Introduction

From where did this imperfect behavior come? No one learned the "hold back" behavior from a driver training manual. No one learned it from applying physics. (If they did, rocket scientists would have chosen a different behavior from fruit pickers.) Instead everyone learned the same behavior from experiencing things sliding forward during abrupt stops. After a couple of spilled grocery bags, arms almost unconsciously pulled back on whatever is on the seat before braking. We, today are still learning the "hold back" behavior from these experiences. We are still putting out our arms during heavy braking. The increase in injuries is avoided only because our kids are wearing seatbelts. If they were not, our normal learning processes would still be adding to the highway death toll.

If you want to understand the physics that explains why holding back children increases their injuries, read the note at the end of this introduction. If you want to understand that learning process and its implications for human wellbeing, read this book. I describe how normal learning results in behavior that causes scarcity, social conflict, and environmental destruction, just as surely and unintentionally as holding back the unrestrained increases their injuries. I show how normal learning fails to accurately predict future conditions, weakens values for predicted conditions, and diminishes ability to connect predicted conditions to behaviors that cause them.

Some of these learning processes are acquired after birth. Each of us learned them from interactions with the environment. Part of this environment was produced by our culture. Therefore, part of six billion sets of imperfect behavior is the unintended by product of cultural activity.

These behaviors might be improved by changing what culture contributes, not to knowledge, but to cognitive development. Finding and implementing these changes might develop a new generation whose thinking and learning takes them to an environmentally balanced, abundant, and peaceful future.

The learning processes on which I focus contain a common thread "time." Our perception of time shapes both our predictions and the values we assign them. When a person fails to use available "temporal information" I call these distortions in prediction and value, "temporal blindness." When an entire society is time blind, and we try to fix social problems by changing the temporal cognitive abilities of all individuals in the next generation, I call the activity "cognition based solutions to global problems."

In the first book I describe the "time blind problem" for the lay reader. I show that our weak abilities to gather, process, and value information, distort our expectations. These distortions cause people to smoke, skid off roads, not wear seat belts, and contribute to global problems. The book's conclusion:

The text is not designed to upgrade the reader’s temporal thinking abilities. The text is designed to get each reader to acknowledge: 1) their own thinking limitations; 2) the implications of these limitations in terms of their behavior’s impact on future conditions; and 3) the utility of creating new "thinking development environments" that prevent these limitations from being a part of a future generation’s cognitive tools.

In the second book I outline a solution to the time blind problem. I connect "behavior selection" to "thought processes," and "thought processes" to "learning environments." I show the ability, to predict and value outcomes of behavior partly results from nurture and partly from nature. Some of our global problems occur because "nurtured cognitive abilities" produce predictions or values that are too weak to compete with those produced by our animal nature. Or they are too weak to compete with the incorrect predictions, and shortsighted values we obtained through rote learning.

I show, using graphs of information flow, how existing development environments unintentionally induce temporal blindness. I hypothesize how alternative learning environments may prevent it. The conclusion:

(physics of injury" note)

During a 30-MPH crash both the car and the child must abruptly stop moving. The stop is like landing on the pavement after jumping off a 3rd story balcony. If any of the parents had a choice between landing on a thick cushion or a thin one, they would all pick the thick one. However, the "pull back behavior" is like picking the thin one. "Not holding back" is like picking the thick one.

In such a crash, the car’s front-end crushes 15-inches. This crush acts like a 15 inch cushion for the dashboard and everything that slows down with it. If the child is against the dashboard at the instant of crash, he or she slows down in 15 inches. The dashboard applies a 500-pound force to the child’s body. It seems large. However it causes no severe injury. Seat belts prevent injury because they act like the dashboard. They apply the 500 pound force for the 15-inch stopping distance.

No parent however, can create 500 pounds of restrain with one arm outstretched to the right. The child’s body overpowers the arm and continues moving forward at 30 MPH. Unfortunately, by the time the child moves from the seat to the dash, the dash has already moved 15 inches and has already slowed to zero MPH. When the child collides with it, he or she slows to zero MPH in the 1 inch or so the dash deforms. What could have been a 15-inch cushion is now little more than a one inch cushion. The forces on the child are ten times higher and so are the injuries.

---> This analysis is confirmed by crash data of unbelted car occupants. Sleeping passengers, drunks, and un restrained children, that slid forward during braking, and were "on-the-dashboard" at the time of collision, walk away from accidents only slightly injured. Passengers that hold themselves, or are held like children, away from the dashboard during braking, and then collide with it during collision, get seriously injured or killed.

Part 1 - What is temporal blindness?

In 1968 my first engineering job was at the Safety Research and Development Laboratory of General Motors. My task was to find ways to reduce occupant injury during automobile crashes.

Engineers ran hundreds of tests. Special movie cameras captured crash dummies colliding with steering columns, instrument panels, and windshields. Or better: not colliding, because of seat belts.

Replaying these movies in slow motion and comparing the impact forces imposed on belted and non-belted dummies, showed that seat belts prevented injuries. Case closed. With public dissemination of crash test findings, everyone would buckle up. Tens of thousands of lives would be saved annually. And I would be out of a job within a matter of weeks.

Unfortunately, it didn't work that way. Inconceivable as it seems, few people in 1968 were willing to wear seat belts no matter what information or movies were available for view.

People chose not to wear seat belts. Sometimes they would go to great lengths to defeat the buzzers, lights, and ignition interlocks that coerced them to wear them.

Why did drivers and passengers choose the dangerous over the safe behavior? Why did the enormous injury reduction provided by seat belt wearing seem unimportant? Why did normal thinking lead to perverse behavior?

This was just the first of many discoveries where a chosen behavior reflected the inability to see or consistently value the predictions made possible by available information. Each distortion of human thinking by itself meant little. However, these views collectively describe a cognitive limitation I call temporal blindness.

Chapter 1. Distorted expectations

Behavior results from expectations of future condition. When individuals are unable or unmotivated to correctly use available information to make predictions, the resulting strange expectations and behavior become starting points to describe our temporal blindness.

1.1. Poor use of data

Long before seat belts first appeared in cars, newspapers reported "weekend state highway death tolls" like baseball scores – in a box in the top right corner of the front page. In the 60's, about 55,000 people were killed on U.S. highways each year. 350,000 were maimed. The numbers may seem impersonal. However, everyone knew of someone that had been killed during the previous year. Everyone knew someone who was severely injured in the last two months. Some were bad drivers. Some were hit by bad drivers. Some were drunks. Some were hit by drunks. The data’s implication was clear. Everyone exposed themselves to injury and death each time they went for a ride.

The data about the injury reduction resulting from wearing seat belts was just as clear. Chances of death over a lifetime were reduced from: 1 in 150 to: 1 in 25,000. Even if people did not understand the difference between these two probabilities, it was explained to them in simple terms. Seat belts were a wonder drug against injury – as powerful as the polio vaccine. The GM accident/injury database contained almost no fatalities for passengers in the belted category.

This data was not kept a secret. It was as impossible to miss the intense dissemination of this information as it is for us today to miss the Surgeon General’s notices on cigarette packs. Yet, while seat belts had been installed in millions of cars and tens of millions of Americans could have chosen to wear them, they didn’t buckle up.

Something was amiss in their thinking process. Something in the individual’s thinking discounted the data’s inference enough that accidents and injury did not appear in the individual’s expected future. Instead each person held two fantasies: "I won t have an accident" and "If I am in an accident I won’t be injured"

1.2. Partial word meaning

How did the individual find it so easy to exclude himself or herself from a danger the data said everyone else faced? There are many answers and they fill volumes. To this ever growing body of knowledge about how or why we think so poorly, I want to add another hypothesis. A potential seatbelt wearer misused the available information because it contained temporal content that was not fully understood and valued. It contained content that did not influence expectations.

This hypothesis is my starting point for describing temporal blindness. Consider a logical argument for wearing a seat belt. Then consider how the argument is weakened when temporal content is misunderstood and undervalued.

The terms we misunderstand or partially value are: sequence, preceding, motion, caused, time, next, change, and expected. The terms seem simple enough. We feel we know what they mean. However, we the temporally blind are like the colorblind person saying he can see the difference between colors when all he sees is differences in grays. We understand and value only a paltry subset of each word's temporal meaning. The reason our "invincibility fantasies" have not been expunged is that arguments made with these partial word meanings are not strong enough to do it.

This lack of word definition also prevents the creation of concepts that would rely on word constructions. For example, when a decision-maker does not see life as a sequence of connected conditions. When a decision maker does not see one set of conditions as the basis for the next. When motions do not have to transpire to change one condition into another, then, future conditions predicted by motions (rates of change and time durations) appear no more compelling than alternative future conditions predicted by fantasies. Also when future events are disconnected from physical causes; when, physical behavior has no role in creating or preventing those events, then, there is no need to choose behavior to alter them.

These generalizations when applied to the auto-crash-injury environment mean, when accidents are not related to car speed or mass of the object impacted. When, injuries are not related to collisions with the instrument panel, windshield or steering wheel. When, the wearing of seatbelts doesn’t have any effect on expected injury, there is no motivation to buckle up. If injury events are random, then not wearing a seatbelt can be OK'd by a belief in one's own invincibility.

1.3. Perception of physical motion

When an automobile crashes and people are injured, there are actually two collisions. The first is between the car and the road obstacle. The second is between the occupant and the car's interior. The first collision everyone understands correctly, the car stops. The second collision, hidden within the simple view of the first, happens when the passengers inside the car, still traveling at the original speed, run into the stopped dashboard, steering wheel and windshield. We saw in the introduction the second collision produces the injury. Being up against the dash or buckled-in prevents the second collision. It follows, that buckling a seat belt, relies heavily on understanding that the second collision creates the serious injury and bucking up prevents the second collision.

Getting a view of this second collision is not easy. The initial pre accident view "cruising down the road," and the post accident views (bent sheet metal, broken glass, and human injury) do not describe two collisions. The second collision can be teased from information three ways. However, none of the ways are easy. First viewing movies taken by auto safety engineers show second collisions. However, these views are not are not easily connected to the initial or final views normal people have of the accident.

