long recognized the importance of getting people to notice objects and information that are not actively being sought. The usual approach is to heighten conspicuousness, the ability of an object to attract and/or hold attention. Roughly speaking, conspicuousness factors fall into two categories, sensory and cognitive. There is little doubt that humans have an innate tendency to notice some sensory input more than others. High contrast, large size, motion/flicker and perhaps colour can all increase conspicuousness at times. Most recently, there has been much interest in the potential of fluorescence as an attention-grabber. Safety professionals often capitalize on these innate predispositions to promote sensory conspicuousness as the cure for inattentional blindness errors.

This has generated a cottage industry in both conspicuousness enhancement products and in research designed to reveal the best colours, shapes, icons, positions or types of motion for creating conspicuousness.

However, sensory conspicuousness factors have limited power. For example, most people's introspection suggests that motion and flashing increase conspicuousness. Everyone has had the experience of suddenly noticing a moving object out of the corner of the eye. (Of course, this impression is dubious because we don't know about all the times when we failed to notice a moving object. We are not aware when we are not aware.)

Early attention studies confirmed the introspection, but more recent evidence suggests that motion and flashing have limited power to improve conspicuousness. Generally, flashing and motion, as well as other factors such as colour and brightness, increase conspicuousness if the viewer is not already focused on any special activity and has some spare mental capacity unused. If the viewer's attention is already tightly focused on work or if expectation is powerful, however, then flashing and motion will have little effect.

At its essence, perception is the search for meaning. It is hardly surprising, then, that attention is primarily attracted by meaningfulness. The classic example is the cocktail party phenomenon. You are at a cocktail party and having a conversation with someone. You understand the words of your partner and may be vaguely aware of the buzz of other, unintelligible conversations, but do not have the capacity to interpret both your partner's speech and those of other conversations at the same time. You can scan the room and switch your attention to someone else and then understand that conversation, but your partner's words become a meaningless buzz. Now, suppose someone behind you says your name. This breaks through the attentional filter, automatically attracting attention to the other conversation.

The cocktail party phenomenon is instructive for several reasons. First, it shows that attention is limited to a unitary stream of consciousness that may be controlled by either endogenous volition or by exogenous imperative. Further, it shows that even though we are not consciously aware of what is being said in the background, we must be constantly monitoring it at some unconscious level in order to pick out a name from the buzz. This further demonstrates that most of our perceptual processing occurs outside of awareness. The concept of meaningfulness is somewhat vague, but studies have isolated a few important factors.

Meaningfulness can be based on self-reference, such as a name, city, children's names, and so on. Studies examining conspicuousness of highway signs found that meaningfulness can be created by communicating an immediate call to action. People almost never notice or remember signs unless they contain an imperative: Stop, Slow Down, Yield. If they merely say, “Caution'' or “Hospital Ahead,'' they do not penetrate our attention very well. (Of course, if the driver were sick, then the “Hospital'' signs would be more likely to be meaningful and get noticed.) Violated expectation in most cases, we learn what is meaningful through direct experience. People are highly adaptive and, when experiencing the same situation repeatedly, quickly develop expectations about what is and is not relevant.

A phenomenon called “violated expectation'' is perhaps the most frequent cause of inattentional blindness. For years, authorities attributed the high rate of motorcycle accidents to low conspicuousness, assuming that the machines' small size limited their visibility. Safety experts strongly encouraged the use of bright clothing and reflective materials. However, accidents still occurred when the motorcycles should, in theory, have been highly visible. It is more likely that motorcycle accidents happen so frequently because car drivers don't expect to see them. They are relatively rare, fewer than one in 120 vehicles on the road. Automobile drivers expect to see more common objects such as road signs and other cars. Statistics on pedestrian accidents show the same trend. Cars hit relatively few pedestrians where they are expected -- on crosswalks. Instead, the majority of accidents occur when pedestrians cross the road at unexpected locations in the middle of the road .

Expectation also affects where we look in the first place. If experience has taught that there will be no useful information in one part of an environment, then people will simply remove that area from the scan path and see it only in low-resolution peripheral vision, if at all. A good example is the common accident scenario of a driver coming to an intersection intending to turn right while a bicycle approaches from the right. Drivers have learned from long experience that approaching cars would come from the left. There is almost never anything coming from the right, so they don't look in that direction and don't see the bicycle until impact. If the drivers did glance to the left, there would be only a quarter second to overcome the expectation and see the bicycle.

Ironically, highly skilled and highly practiced experts are more likely to commit inattentional blindness errors than are beginners. They have had more chance to develop expectations and to have the expectations repeatedly reinforced by experience. Inattentional blindness accidents are especially likely when expectation and distraction combine.

Cmdr. Scott D. Waddle, of the submarine USS Greenville, gave the order to surface, and overturned a Japanese fishing boat, killing nine people. With visitors observing onboard, the sonar operator told Waddle that there were no blips on the screen. Waddle had learned through long experience that a clear sonar screen meant no ships in the area. When he subsequently scanned the ocean through the periscope prior to surfacing, he did not expect to see a relatively small fishing boat. Further, the visitors doubtless occupied his attention, further reducing the likelihood that he would see the boat.

The human tendency toward “confirmation bias'' strengthens the expectancy effect. People who hold a belief or expectation tend to seek evidence that confirms it and ignore or avoid evidence that refutes it. For example, people who favour one political candidate will listen to his or her speeches but immediately turn off the TV if someone starts extolling the virtues of the opponent. The same effect occurs unconsciously. Cmdr. Waddle peered through the periscope, unconsciously looking not for ships, but for evidence to confirm the sonar reading that there were no ships in the near vicinity.