Second, the laws of physics can convert the initial and final speeds and geometry into views of the second collision. However, most of us are not physicists and even if we were, the existing pre and post accident information does not motivate even physicists to investigate and use their tools to discover the second collision.

Third, our physiology understands some collisions and injury. For example we all duck fly balls. However, our physiology can not help us understand and respond to the second collisions in auto accidents. Human physiology can not obtain a view of the auto crash as a two collisions because too much change happens in too short a period of time. Both collisions are over in the pop of two fingers. Our physiology can not capture and process information fast enough.

To describe our temporal blinds we will eventually expand all three. Here I will limit the discussion to just physiological reasons.

Our physiology can not convert feelings of motion before the accident to into potential for injury. A crash at 50 mph is 100 times as violent as a crash at five. Yet a car going 5 MPH "feels only slightly different that a car going 50MPH. People can feel that it takes more time to speed up to 50 miles per hour than 5 MPH. However, a blindfolded passenger can not tell the difference between 5 mph on a bumpy road or 50 mph on a smooth road. Habituation soon dissolves differences initially sensed in wind and tire noise. The blur of the pavement just below the side window at 10 mph can not be distinguish from a blur at 50. The speed of a cruising commercial jet feels about the same as a car at 50mph yet the speed is 8 times faster and a collision at that speed is 100 times more violent. Changing lanes in your car at 50 MPH produces the same feelings of motion as swinging on a porch swing. No wonder no one wanted to wear seat belts in 1968, they were visualizing themselves wearing seat belts on the porch swing.

Our physiology can not covert the concept of impact with the car's interior into the correct protection or avoidance behavior. When a ball is seen coming toward one's face most people duck or put up a hand to deflect the impact. How that head ducks, or how that hand gets moved to the right place and at the right time to catch or deflect the ball is not a conscious computation. The computations are done at the physiological level. However, these computations are not universally correct. Physiological computation also causes drivers and passengers to "stiff arm the instrument panel" during braking before crashing and this increases their injuries ten fold.

Why the difference in capacity? In the ball case the ball’s approach could be sensed by the body’s physiology. The "ball’s approach," approximated hundreds if not thousands of previous experiences. Using these present and historic data, physiological computation can correctly determine the impact location, the damage, and even the ducking or blocking behavior.

When a car accident occurs there is no previous data of the human impact and injury part of the event. Even after an accident, if the physiology made a record of the events sequence it would be only a single instance. This limited data would not allow computation of a protecting behavior like buckling up. To generalize the limitations of our physiology, it will not work went there are two few experiences of the event to be felt, or when the event happens too fast or too slow to capture the data by direct sensation.

1.4. Even true movies produce fantasies

In 1968, I though we circumvented the thinking limitations caused by our physiology, our lack of temporal meaning for words, and our inability to appreciate compiled data. I thought the super slow motion movies of an unbelted dummy using his head as a sledge hammer on the dash board, held in contrast to no impact for the belted dummy, provided a visual and conceptual route to seat belt wearing. What surprised me was that when these movies were disseminated they did not significantly influence seat belt behavior.

Psychologists were not surprised. They knew a viewer of a violent adventure movie, sees no connection between the movie’s murdered victims and themselves. The psychologist would say the viewer of the dummies has no belief that he or she will be in the accident predicament of the dummy any more than he or she sees himself or herself in the same predicament as the murdered movie victim.

The viewer has two automatic mechanisms that shielded him or her from association with the dummy’s condition. The first is that the dummies are like actors. We are accustomed to believing that actors have only "make believe" injuries and pain.

The second is the belief that the character that got shoot in the movie was in the bad part of the town. A part of town the viewer would never visit. It follows that the viewer of the crash movie believes that while the dummy was in the crash environment the viewer will never be in that crash environment. Therefore he or she doesn’t have exposure that needs to be mitigated by wearing a seat belt.

1.5. Distributed exposure

"Behaving differently than our fantasies dictate" also relates to an inability to understand how the passage of time should affect the certainty of a prediction. For example, if you are playing Russian roulette and there is one bullet in the six chambers of a six-shooter, then your chances of not shooting yourself if you play "once" are 5 chances in 6 or 83 percent. However, if you play once everyday for 25 days your chances of not shooting yourself drop to less than half a percent.

Think of it this way, if you play with 100 friends on the first day 83 live to play the second day. On that day 83 percent of the 83 people live to play the third day. Then 83% of those left living on the third day... etc. By day 25, most likely none of the original 100 friends will be left to play. Time (number of days of exposure) changes the probability of life from 83% to .5 %. However if you handle the temporal aspects of expectation poorly you don’t see it.

If we review how people see death in car accidents we see the same analysis. The probability that you die in a car accident on any trip is about 1 in 8 million. This seems like a very small chance for you to take. Why buckle up? However, if you expose yourself to this probability 100,000 times (the number of car trips per lifetime) your chances of being killed are one in 150. Your chances of being injured (6-8 times higher than being killed) are 1/25.

It is only after you include the temporal aspects of the environment that the data argument to wear seat belts acquires the strength to overcome fantasies of immortality.

1.6. Integration fialures

Each of us in 1968 had several views of the accident environment; driving experiences with near misses; personal experience of accidents or injuries; crash tests movies of unbelted dummies hitting the dash board; and ever refreshed views of other people’s accidents and injuries. Taken singularly each did not influence seat belt wearing.

More surprising was that the group working together did not either. I suggest that the temporal aspects of each prevented their integration. To connect these views together to make a stronger influence than any one image could produce singularly, requires an additional cognitive ability we still don’t have. We are missing an ability that takes the separate images and visualizes them as a part of the same sequence.

"Sequence," that’s one of the words for which we have only partial meaning. For example, when we do get the entire sequence, as we do with fly balls, we do choose the correct ducking behavior. However, this vision of sequence was built by our physiological capabilities. In the auto accident/injury environment this physiology can not provide sequence. Sequence must be provided abstractly. To integrate the accident environment's independent views, an object’s mass and motion must be visualized as part of a system which obeys the laws of nature. The object’s motion must be visualized as continuous. The object previous condition must be seen as the basis for the next. A behavior must be seen as something that transforms one prediction into another. Behavior musts be seen as an additional force working (during a time interval) to cause this transformation.

A long as the accidents reported in the newspaper are not connected, for example, to initial car speed, or as long as injuries are not connected to human movements similar to those seen in movies of crash dummies, we should not expect any internal motivation on the part of an individual to wear seatbelts. Today seatbelt wearing is not influenced by the abstract visualization of complex motion. Instead seat belt wearing is culturally coerced. Without our culture we still would not wear them.

1.7. Direct experience failures

What we learn from experience does not always accurately predict the future.

Repetitive experience is our most powerful teacher. Most of us buckle up because of habit. However, before this habit is in place. During the habit learning period, part of the experience forms a logical path to not wearing a seatbelt. From the many repetitions of putting on a seat belt and not being in an accident, we learn that the belt buckling causes delays, wrinkled cloths, chafed necks, and feelings of confinement. The absence of an "experienced-accident" prevents learning that a seatbelt could also produce injury reduction.

How would you know when an experience produced spurious learning? How would you know that an experience happened too fast for you to register the sequence of crash events? How would you create a record of what transpired?

With no logical path to create understanding of cause and effect, there is no way to determine that the absence of the seatbelt causes injury and that the wearing of the seatbelt prevents injury. The more common "learning" resulting from an accident experience is "being in an accident is a random event." "Being injured is random."

These examples describe domains where our experiential capacities can not learn and thus describe part of our temporal blindness.

It is a good story and maybe it even happened once. However, it does not match the facts. Being thrown out almost always guarantees serious injury. Getting thrown out of a moving car is the same as jumping off a bicycle at the same speed. (Except cars are bigger and crush you when they roll over on you.) Individuals who were thrown out during accidents accounted for about half of all traffic fatalities in 1960. Car fires happened only once in every 5,000 injury-accidents. Which condition do you prefer, taking your chances getting thrown out, or taking your chances getting burned to death in a car fire

We see the same misuse of information when the decision-maker decides not to wear a seat belt because of a story where a belted passenger was crushed to death between two surfaces of the car’s interior. Again, the incorrect conclusion is that the passenger would have survived had he not been belted into the crushed spot. Without the belt he or she would have moved away from the crushed area of the interior. The facts: for every accident with a crushing injury, there are 50,000 where the belts reduced injury.

So why are these stories so influential in the seatbelt decision? You guessed it, "temporal aspects." If the true images of unbelted people bouncing around inside cars and getting injured during accidents, or the true images of belted individuals not being injured in accidents, remain invisible, all that is left visible is the silly mythology offered by the officer.

With our temporal blindness we don't see the benefits for wearing seatbelts. We can not convert post accident images of bandaged, bruised, scarred, hobbled, wheel chaired, or buried bodies into "caused by not wearing seat belts." In this vacuum, the police officer’s story is most convincing and influential.

Temporal distortions in expectations from the misuse of information extends beyond the auto safety world. I reviewed injury data related to the broad spectrum of human activity and realized that people cannot tell the difference between safe and unsafe activities.

This discovery got a few engineers at the Safety Laboratory to run an informal experiment. We wanted to see what people thought were dangerous activities, and what they thought were safe. Maybe then we could explain why people would not wear seat belts.

We made a list of common weekend activities:

• Baking a cake:
• Mowing the lawn.
• Washing a second story window.
• Fertilizing the garden.
• Splitting firewood.
• Rotor tilling the garden.
• Taking out the garbage.
• Using power tools in a wood shop.
• Changing a light bulb on a ladder.
• Driving to the grocery store,
• and 20 more items.

We put our friends, colleagues, and neighbors to the test, and asked them to rank the items according to relative danger. They were to put a "1" next to the most dangerous class of activities and a "30" next to the safest class of activities.

This survey experiment would never withstand rigorous scientific scrutiny. Almost everyone knew we were auto safety engineers and should have been biased to raise "driving to the store" higher on the list than they thought it should be and maybe they did. But our point was made loud and clear in spite of this, because car driving, the most dangerous item, never made it near the top of anyone’s list. That’s right, driving to the grocery store was far more likely to cause injury than was anything else on the list.

Why did rotor-tilling the garden appear to most people more dangerous than driving to the grocery store? Maybe because the rotor tiller is a noisy growling gasoline powered machine, which jumps and bumps along as its mix master blades claw and churn the soil. Certainly not because the little guy has 1/50 of the horsepower, weight and speed of your car. Not because there are many fewer and less severe injuries derived from rotor tilling than car driving.

The answer is related to the temporal aspects of the two experiences. You drive your car several times every day. It is your faithful servant. The daily experience habituates any sense of danger. The routine creates a belief that injuries never happen and car driving is safe.

Rotor tillers come to the front of conscious but once a year. Each spring just the thought of the whirling blades creates images of being eaten or maimed. Just starting the thing takes an extra portion of breakfast courage. This fear, even though the potential for injury is much smaller, is never reduced by habituation and thus roto tilling was ranked higher (more dangerous) on our test.

1.8. Extending experience failures

Experiences are sequences of conditions. We remember condition "A" was followed by condition "B." This memory allows us to predict "B" the next time we see "A." We also remember that condition "A" acted upon by behavior "X" produces condition "C" instead of "B." This allows us to predict a difference in outcome that depends on our behavior.

We can also guess that a smaller behavior than "X" will get a future condition with less change than "C." A slightly larger behavior than "X" will produce more change. These predicted, yet non experienced conditions, I call extensions of our experience.

When the behavior is tried and the expected condition does not occur it shows us that our thinking processes distorted the prediction and thus the expectation. In the dynamic systems presented next these distortions are so consistent they describe temporal blindness.

When I was a small child there where no seat belts in cars. I liked to ride standing with my chest and my chin on the instrument panel. I was extending my experiences where braking threw me into the panel. I learned that it hurt more when I was farther away and less when I was closer.

My father and mother had another view. They made me sit far away from the panel on the seat next to them. That way they could hold me back when they had to step hard on the brakes. My parents were extending their experience. They had the leg strength and arm reach to keep themselves in place on the seat during extreme braking. They learned to hold themselves back. And they extended "holding themselves back" to "holding me back."

Each of us, looking at the same environment, saw different things then extended them to different predictions, expectations and behavior. If this could happen, there is a good chance there were additional parts of the existing environment that neither of us saw and extensions that neither of us made. For example, neither of us could or did extend our experience with coasting and braking stopping distances to the very short stopping distances that occur during a crash. Neither of us extended our experiences of forces created by coasting stops, and braking stops, to predictions of very high forces during a crash. If we had, our expectations would have had us both diving for contact with the dash board just before a crash.

Let me describe the kind of thinking that might have transpired. Coasting stops, hard braking stops, and crash stops are different stopping distances. For a car going 30 mph, coasting to a stop could take a 1000 feet. Hard breaking a car stops it in 50 to 60 feet. Crashing stops a car in one or two feet. These are not just instances. Stopping distance is a continuous range from 1000 feet to 1 foot. Extension processes, if they are working properly, should be able to take the cases that have been experienced and extend them to cases that have not. This is partially true. Even though we have not experienced every braking case with every object on the seat , we know when to put out our hand to hold objects back and when not to worry about them.

However, there are limits to this extension processes. For stopping distances during crashes(those between 49 feet and 1 foot) the extension processes does not produce correct predictions or expectations for the forces. That's why we hold ourselves and our kids back just before crashes.

In the next section, I show how our physiological thought process makes these incorrect extensions. I show that our physiology plays a pivotal role in our temporal blindness. However, that discussion depends on an understanding of the information that is available but not used.

To create and present this information I rely on a law of nature that connects stopping distance to force needed to impose the stop. By connected, I mean that if one variable is known the other is determined by law. A simple example of such a relationship (law) is included in the example below.

In the stopping distance/stopping force law, one variable goes up the other must go down. From the table in Figure 1.8-6 you can see that the product of the stopping distance in feet and the stopping force in g's is equal to a constant. In this case the constant is "32." Choosing a many stopping distances and calculating many stopping forces in the g's produces the graph.

Normal driving and hard breaking experiences of stopping forces and distances are depicted by the far right portion of the line and the line's extension off the right side of the page. For most of the line in the figure, which depict stopping distance and forces that occur only in crashes, most people have little or no direct experience. The relationship between the experienced and non experienced portion of the line is useful for describing the limitations of our mental process that perform extension on our experiences.

For example, when you slam on the brakes and the car stops in 64 feet everything in the car experiences (from the table) .5 g's. To stop a 200-pound man requires (200 times .5) a 100 pound force. To stop 100 pound woman requires (200 times .5) a 50 pound force.

This means, when slamming on the brakes the 200 pound man, between his feet and his hands has to produce 100 pounds of restraint to stay on the seat. This is not a big number. When the 200 pound man is standing up, his legs have to create twice that force just to keep him from falling down. However, during a crash into a tree (a stop that brings the car to a halt in two feet (see table) the force in g's is 16. He will have to produce 16 times his weight (16 times 200) or 3200 pounds of restraint to stay on the seat. A seat belt can do that but he, with just his arms and legs, can not.

When a driver tries to hold himself or herself on the seat in a crash environment he or she has incorrectly extended his daily driving stopping experience (those on the far right portion of the graph) to the crash situations described on the left.

In summary, the driver’s view of this "stopping distance/stopping force problem is so partial, it has been extended over such a small part of the range of stopping distances, it explains why our parents learned to put out their arm to protect us in both heavy braking where it worked and in crashing where it didn’t.

–––––

At the proving grounds I did not interpret this discovery as grounds for rewriting the driver-training manual. Instead I realized some additional general descriptors of our human thinking.

A full description of why we do this is the discussion of the second book in this series. In this book a brief description of limits in our extension processes is useful as a means for describing our temporal blindness.

Next I show that total reliance on our physiology to perform the extensions that help us choose behavior works only in a part of the domain in which we are immersed. Outside of this part, the choice of good behavior requires manipulation of abstractions. If we can not find and manipulate these abstractions we can not keep ourselves from taking tragic behaviors.

That people can not physiology extend their "stopping experiences" across the entire stopping domain to which they were exposed, was just the first of many examples I observed at the proving grounds. The next example, while a little long and complicated, is included because it was pivotal in helping me describe temporal blindness. Your attention to its mechanical and sometimes abstract issues will be rewarded beyond learning something about your temporal sight. You will learn how to drive your car when it is skidding out of control. I will teach you why having your graduation tassel hanging from your rear view mirror makes managing skids much more difficult.

A skidding car

In normal everyday driving, we place ourselves in serious jeopardy. Sometimes we know it and sometimes we don’t. On icy roads we know it. This knowledge makes us focus on learning the special behavior required to not skid into things on the road or into the ditch. Our learning processes determine how much we do learn – and how much we leave unlearned. They describe how we gather information, how we assemble it, and how we arrive at behavior. When that behavior does not implement our expectations, the processes describe our temporal blindness.

At GM, for a short while, I was part of a group that taught performance drivers how to drive safely at high speeds. Drivers normally learn through practice. Each driver experiences the unsafe condition over and over until he or she learns to control it. "Skidding around" is the process by which they learned how to manipulate the gas pedal, steering wheel, and brakes to control the car.

The learning process is little different than that used by hockey moms to manage the skids on a snow day. Except the GM driver’s exercises included more vehicles, road surfaces, speeds, etc. And the GM drivers had special roads with wide grass lawns on each side. If the GM driver lost control, he or she took an unexpected excursion onto the grass. Cars and drivers generally made it through without a scratch. This excursion injured only the driver’s pride.

These special roads had their limitations. Each modest increment in speed required vastly greater lawns to accommodate the car’s wild antics. As testing speeds increased even the shoulders of the GM’s Milford proving grounds could not be made any wider. Suddenly the training used for 40 years did not seem adequate for the powerful sedans of the 60's.

So a group of engineers began to design a new training program. One that would make it possible for performance drivers to handle emergency conditions for the higher speed driving conditions without direct road experience. That is, without learning to drive the event by trial and error; or as they called it "by the seat of their pants." The exercise was an engineering challenge and it contributed to my understanding of temporal blindness.

We began by reviewing in detail the "seat of their pants" practice sessions. The only mental skill required was the motivation to do it again and again, until senses and motor coordination got unconsciously connected. An aspect of this learning process, possibly one I fully understood only after 20 years of learning research, was that it allowed the drivers to successfully controlled skids for only those cases directly mastered in practice. If the car design, road surface, speed, or severity of the initial skid deviated more than a small percentage from the cases they had mastered, the behaviors they learned would not help recover from the skid. For example, for the same skid in a car with different power steering than the one they had practiced, performance drivers and soccer moms respectively turned the steering wheel in what they thought were the right directions and timing, and it did not keep them on the road.

What each had learned about skid control was not the whole story. The behaviors did not recover from every skid into which they could drive. Their skid recovery training produced behaviors that were a lot like the behaviors of individuals who learned from sliding grocery bags to put their arms out to reduce child injuries. It worked for the skids they practiced but not much more.

Examination of these learning limitations describe several additional parts of our temporal blindness. However, for you to understand, I will have to explain a little more about, skidding cars, behaviors that respond to the skids, and ways to learn those behaviors.

First let me describe a way of choosing behavior called "pattern matching." In pattern matching a car’s present skidding condition is subconsciously connected to a memory of a skidding condition experienced in the past. This memory in turn, is subconsciously linked to a memory of the executed behavior that resolved that skid.