There are some extraordinary cases of confirmation bias leading to accidents. In one that killed 500 people, a train collided with another when seven highly trained and experienced members of the crew failed to notice that they had pulled out of the station going in the wrong direction. Even though familiar with the station and the surrounding area, they proceeded for many minutes before running into another train on the track. Although attention is limited, its capacity varies with circumstances.

Aging lowers attentional capacity and impairs perception. People in a low state of arousal due to alcohol, drugs or fatigue have less attention. People performing dull and monotonous tasks also tend to lose arousal and the ability to perceive new information. While safety experts have traditionally focused on attentional overload, it is becoming a more common problem. The major culprits are computers and other technological advances that have allowed automation of an increasing number of job functions. In theory, this should reduce human error by transferring important tasks from the fallible human to the reliable machine. The human then switches from the active role of operating the system to the passive role of merely monitoring the equipment.

Unfortunately, people invariably suffer “vigilance decrement,'' a rapid decline in ability to detect information after half an hour of monitoring. Viewers then quickly lose arousal and fail to notice new information. In England , for example, there have been major train collisions that occurred when the crew simply failed to see a plainly visible signal. This type of accident occurs so often that it has its own name, SPAD, for “signal passed at danger''.

Accident inquiries usually attribute SPADs to low arousal caused by a combination of fatigue and boredom. Moreover, the safety benefits of automation are often consumed in the name of efficiency. Autopilots should make flying safer by automating functions, leaving pilots free to attend to other tasks. The airlines, however, soon started using automation as a means for reducing safety margins, such as the time between takeoffs and landings on the same runway. Now, pilots have even less room for error than before. Of course, the autopilot also means that the human pilots have one more instrument to monitor and one more piece of equipment that can go wrong.

Automatic behaviour learning can, in a sense, expand attentional capacity. As already described, developing expectations helps by removing the need to waste attention on the irrelevant. Learning can also lower the attentional demand by creating “automatic behaviour''. With extended experience, people learn to perform tasks with little or no attention.

For example, an operator learning to use a device must initially pay very close attention in order to know where the hands and fingers go, read instruction or warnings, locate switches and so on. With extensive practice, the response becomes automatic, meaning that the individual movements become linked in a chain which, once started, seems to run of its own accord. The need for external visual information is replaced by ``muscle memory'' as internal (“proprioceptive'') feedback from each hand movement guides the next. The operator seemingly performs the task without attention and can simultaneously carry on other activities, have a conversation, read or listen to the radio. Virtually all skilled performance involves this replacement of conscious, attentive response with automatic behaviour. Automatic behaviour is a usually beneficial because it circumvents attentional limitations and improves performance, but it can also produce disastrous results. Since attention is no longer required, people become inattentionally blind to situational changes such as wear, equipment fault and warnings.

Dr. Jeff Justis, an American hand surgeon who has treated hundreds of shop-related injuries from power tools, reports that accidents most often occur when an experienced person is performing a highly repetitive task. He further notes accidents often happen when an unexpected event occurs, probably because it interrupts the response chain at a critical point. A similar situation occurs when an operator is highly practiced with one machine, but then must use a similar but slightly different one. The operator continues to rely on muscle memory and fails to notice differences in the new situation because of inattentional blindness.

One employee operating a press brake, for example, lost fingers from both hands using a new, unguarded press. His usual machine had a guard, but he failed to notice the difference. Old habits die hard.

From what we know about inattentional blindness, it is obvious why investigators and onlookers often attribute accidents to the individual involved rather than to the situation. The inattentionally blind person had a strong set of expectations and habits. Perception may have been limited by clutter, low conspicuousness, poor legibility or low contrast and there was likely a limited amount of time to detect and assess important events. He or she may have been overloaded by performing several tasks simultaneously or under low arousal from boredom and fatigue.

In contrast, people performing the analysis after the fact know in advance what has happened, so they know where to look and what to inspect. They are free of the accident victim's expectations and automatic behaviours. The investigators view with focused attention. Their vision is optimized and they can examine the scene in a leisurely fashion with no time limitations. They are alert and do not have to worry about attentional overload from simultaneous tasks. They are aware while the accident victim was unaware. Instead of blaming the individual, it is a better strategy to analyse when and how inattentional blindness is likely to occur and to design an environment that prevents accidents when the inevitable errors occur.

Perhaps the best advice on this subject comes from the Swedish-Canadian arctic explorer, Vilhjalmur Steffanson. He noted that there are two kinds of hazards, real and mental. The real hazards arise from physical circumstances while the mental ones from our perception, expectation and beliefs about hazards. Steffanson said that we would be more successful in changing the real hazards, because man has always found it easier to change the world than to change himself.

Marc Green Ph.D. is a consultant with ERGO/GERO Human Factors Science. He is also an adjunct faculty member at the Department of Ophthalmology, University of West Virginia Medical School in Morgantown , West Virginia , and at the Centre for Vision Research at York University . He can be contacted through www.ergogero.com.

Idnumber: 200202015639
Subject: Attention; Perception; Accident prevention; Occupational health and safety -- Research and development
Version: CBCA: Fulltext Business; CBCA: Fulltext Reference; CBCA: Index
Feature: Illustrations