Pattern matching finds successful behavior for any well practiced skids. It fails when we get into a skid which we have not practiced. And as I have explained above each of us has practiced skid control in only a small portion of the situations we may experience in the real world. For example, recovering from most emergency skids is beyond the little slipping and sliding to the grocery store stuff most of us learned.

To choose behaviors for a non-experienced skid, at least from our direct experience, requires that our memories be interpolated or extrapolated. These processes are not difficult to perform on a 2 variables system like the "Give /Eat cookies" problem (see footnote previous section.) If you eat two and half cookies it is not difficult to calculate that you can give away "a half." However, in the skidding problem where there are many variables, when they change continuously and independently, as the timing of a behavior becomes as important as its direction and magnitude, the extrapolation or interpolation process becomes complex if not impossible to perform. Physiological extrapolation or interpolation manages skids that are only slightly different then those we practiced. For skids more than slightly different and our behaviors exacerbate them.

Our research found that controlling any skid, requires a driver to compute a series of behaviors from unfolding information. If direct pattern matching could not perform this task. If physiological extrapolation or interpolation, could not perform this task, maybe there were some things we can teach (transmit) that will help drivers control skids.

To understand how hard this transmission solution is consider the following. You may be able to ride a bike, however, if you learned to ride it subconsciously (we all did) you have no idea what physical world variables you are measuring or how you are manipulating them to produce the controlling behaviors that keep the bike from falling over. This is painfully clear when you try and help your child get started on his two-wheeler. What meaningless gibberish do you yell from a distance when your child is headed toward the play ground flag pole, "turn!" "lean!" …? Those of us that have had this experience know it’s a terrible feeling even if we do not understand the learning theory.

It is the same for teaching skidding. Experienced drivers could not sit with skidding student drivers to tell them what to do next. At the conscious level, the experienced drivers did not know what they measured or how they used those measurements to produce appropriately scaled and timed behaviors. And even if they could correctly find ways to convert their knowledge into sentences, the student driver, normally operating under a physiologically driven experiential learning process, had no way to use it. For both instructor and student it was an impossible situation.

If all the processes that we have in learning tool box, pattern matching, physiological extensions of direct experience, and transmission could not accomplish the task of teaching skid control, we had to search for a new process. Next we considered a process that allows the driver to consciously compute and execute behavior by manipulating an abstract mechanism. The mechanism describes the car/driver system. It consists of connected variables whose values can be extracted real time from the unfolding skid information.

This alternative behavior creation process is far different from those we all so easily accomplish. Creating a manipulatable abstraction, a mechanism of connected variables, is conscious work. It is not easy work. Physiological sensitivity to mechanistic relationship is not something natural selection has treated kindly. Most people can’t do it.

No trainer wants want to consider this alternative as the only way to attain safe drivers. Yet with the physiology based learning models unusable. The transmission model, based on the experience of gifted performance drivers, unusable, GM had no other choice. The alternative meant GM was going to send performance drivers out to drive in domains for which they had no experience and for which they could not be trained. Forty years into a driver training program, this was not great news. However, the revelation motivated the launching of a very intensive effort to find a better way to train drivers, so their driving skills would extend over the range of circumstances to which they were exposed. (Read this paragraph four times. It is an analogy for "cognition based solutions to global problems." It is probably the most important paragraph in the whole book.)

To design a driver training program I began by looking analytically at simple skids. For example, consider a car traveling in the right lane of a straight 4–lane divided road Figure 1.8-10. A deer jumps out from the shoulder and blocks the right lane. The driver has just enough time to steer the car to the left of the deer.

The car is now pointed, and is traveling, straight down the center of the left lane as shown in both the figure above and in the figure below. However, it is rotating clockwise. Within a second, Figure 1.8–20, while the center of the vehicle moves neither right nor left in the lane, the rear moves left and the front moves right.

We all know what to do in this situation. It’s the same as normal driving. We turn left to get the car to point down the road.

We also know that if we turn too little – or too late – the clockwise motion will continue. The rear of the car will pass the front on the left side, and the car most lightly will end up in the right ditch as shown in Figure 1.8-30.

We also know from experience that if the steering wheel is turned to the left too much – or the correction is held too long – the clockwise rotation will be replaced by an even larger counterclockwise rotation. The car will go from pointing right to pointing so much to the left that the rear of the car will pass the front end on the right side. The car will end up in the left ditch. Thus, in the skidding condition that immediately follows missing the deer, too much or too little steer – or taking out the steer too early or too late – will cause loss of control.

This is where the "skidding around" learning environment is a good teacher. Drivers learn through practice precisely the amount of turn and the timing required to prevent loss of control. Trial and error experience allows the body to subconsciously measure all the needed variables and find timely behaviors that allow the car to get "straightened up" with the road.

However, this learning design could not be used on public roads. It could not be used even be used on GM roads. Even with wide grass lawns they were not big enough. In the new learning design, control behaviors would have to be created without previous experience. Behaviors would have to be created from information consciously gathered and consciously manipulated as the skid unfolded.

Not relying on subconscious processing, imposes two constraints on the new training design. First at a conscious level a driver can measure little more than one or two variables. Second, at the conscious level calculation could be performed on little more than one or two variables. Within these limitations we set to work to find: which variable provided the most important information about car skids, and what relationship could convert this measurement to performable and successful behaviors.

The critical variable to measure was found to be "heading movement"– this can be determined by the horizon movement (right or left) in the windshield frame. This discovery is within the capacity of a high school physics student. It is found in Appendix A: "Finding the Skid Indicator." However, for our search for understanding about our temporal blindness the derivation is not important. What is important is that the variable "heading movement" or more specially "not finding and understanding the variable," demonstrates a blind side to our cognitive process. It is not a small mistake. It is responsible for hundreds of thousands of lives in the past and untold lives in the future.

To understand the concept of "heading movement" paint a black arrow on the car’s hood with the tail near the steering wheel and the head straight forward the length of the hood as show in Figure 1.8-40.

Assume the steering wheel is straight and the arrow points at a mountain peak. If after an instant of driving the arrow points at the same mountain peak then heading movement is zero. If instead the arrow points to the right of the mountain peak, this heading movement would indicate a clockwise skid. This measurement can be made without any road reference. All that is needed is a reference point somewhere on the horizon. It does not have to be at the end of the arrow just that the end of the arrow moves toward or away from the point.

Thus skid control, at least the hard part of skid control, is to stop the heading movement. Getting realigned with the road after the car stops spinning is a secondary and much simpler problem every driver can solve.

The table in Figure 1.8 - 50 summarizes what to do with the steering wheel for each detection of undesired rotation. If the car is rotating clockwise – turn left. If the car is rotating counter clockwise – turn right. If there is no rotation (no heading movement), return the steering wheel so that the front wheels are pointed in the direction of the headlights.

Even if the headlights are pointing off the road when the heading movement stops, that is the car is sliding down the road sideways, take the steer out. You can visualize this in the figure above.

You are probably asking yourself two questions about controlling car skids. If the car is not rotating but is sliding down the road "sideways," and the front wheels are pointed off the road, what behavior realigns the car with the road? This is simple. You steer in the direction you want to go. If the road is to the left – you steer to the left. Slowly and carefully to get the car to go there with out creating any large buildup in counter clockwise momentum. Have confidence you will be able to master this part of skid recovery with no additional training or practice.

The second question is "How much steer to put in?" The amount to steer (proved in Appendix "A")is:

such that when there is no heading movement, the steer can be removed to allow the front wheels to be pointed in the direction of the headlights."

The table in Figure 1.8-50 when combined with the "Turn the steering wheel as much as possible rule" seems too simple a solution to control car skids. How does a three condition three behavior table do the work of sensing all those variables and computing the magnitude and timing of all those behaviors. The proof, besides that contained in physics, is that the table works extremely well in practice. It is very easy to learn. Timid drivers can learn it and perform it just as well as hot shot performance drivers, Using it, both groups can handle in skids neither has previously experienced. Furthermore the table solution is very robust. It resolves the skidding vehicle problem for busses and go-karts, on icy roads and dry pavement. It works for high speeds and low speeds, and little skids and big skids. It works for airplanes, space shuttles, submarines and oil tankers.

However, another useful aspect of the successful use of this table, is that it contributes a critical piece to my description of temporal blindness. It illuminated the difference between two pieces of information. One, the "misalignment with the road," and the second "heading movement." Misalignment with the road, most of us understand and can use to steer our vehicles "Heading movement," few if any of us use because its power to manage the skid control, remains invisible at the conscious level. My proof of this conjecture is that when we are in an unpracticed skid, we use the first variable and not the second almost every time.

Short course in skid control

Describing our temporal blindness is the goal of this book. I am less interested in teaching you new ways to learn to control car skids. However, I promised that because you concentrated on the primary task there would be a reward in the form of a short course in skid control. (If you are not interested just skip forward a few pages to Section 1.9)

What follows is the remaining portion of a skid recovery course. It develops three supporting skills that allow a driver to implement the behaviors dictated by the table in Figure 1.8-50:

To determine heading movement, a person simply watches the horizon and determines if it is moving right or left. Optimally I have suggested a long stripe painted on the hood of the vehicle, from steering wheel to a spot left of the hood ornament.. If you look at pictures of old race cars (when the driver had to sit right or left of the engine and transmission) you will find they where painted that way. However, most family sedans lack such a stripe. GM if your listening 32 years later you can still follow up on my recommendation.

Lacking such a paint job, drivers might use any two points fixed relative to the centerline of the car. The two points should make the longest line possible and be close to the line between the driver’s eye and the horizon maker.

Without the special paint jobs, one unconscious line selection is accomplished by assuming that the driver’s head is fixed relative to the car’s centerline and the far end of the point is the hood ornament. With down sloping of hoods higher seated positions and the removal of hood ornaments this line was removed from the choices. Drivers were forced to use a dirt spot on the windshield or the edge of the rear view mirror. These choices make the detection of heading movement weaker than a hood stripe first because the line is shorter. Furthermore the eye socket of the driver is really not fixed relative to the vehicle's centerline. In an emergency maneuver, the head flops around like a jack in the box creating all kinds of car "heading movement" indications that do not exist.

The worst but most obvious choice of front reference point is the thing hanging from the rearview mirror, for example a graduation tassel. It’s better than a point in that it is a vertical line. It cuts through the horizon no mater what its elevation in the field of view. At slow speeds, when it hangs straight down and you are making regular turns it is quite useful in smoothing out your driving. The fact that it moves like a crazy pendulum when the car is skidding means that both ends of the reference line (your eye and the tassel) are moving relative to the car’s centerline and not necessarily in the same direction.

Heading movement, if it is to be useful in vehicle control must be delivered accurately and timely. All this secondary moving of the reference line requires so much computational correction its utility is greatly reduced.,

To summarize; skid control behavior depends on determining "heading movement" several times a second. This is done using a reference a long distance from the car (e.g. mountain on the horizon) and a line fixed to the car that can be used to report small changes in the car’s heading relative to the reference. A student upgrades his or her means of determining heading movement when he or she knows to choose long lines that are stable relative to the vehicle’s center line, and near the line between the eye socket and the distant point.

Steering right or left as much as you can seems like a pretty rash thing to do. Timid drivers, and drivers that love their cars are loath to turn the steering wheel so violently. These turns make the car tip and the tires squeal. It spills your coffee and throws everything right or left. Such violent steering does not have to be learned – just justified. The justification is one paragraph long.

Assume a driver slammed on the brakes in front of you and you needed to stop to keep from running into him. You would step on the brakes as hard as you could. It would make no difference if your car was going 60 mph or 30 mph. So it is with turning the steering wheel to stop the rotational skid. The steer wheel correction should create the maximum force no mater what the size of the angle between the car and the road.

Getting the steering wheel back to center when the heading movement stops changing is a little trickier. It is easy when the car is not skidding and the direction the car is travelling in the direction the car is pointed. However, as in Figure 1.8-60, when the car is skidding sideways as the heading movement stops changing, the car is pointed off the side of the road. Then aligning the front wheels so they point in the direction of the headlights is a whole different problem.

All the experience in the world is not going to help you. The task is accomplished not by feel. The driver must count the steering wheel turns "put in" from straight ahead to the "as much as you can" counter steer. Then when the heading movement stops changing, the same number of turns can be removed without any other cues from the environment.

Let me explain one more detail about learning to control skids. Then back to temporal blindness. During the development of the course many performance drivers said they used the heading movement to choose how much and when they turned the steering wheel. It was difficult to show what they were, or were not, using because they used so many things and they were using them subconsciously. However, we could prove to them that even if they were using heading movement they were not using it to tell them when to make the front wheels parallel to the vehicles centerline. We could prove to them that they could not align the front wheels with the vehicle centerline any time the car was skidding.

To prove this to the most adamant drivers we ran the following demonstration. We had them driving very slowly in a parking lot. We had them turn the steering wheel to the right very slowly until the vehicle was making it sharpest turn to the right. While in this very slow tight turn, we asked them to stop the car and set the parking brake. After sitting there a minute we asked them to make the front wheels point straight ahead without releasing the brake (that is with out allowing the car to roll forward and get a sense of where the wheels were pointed.) This being GM, most of the students turned the wheel until the Chevrolet logo was right side up. But they still didn't know if the wheels were straight ahead, one full steering wheel turn to the right or one full steering wheel turn to the left.

• Consciously gather heading movement information.
• Turn the steering wheel counter to heading movement as quickly and as much as possible but in a manner that would allow removing all of the counter steer when the heading movement stops changing. (This occurs when the car makes its maximum angle with the road.)
• To accomplish the removal, count and memorize how many steering wheel turns were put into the counter measure and how to remove the steering input using the count as a reference.

1.9. Learning to learning failures

The table in Figure 1.8 - 50 helped resolve the driver-training problem at GM. It also helped me learn about how people handle skids. I learned about the tools they use to learn and even how they learned those tools. These "learning to learn" processes describe some additional distortions in our expectations, which in turn describe aspects of our temporal blindness.

The table in Figure 1.8 – 50 was created through a learning process that identified and manipulated symbols that represented the driver/skidding-car-system. It is a strange process that few of us have the motivation or capacity to perform. Physicists who could create the table were not motivated do so from their exposure to car skids. Even my motivation was not my well being during driving. My motivation came from being assigned an unsolved training problem.

This common limitation in people's learning to learn capabilities is the starting point of Time blind – the solution the second book in this series. Here, I have a simpler use of my discoveries. I assume that the table correctly describes the magnitude and timing of behaviors that implement skid recovery. Then, use the table as a reference by which to measure the performance of learning processes we all have and use.

Let me begin by dividing all the skids that a driver might have into those she "has practiced" and those she "has never experienced." This is a useful division because the behaviors we learn to recover from skids we have experienced, succeed (not surprising they match what is in the table.) While the behaviors that we learn to recover from skids we have not experienced, cause us to drive off the road (not surprising they don’t match the behaviors in the table.) By looking more fully at these two cases we can add to our view of temporal blindness.

When the driver experiences a skid "over and over" she learns to control it using "feel." We say the driver is using physiology based learning tools. Her physiology learns to sense and create behaviors subconsciously. When the driver gets into a previously experienced skid the correct behaviors are magically performed.

Why do I say magically? Because, the driver, while being able to perform these behaviors, can not "explicitly describe" how much and when she turns the wheel to recover from a skid. She can not describe measurements used to dictate this behavior. The driver can not describe how she "figured out the connections" between "measurement and behavior."

Even learning theorists can not explain how the connections between measurement and behavior were created during practice, and how they are retrieved and executed during a sudden emergency. And they are certainly not able to describe how the physiological learning capacity was put in place. The only thing we are sure about is that it happened long before it was used to learn to control skids.

Now lets look at the other part of the skid domain. When the driver gets into a skid which she has not practiced, her powerful physiology fails to provide any behavior. If the driver gets beyond "freezing up," her conscious learning tools have to "kick in" and create the behavior. These tools it seems produce two guides for behavior selection.

Notice that both guides use "angle between the car and road centerline" to determine behavior." This choice of variable is different from the variable "heading movement"suggested in the table in Figure 1.8 – 50. If the table suggests correct behavior and normal conscious learning suggests something else, it would appear that our conscious learning process produces incorrect behavior.

If in everyday driving, we can enter into a skid for which we have no experience, then, our lives, the lives of our family, the lives of other people on or near the road all depend on our abilities to regain control of a couple of tons of skidding steel. We don't have these abilities. The ones that we do have will choose behaviors that will probably make the skid worse.

If we knew about these limitations in our skid control capabilities, we would probably slow down. We would take driving more seriously. We would develop a much more powerful driver training course for everyone to take before they got behind the wheel of their giant SUV. We might even limit drivers without this special training to smaller vehicles, bicycles, motor scooters or mini cars.

Studying the skid control learning environment illuminates many learning activities. Some, interesting in their presence – some, interesting in their absence. These include:

• the conscious and physiological cognitive development that proceeds any skid domain learning,
• the conscious and physiological learning that happens during normal driving and skids,
• not learning the limitations of each learning process,
• taking a course in driving to learn "table" skid management,
• not learning the implications of learning "table skid management" to choose behavior relative to our natural learning,
• learning required for engineers to create the table,
• content and process that engineers learn from their activities to create the table,
• learning of "learning-theorists" about the interactions among all these levels of learning.

Learning processes produce expectations that shape behavior. If there is absence of, or a limitation in, learning processes, expectations are distorted. Distorted expectations, result in behavior that produces the unexpected.

For example, if two learning activities are operating in parallel and produce different expectations for the same behavior. At most only one can be correct. In the competition to be the guiding expectation, the more well developed learning activity wins whether it is correct or not. The weaker expectation is not believed even if it is correct. Similar distortions happen when the learning process that could produce the correct expectation is absent all together.

All of these failures were present in the car skidding problem. That is why it provides a view of undeveloped or underdeveloped learning processes and an avenue to understand our temporal blindness.

Learning processes can create a view of themselves. In learning to learn each of us asks, possibly subconsciously, questions like, "Is my learning process working properly?" "What are the components of my learning activity?" "How did I acquire these components?" Any questions that can be answered become starting places for another round of the same questions. Learning about learning is like investigating the layers of an onion. The learning we all see is preceded by learning …that is preceded by learning … that is preceded by learning, …that is …. Any weakness or limitation in a visible learning process can be explained by studying the preceding supporting processes.

Skid control learning, shows strengths and weaknesses of learning activities at many levels. We have seen that 1) some learning at some levels in some parts of the skid domain automatically produce the correct expectations. For example, our seat of the pants (physiological learning activities) work pretty well. 2) Learning at some levels in other parts of the skid domain produces incorrect expectations. For example, our conscious learning produces for all of us two rules of thumb which do us no good. And 3) some learning at some levels does not progress and does not produce expectations. For example, our abstract learning fails to make the skid recovery table.

Let me focus on learning that does not result from the skid control environment. Most of us did not understand that we had subconscious and conscious learning processes helping us choose behaviors to recover from skids. We did not realize skids fell in two domains – experienced and not experienced. We did not realize that physiological based learning worked only in the experienced domain. We did not realize that conscious based learning failed to provide correct behavior in the un-experienced domain.

Most of us never realized that our physiology was subconsciously choosing which information to collect. Or that subconsciously our physiology was creating ways to convert sensed information into correct behavior.

Most of us never realized our conscious learning process gathers the wrong information. We don’t realize that it divides the system into skids with different "angles with the road." Or that it creates incorrect ways of converting that information into behavior. We didn’t realize that when we ended up in the ditch, it was the behaviors that we choose that got us there. We do not know if some other behavior (letting go of the wheel) would have been a better solution to the skid than the behaviors we took.

By reading the previous section of this chapter, we learned there is an abstract learning capacity that can build a table to control skids. It provided us with a view that an abstract learning process produces very robust solutions to the skidding car recovery problem. The solutions are more robust than either of the two learning processes we normally use. However, we never realized that we do not have that learning capacity. Nor did we realize that "temporal abstract learning" could or should be part of any normal person’s abilities if he or she lives in our dynamic world. Even after reading this, most of us still feel these special abilities should only be the capacities of engineers.

However, engineers have all of the above temporal blindness and more. Engineers that created the table normally never learn the huge difference between the physiological and conscious learning processes. It just does not seem important to the problem they are solving.

Even the learning theorist seldom if ever learns, that learning done by engineers requires several levels of abstraction. First realizing the skidding car is not under control. Second, dividing the physical system into variables and operations. Third transforming them into symbols and functions. Fourth, connecting these into a manipulatable simulation. Fifth, "distilling generalizations" from the simulation experience that do not exceed either our physical abilities to measure sensations and manipulate the controls at the conscious level, or the physical limits of the car.

Most drivers, even after being shown the table, even after learning to use it, will not realize that the "table" and their in place "conscious learning process" create different skid recovery behaviors for the same skid. Most drivers will not recognize that both behaviors can not be correct. They will not recognize that they have to consciously make a choice as to which process they are going to use to choose behavior in the next un-experienced skid. Most drivers, without practice using the skid management table, still regress to "2 rules" or freeze when faced with a new skid.

Finally, most drivers, engineers, and even learning theorists, won't see most limitations in their learning processes. This makes it impossible for them to learn the limits of what each has learned.

In summary "distortion in expectations" arise from the presence or absence of each level of learning as well as imperfections within each level. The details learned from the skid recovery problem are so extensive they provide the foundation of the second book in this series. (For the reader who can not wait I provide a few examples in the footnotes below ^,,.)

What do you call it when drivers never realize the domain where they can control skids is partial. What do you call it when 50 years of experience driving on icy roads never develops the table solution. What do you call it when even after the table training is completed there is no view that there has to be a conscious decision as whether to use the behaviors supplied by the table or those supplied by old conscious processing that has been in place for years. It might be called temporal blindness.

When these missing views of learning are absent, there is no motivation to learn more. There is no motivation to go looking for ways to overcome these invisible limitations in our learning processes. Without this motivation learning content and process stops. Cognitive development stops.

Even thought millions of people have been injured because of these "limitations in cognition" we still do not realize, these limitations, exist. In this sense our temporal blindness shapes our behavior, prevents us from discovering that our behavior is inappropriate, and prevents us from ever creating learning tools to create the discovery.

1.10. A journey toward temporal sight ***(does this belong here)

What is wrong with our learning? What is wrong with our education? What is wrong with our judgment? What is wrong with our brains? The questions kept mounting. When people were asked to judge, they seemed to misjudge. Why did relatively safe activities look dangerous? How could relatively dangerous activities look safe? When something was obviously dangerous – what crippled thought process, what spuriously learned knowledge, allowed people to dismiss the danger?

Tens of thousands of injured and dead children each year did not make car driving dangerous. Millions of minor skids did not provide individuals either images of future crashes or the motivation to learn to control them.

My faith in the rational mind was shaken. How could people misjudge the potential danger of rotor tilling the garden? How could they believe it was more deadly than driving a car? It was inconceivable. And yet people on the street did it. Doctors, lawyers and college professors did it. Even automotive safety engineers, and professional drivers paid more attention to safety procedures during rotor tilling than while driving their cars.

Thus began a journey to discover the limitations of human thinking in the temporal domain; the limitations of temporal learning as they exist today and the study of temporal learning as it must exist in the future if we are to avoid tragedy.

I have studied temporal blindness for several decades. I have arrived at a common sense description of temporal blindness. I can produce the first feeble answers to questions like: What is it? Who has it? How did we get it? (And we all did!) How does it affect personal decisions? How do these decisions shape the human condition? And – most importantly – what can we do for the next generation to prevent their becoming temporally blind?

Chapter 2. Hidden by common absence

I can not prove that temporal blindness is a universal affliction of the human population. Yet my experiences drive me to find ways of making such a proof.

2.1. First attempts to measure temporal blindness

As a graduate student years, I implemented a crude test to evaluate temporal blindness. While the test never progressed beyond an experimentally flawed prototype, the experience suggests that most individuals, independent of their culture, education, or intellect, are equally time blind. That is they share the same limits when choosing behaviors to respond to dynamic environments in which they were immersed.

The test was specifically written for sixth graders. It required fifth grade math skills and sixth grade reading skills. Each young subject of the experiment was given an island, a herd of bison, and some beans. The goal was to use the beans to keep the bison from starving to death. The beans could be fed directly to the bison as well as planted, harvested, and stored. At the end of each year, points were awarded for each bison living and subtracted for bison that had died. The object of the test was to maximize total points earned over a decade.

There were some obvious pitfalls. If all the beans where fed to the bison the first year, and no beans were planted, then at the end of the first year there would be no bean harvest. In the second year, there would be no beans to feed the bison and all would starve to death. Conversely, if all the beans where planted and none were fed to the bison in the first year, all the bison would die. It was a planning problem. The decision makers had to live in the second year with the results of the decisions they made in the first year. During the fourth year they had to live with the results of their decisions in the previous three years.

It sounds like life. However there was one new twist. In this management problem there was no uncertainty. The result of each action was completely predictable. Which means the test taker could figure out exactly what each round of decisions produced without having to try them. He or she could figure out before the first year's decisions what would be the size of the herd, the bean reserves, and the point standing at the end of the first year. With this information they could make decisions and figure out what would be the size of the herd, the bean reserves, and the point standing at the end of the second year for the combined hypothetical decisions of the first year and second year. Using the same procedures they could actually figure out their final score at the end of the tenth year before making the first irreversible decision in the first year.

The test was designed to give people who "guessed" – terrible scores. Only subjects that planned their way through the first four years, only those that discovered the potential pitfalls before experiencing them, survived to manage ten years. The mathematics where quite simple. There were two hard challenges. The first was for the test taker to visualize that the problem allowed itself to be solved without guessing. Second was to visualize a process to separate the behaviors that lead to desirable results form those behaviors that lead to tragedy.

The bison/island test was a little like the driver/deer/skid/recovery test. In both the decision maker was presented with conditions they had never experienced and future conditions depended on immediate behaviors. The two problems were different in that the bison island system did not change while the thinking process proceeded to figure out what these conditions were and what behaviors to take to obtain desirable conditions.

A great deal of effort was expended to make the crux of the test these temporal inference aspects. The number of relationships and the numbers that describe the states of the system were chosen to make visualization and calculation easy. The text was written to minimize errors in reading.

The bison/Island was suppose to invite people to plan – not guess. The directions told them not to guess. However, subjects guessed anyway. They probably thought that they would be given a second chance. However, for my purposes these secondary scores were not indicators of their temporal inference skills. How well people improved over multiple tries on the test would predict the quality of their experiential learning skills not their inference learning skills.

The computer based test kept meticulous notes on what each test taker read and reread and how much time was allocated to each reading and computation task. Providing an on screen calculator, allowed the record to show exactly what numbers were used and how they were used at each decision point. If they failed because they miss read a piece of data or a relationship and then used it to make decisions their score was thrown out. Their failure to "do well" was tainted by factors other than poor inference skills.

Test takers with temporal sight, should succeed in a clear way. Test takers that fail should fail, not because of some mistake in reading or adding and subtracting, but because they simply did not realize the form of the problem, and or were not motivated to implement a planning solution which was within their capacity.

Even after throwing out the tests where other factors appeared to influence the final score, the test results were so uniformly dismal at the sixth grade level, that I tried seeing if Stanford students could perform better. Their distressing results next led me to give the same test to a diverse group of Stanford professors. These were exceptionally bright, well–educated people from diverse backgrounds including operations research, engineering, psychology, political science, economics, computer science, education and business.

Comparing the results was astounding. The professors' scores could not be distinguished from the sixth grader’s scores. Both groups, the sixth graders and the Stanford professors, did poorly. By preventing the starvation of any bison in the first years they (due to lack of full use of the information) set in motion a plan that predictably sacrificed most of the herd in the third or fourth year.

The hypothesis that needed proving was that, temporal inference (not experience or transmitted knowledge) is equally undeveloped in both groups. However this was not to be. Word quickly spread that there was a simple test making the rounds, one which implied that Stanford professors were no smarter than sixth graders. Faculty, initially inquisitive and amused, became hesitant to volunteer. Professors I didn't even know would run the other way if I happened to wander toward them in the corridors, carrying the computer monitor under my arm.

From this experience I believe that a more rigorous test design would show, most adults perform no better than sixth graders in this type of problem solving. A temporally blind sixth grader sees no better than a temporally blind college professor. In some domains that require planned behavior we are all equally blind.

2.2. Even temporal experts exhibit temporal blindness

Consider the learning implications implied by the fact that most technically trained individuals between 1920 and 1968 did not request seat belts in cars. The computation of motion and forces of a human body inside a car during an accident is well within their capacities. It is well within the capacity of a high school physics student. However, physicists and engineers, even safety engineers did not request seat belts for themselves or their families.

Even thought they knew about accidents and injuries, it never occurred to them that there was an opportunity for an application of their temporal analysis capabilities. Even thought they could have easily calculated the reduction in forces provided by seat belts, they just did not do it. Like everyone else, they continued to perish on the highways and to insist that their children sit back on the front seat, so they could be manually restrained during braking and god forbid collision. They fared no better in accidents than the general populace.

It appears temporal blindness exists even for those that have had training that makes explicit the motion and forces in dynamic systems. There is a difference between specialized training in temporal problems and temporal cognitive skills. While this difference is described in detail in Time blind - the solution the second book in this series, here the inference of its existence helps describe our temporal blindness.

2.3. If you are like everyone else?

OK smart guy how come you think you are such a good driver when you really stink? How come you never figured out that your "skid control skills" only worked in a very small part of the skid situations into which you could drive? How come you were not motivated to understand skid control by accident injuries you saw every day of your life?

The inference is, if your temporal inference skills were a little better you would not make these kinds of errors. More importantly you, at least in the temporal inference domain, are about as smart as everyone else.

The fact that no one developed and used the "heading movement" solution to skid recovery after a half century of individuals driving cars (and there were billions of drivers during this time period) contributes to my belief that temporal blindness at least at a level high enough to resolve skids is probably universal.

Chapter 3. Boundaries between ability and disability

Human beings do understand some motions and use them to choose behavior. We understand that shaking baby rattles make noise, thrown balls have trajectories, and planted seeds grow to be flowers. We choose behavior that correctly reflect these diverse motions.

However, there are motions, we fail to see. There are motions we see and still fail to predict conditions they produce. We predict future conditions and still fail to give them value. In the competition among alternative behaviors, these distortions influence choice.

The following examples show changes in behavior that would result from underused or incorrectly used information when it describes temporal aspects of the environment. The examples show where our cognitive abilities accomplish their tasks well and where they fail. They describe the boundary between our temporal abilities and our temporal disabilities.

3.1. Boundaries from limitations in learning

Learning processes like learning to play catch or riding a bicycle demonstrate both the power and the limitations of our temporal thinking abilities.

Learning to play catch with a baseball is within our temporal capacities. However, it lies very close to the boundary. For instance, too few practice throws prevents learning. Too much time between iterations (a week between each throw and catch) prevents learning. If the object is traveling two fast to see (like a bullet) it prevents learning. If the object was light enough that its trajectory was greatly affected by the slightest air currents it would prevent learning.

These limits describe temporal boundaries of our learning capacities in understanding trajectories of thrown objects. They explain why we all learned to play catch, and why we all failed to wear seat belts in the 60's. These explanations will eventually show, in Time Blind - the solution, why what we learn in early in childhood supports learning in environments like baseball and not in environments like auto accidents; why we learn to duck fly balls so easily. It also explains why we don’t learn to wear seat belts from being in or observing car accidents.

A child plays with rattles and forms her understanding of physical motion. She is not temporally blind to these physical motions. The proof is that she gained cognitive capacities to learn more complicated systems of motion like playing catch.

These learning capacities have limits. Consider the case where after playing with many toys, but prior to learning to play catch, our subject moves to a different world. In the different world it takes six months for a thrown ball to travel the distance between pitcher and catcher.

In this slowed down world she can never learn to play. The temporal capacity previously developed while playing with toys, cannot support experiential learning in this different world. She would not live long enough to get in enough throws and catches. In a lifetime, they would not equal the learning iterations she would have gained in just one quarter hour of backyard practice on earth.

Furthermore, if a consequence iterates every day and the behavior that made it happen occurred everyday six months ago we can not learn it. As the delay between behavior and action increases the learning problem gets harder.

For example, it is impossible to learn experientially to control wars if they happen only once every 30 years. Even if you lived 50,000 years and experience as many wars as you have throws and catches. It would still be impossible to learn to control wars. There are simultaneously many actors' behaviors contributing to the war. Even an individual's behavior contribution is not singular but the accumulation of behaviors over the 30 year time period.

Thousands of historians, limiting themselves to experiential records of past events, have little chance of producing either: a) predictions that illuminate that "contained in even peaceful conditions is the trend toward war," or b) a structure that shows which behaviors change the trend. In this sense the historian is like the seatbelt non-wearer, they have the data that suggest wars happen in every future but can not discover the social causality nor the behaviors to prevent war.

Why can't a child that lives 50,000 years in the slowed down world learn to playing catch? There are enough iterations. I think this can be explained by limitations in what I call physiological memory.

To learn how to throw, we establish a destination for the throw. We throw the ball. We perceive where the throw landed. We measure the difference between desired and actual destination. With this difference we modify the previous throw.

One of the cognitive capacities that makes this process work is that the initial throw was "remembered." This includes precisely how we held the object, how the muscles tensed, and when and how it was released from our grasp, etc. It is this "remembered" throw that must be modified by the difference between desired destination and actual. We change the trajectory by modifying the combination of remembered physical actions.

Remembering the throw is, in fact, quite complicated. Far too complicated to do consciously. Fortunately, human physiology does the remembering for us. The body remembers how all the muscles felt. It remembers if we started on the back foot and changed weight to the front foot, how we swung our arm.... etc.

While the physiological memory process is beautiful, it does have limitations. Physiological memory, at least that part we use to learn to play catch, is stored for only a few seconds. In our world this short life is of no problem. The next throw is well within this brief period. However, in the slowed down world the physiological memory is long gone when it is needed as the basis to design the next throw six months later.

We can’t learn to play catch in the slowed down world even if we lived long enough to get in a normal complement of throws and catches.

From this we can identify another boundary of our temporal abilities. We can learn to perceive and control motions using direct experience only if we can rely on physiological memory to remember the intricate details of the actions that we took to implement the last physical action.

To implement trial and error learning, the student must proceed through a series of iterative adjustments. In learning to play catch, these calculations modify each successive throw, so that the next will have a destination closer to that originally targeted. How these calculations are performed also defines part of the temporal sight/blindness boundary.

As the "memory" required for learning exists at the physiological level, not at an explicit conscious level, so too is the "computation" of each successive throw done at the physiological level. The amount a given throw must be changed is physiologically computed using the perceived error in the first throw. Also done physiologically is the translation of the error into physical actions for the second throw.

Doing this computation in an explicit and conscious way would be an extraordinary task. Few of us could be encouraged to even attempt the physics and math to accomplish it.

Consider for a moment how hard it would be to try to explain to a novice which arm muscles to contract to throw a ball five degrees to the right and 10 feet farther. The modification of the throw "by feel" may be quite easy to perform. But it remains devilishly hard to make explicit.

This dependence upon physiological computation helps describe a third limitation in our abilities to understand and control motion. In cases where physiological computation cannot be used (situations where motions can not be physiologically sensed or remembered) we will have greatly diminished capabilities to modify "motion-controlling-behaviors." The next example of riding a bike makes this abundantly clear.

Learning to ride a bicycle is another example which illustrates the boundary between those motions we manage well, and those which exceeded our cognitive abilities. Most of us have learned to ride a bicycle. Frightened as we were, after a few close calls, we tamed our two–wheelers.

To begin, we had to be rolling fast enough to make balancing easy. If we attempted to travel at very slow speeds, learning became impossible. Learning to bicycle may have been the first time in our lives that we could not master the learning task muscle by muscle, at a very slow speed first.

We all took our bruises in the learning process (we fell down) and were eventually glad we did. For that bicycle, once tamed, extended our speed of travel by more than three times, and the area we could explore without a ride from parents perhaps ten–fold.

The quickness and the almost universal success of people learning to ride a bike belies the complexity of the learning task. To realize the boundaries of those capacities brought into play, we must recognize both what we are doing when we bicycle; and also why we fail to learn to ride when we can not use these capacities to help us.

Physiological memory, and physiological computation are playing indispensable roles. These capacities together accomplish a task equivalent to that performed by a very well trained and disciplined mind first mastering physics, understanding all of the mechanics of the bicycle, the human body’s sensors and muscles, and formulating them into a set of interacting equations called a program. The program’s job is to read the sensors that describe the condition of the bike and rider, and dictate a stream of behaviors for the muscles to execute.

Bike riding is so "computation -intensive" that until recently even the fastest computer could not perform the program’s calculations fast enough to keep a robot-driven bicycle upright. Even with the computer power to run the program, the computer is doing only a fraction of the computation that a young child does when she learns to ride a bike. The computer already has a fully designed program installed. The child, during his or her efforts to ride the bike did not have this finished program. The child in learning how to execute the required behaviors to ride the bike, had to simultaneously write program. During learning the child had to create a primitive version of the program and then write improvements from riding experiences.

The combined learning task, as the child performs it, is almost beyond comprehension. Yet a young child, with the temporal capabilities she gained by learning to walk, playing with toys, etc. can learn to ride a bicycle in an afternoon. While this impressive task shows immense design and computational power in a temporal domain, riding a bike lies near the limit of these capabilities. Tasks just slightly more complicated or tasks which can not depend on the human body’s physiological computation will be beyond human ability.

To visualize such a task, consider a new world wherein falling off a bike is fatal. While riding the bike remains about as difficult as in the normal world, the consequences make successful learning unlikely. Almost no one masters a bike without at least one fall.

In a world where falling off a bike is lethal, instead of "feeling our way," we must make explicit the physical parameters of the system. We must unerringly infer from analysis the correct behaviors for each state of the bicycle, so that we can perform them correctly the first time, without experience.

This may seem like an impossible task. You and I may never be able to learn to ride without relying on our physiological capacities. However, given a situation where bungling our first attempt will kill us, we have no choice but to learn to first explicitly express and then manage the motions in which we are immersed.

Given such circumstances, developing a program to make predictions of outcomes based on a range of behaviors is extraordinarily complex. It requires realization or discovery that we are immersed in motion, followed by analysis which connects the causal mechanisms of system motion. It requires not only an understanding of the temporal aspects of the physical world but also a more advanced form of temporal capabilities all together, what may be called temporal sight. Temporal sight must include the capacity to recognize when trial and error learning will be expeditious, or wildly inappropriate.

Thus one boundary between temporal sight and temporal blindness can be described as the difference between individuals who expertly handle the physiologically–based bicycle learning (that is most of us) and that individual who can handle temporal learning tasks which are not physiologically based trial–and–error (this excludes most of us.)

3.2. Boundaries from idiosyncrasies in use of information

"System change" can be derived from observations of a variable at two points in time. When the change is large and the time interval is small as with a thrown object, physiological computation can produce feelings of motion. However, when the change is small and the time is long, as with the stars moving across the heavens, physiological computation can not produce feelings of motion. In the latter cases the motion (change during a time period) can only be computed using conscious cognitive processing. That is recording position and time at two instances. Computing the differences in position and time interval. Dividing distance by time to get average velocity for the interval (inches per second, ft per minute, miles per hour, etc.)

Sometimes the "difference descriptor" of an object is not its location but its amount. For example gallons of water in a pail under a roof drip. Whether it is a distance or an amount, what is being calculated is a trend - a change in a value per unit time.

Trends can be depicted by graphs. The size of a descriptor at several points in time are plotted on a graph. If the line connecting the points rises we call it a positive trend. A horizontal line denotes a zero trend. And a downward line denotes a negative trend.

If you know the value of a variable and the magnitude of its trend at a point in time, "without any more information," you can predict the value of the variable at a future point in time.

This prediction can be used to choose behavior. This behavior is based, not on the experience of the event, but on a image produced by a computation of consciously gathered environmental information. Even the algorithm that facilitates the computation had to be acquired, stored in memory, and retrieved consciously.

Each of these conscious activities, each successive product of these activities, like their subconscious physiology counter parts, impose limitations on the behavior selection process. Inspections of these "conscious abstractions" help define additional portions of the boundary between our temporal blindness and temporal sight. Let me provide three examples of limitations in our capacity to give meaning to motions within our environment or trends that describe these motions.

All trends are not equally meaningful. Some trends predict future values of a variable. Some trends only quantify history. This subtle difference delineates yet another boundary between our temporal sight and our temporal blindness.

Above, are two sets of measurements have been graphed. The first shows the account balance of a Girl Scout troop. The second shows the distance a train has traveled. These two graphs, while appearing similar, are very different. One shows a non–predictive historical record. The other presents a historical record and predicts the measurement’s value at future points in time.

The money in the girl scout account graph is money raised by a Girl Scout cookie sale. The vertical axis is the amount of money. The horizontal axis is the number of weeks duration of the sale.

At the start of the sale, the amount of money in the account is zero. Each week, for five successive weeks, the amount climbs. The graph of the amount when plotted is a rising line. The slope of the line is the trend.

If the girl scouts make $250 dollars on their cookie sale after five weeks, we can derive a trend from the historical data. We say that they made $50 dollars per week. The trend describing sales is 50 dollars per week.

The second graph tells the distance a train has traveled along an interurban rail loop that runs 24 hours a day. Each hour the distance traveled is plotted on the graph. Similar to the Girl Scout’s graph, the points form a rising line to the right. The graph starts at 0. Five hours later, it reads 250 miles. Like the first graph, the trend in the variable "total distance traveled" is 50 MPH.

The trends in both graphs appear identical. Both lines slope upward to the right – one at 50 dollars per week and the other at 50 miles per hour. But do the two trends mean the same thing?

The girl scouts have been selling cookies to their neighbors. Let's say they sell cookies to 1/5 of the households in town each week. The amount of cookies sold each week remains about the same, and the graph of the dollars in the sales account rises evenly at 50 dollars a week, to 250 dollars.

Does the line’s slope upward (its trend of $50 per week) predict how many dollars the troop will make after 10 weeks? Of course not, the girls have canvassed and sold cookies to the entire town by the end of five weeks. Guilt and civic appreciation have their limits, as does one's appetite for butter cookies. Sales per week will plummet. The trend in the Girl Scout graph is not predictive of sales in the sixth week.

The train’s 50-MPH trend, however, is predictive. It correctly predicts the train will have traveled 500 miles after 10 hours.

Whereas the number of weeks into the cookie sale eventually does affect how many cookies can be sold, the distance traveled by the train in the future is not affected by how many miles the train has already traveled. Thus the two trends are different – one is predictive and the other is not. Successfully distinguishing between predictive and non-–predictive trends is not always easy. The capacity to make the distinction is another demarcation of the boundary between our temporal sight and temporal blindness.

The Girl Scout example displayed how we might give relatively rapid trends indicated by historical data undeserved credibility. Below, I show how we give historical data concerning slow trends too little credibility.

In the US in 1993, we suffered two major natural disasters. Hurricane Andrew destroyed 70 thousand homes in Florida. The Mississippi River flooded and destroyed thousands of homes in the Midwest.

These are not unknown natural events. They are recurring events. Floods and hurricanes of equal severity have occurred over and over again in the very same regions, for many thousands of years.

Such events have diminished in neither magnitude nor frequency. Thus the historical record predicts that they will continue in the future.

Temporal blindness is indicated by a belief that the predicted future will not come to pass. Temporal blindness is indicated when people build or buy their homes in the probable path of such disasters. They just don’t believe that the recurrent phenomenon will ever recur.

A mechanism is a group of parts that are connected together. A clock’s hands are a good example. Gears make the hour hand move if the minute hand moves. Gears make the minute hand move if the second hand moves. Because of the interconnectedness of the gears and by knowing how many seconds have ticked off, you also know how many minutes and how many hours have elapsed.

Parts of our world are like a very complex watch. We know the value of some variables (hand positions), and some functions that relate the variables (gears.) If one of these variables changes by an amount, then we know variables, connected to the first also change proportionally. For example if a car gets 10 miles per gallon and a trip odometer reads fifty miles, then it is a good bet that five gallons of gas are missing from the gas tank too.

If adjacent variables change, then it is possible to trace changes through chains of connected variables. Many predictions can be made, not through direct experience but through computation.

Gears make the relational movements of watch hands immutable. None of us believe that the hour hand can change with out the minute hand changing. In our world, there are mechanistic systems for which we should feel equal strength in our predictions. Part of our temporal blindness is revealed when we fail to see that a system is mechanistic, that some of the variables have causal connections, that they form immutable predictions. Part of our temporal blindness is that our predictions, our estimate of the feelings that we will have when a predicted event occurs, will be less than those feelings if the event is happening now.

Below I will give examples where predictable motions based on structure were invisible or visible but ineffective in influencing behavior. As a result, the chosen behaviors produced an undesired future. These examples reflect limits in our cognitive abilities. They reflect the boundary between temporal sight and temporal blindness.

Electric blankets maintain a bed temperature set on the blanket control. If the room is very cold, the bed temperature is maintained at the set temperature using lots of electrical energy. When the room temperature is just below the set temperature, little electric power is fed to the blanket. The person under the blanket did not request more energy on cold nights and less energy on hot. They requested a bed temperature. It was the blanket’s control that requested the energy to maintain "that" temperature. The blanket’s control acts like the thermostat on the wall of your house.

The blanket system, Figure 3.2–30 is sketched as three components; the thermostat the blanket heater and the sleeper. The sketch can be converted into ovals and arrows. Each oval is a state of a component. Each arrow is a mechanistic connection.

Since the last variable effects the first, the chain of affects is shown as a loop.

Let’s see how this loop system works. Let's say a person comes to bed feeling chilled. He turns up the dial, The "higher" dial setting makes the blanket warmer. Then his chill goes away. The blanket makes him feel too warm and he turns the dial down to the original set temperature. In this system, the sleeper can correctly predict what will happen if he changes the dial.

The first electric blankets for double beds had only one controller Figure 3.2-40. Couples negotiated the set temperature. The system reflects a delicate equilibrium between the always too cold spouse and always too hot spouse. No one dared change the dial again. A cold spouse would wear socks. A hot spouse would sleep nude. If someone came to bed chilled and tried to change the dial there would be unhappiness. Everyone knew it. They understood how the structure controlled the system, or if they did not, they had rules to govern behavior.

It was not a good design to have two people under a single control blanket. Eventually, electric blanket designers developed an alternative design. They created distinct right and left side heating coils in the blanket, and hooked the right side coils to the right controls and the left side coils to the left control as seen in Figure 3.2-50. Each partner could pick his or her own blanket temperature, independently. And if they wanted to change it, they could without affecting the blanket temperature of their spouse.

While engineers solved the crisis created by the single control blanket they made possible the creation of a completely new and unintended system. The two independent and stable control systems shown in Figure 3.2–50 could be changed into one unstable system simply by turning the blanket over - something that could be accidentally accomplished every time the bed was remade.

After the blanket is turned over a trace of the information path among the six components shows the wife's controller now affecting the heating coils over the husband. The husband’s controller now affecting the heating coils over the wife. The connections now make a single loop as shown in Figure 3.2–60.

Let's see what happens when the husband feels chilled and turns up his dial. He makes his wife’s side of the blanket warmer. She feels too warm and turns down her dial setting. This in turn makes his blanket colder. He responds by turning his dial higher. His wife, feeling the extra heat, turns her dial down again. Each variable keeps bumping the next until she is roasting and he is freezing.

The new loop no longer causes either individual’s feelings of warmth to hover about set values. Instead, any change in any of the variables will cause each person’s feeling of warmth to change continuously away from their desired value.

The structure predicts that any change will lead to an unending chain of undesired changes. The structure of the system predicts this instability before any change exists. It shows the succession of change without any change happening.

The structure cannot predict "when" this instability will occur; that is, when one spouse will come to bed chilled or over heated. If both spouses have their controller set to the same number, when the blanket is turned over neither experiences any change in comfort. They could live with the unstable system for months before a chilled or overheated spouse starts the spiral toward undesirable thermal conditions.

The structure cannot predict the direction the system will travel. That is it can not predict who will freeze and who will roast until one spouse reaches out and requests a new temperature, cooler or warmer for his or her blanket.

If system structure was a normal tool for understanding the physical world, then the system's structural change (caused by the blanket flipping) could have been recognized and used to prediction its instability before any changes were experienced. Figure 3.2-60 does describe the potential trends and final conditions before they occur. However, how many of us do create structural models for physical systems that surround us? Just about as many as those who would not put their arms to stop a flying child during braking; or learn to ride a bicycle by reading a book.

The capacity to predict the consequences of change (which have not occurred but which are stored in our system's structure) has been beyond normal human cognitive capacity. However, like learning to protect our unbelted kids, the ability to predict consequence from structure may have to be second nature for us to survive in our world.

The undesired conditions produced when a husband and wife have no structural model for their electric blanket, provide us with an