Essays & Articles


Chapter 11: Turn Signals Are the Facial Expressions of Automobiles

Copyright © 1992 by Donald A. Norman. All rights reserved. Originally published by Addison Wesley. Now out of print.

Nature produces a varied assortment of creatures. One that has long fascinated me is the red-tailed baboon. You know, the one with the, umm, red-colored rear? All that color and display, but for what purpose? Well, looking at the rear end of some automobiles reminds me of looking at the rear end of some baboons.

Social cooperation requires signals, ways of letting others know one’s actions and intentions. Moreover, it is useful to know reactions to actions: how do others perceive them? The most powerful method of signaling, of course, is through language. Emotions, especially the outward signaling of emotions, play equally important roles. Emotional and facial expressions are simple signal systems that allow us to communicate to others our own internal states. In fact, emotions can act as a communication medium within an individual, helping bridge the gap between internal, subconscious states and conscious ones.

As I study the interaction of people with technology, I am not happy with what I see. In some sense, you might say, my goal is to socialize technology. Right now, technology lacks social graces. The machine sits there, placid, demanding. It tends to interact only in order to demand attention, not to communicate, not to interact gracefully. People and social animals have evolved a wide range of signaling systems, the better to make their interactions pleasant and productive. One way to understand the deficiencies of today’s technologies and to see how they might improve is to examine the route that natural evolution has taken. You know the old saying that history repeats itself, that those that who fail to study the lessons of history are doomed to repeat its failures? Well, I think the analogous statement applies to evolution and technology: those who are unaware of the lessons of biological evolution are doomed to repeat its failures.

Evolution works its way slowly, ponderously. Even those who believe that it progresses in rapid jumps think that these jumps take tens of thousands of years. By the standards of an individual human, even the most rapid evolutionary changes are too slow to have any impact on the individual. But the study of evolution might aid us as we design artificial devices, enabling us to profit from evolution’s experiences, letting is accomplish in years what has taken millennia for evolution. Over time, evolution tries out a wide variety of methods to ensure survival, some that modify the animal or plant, others that modify behavior, and still others that affect the cooperative, interacting nature of social structures. The baboon’s red rear is one result of this process, as are other signaling systems, such as the calls and cries of animals, gestures, facial expressions and the speech of humans. We can learn from evolution’s successes.

Animal Signals

Animals require a number of different signaling systems to communicate internal states both within the animal itself and to others. Plants signal their maturation and ripening through colorful displays that attract the attention of insects and birds, the better to pollinate and propagate their seeds. These signals need not be consciously given nor received to be effective: all that matters is that there be some perceivable change of state that other organisms can make use of. Thus, the presence of snow, ice, and objects waving in the wind signal a state of weather without any conscious volition on the part of the atmosphere. But the signals, nonetheless, are valuable ones.

The red tail of the baboon is certainly not a conscious signal, and for that matter, neither are many of our facial expressions. Facial expressions arose as side effects of the facial muscles as they prepared the mouth, lips, and teeth for activity. But as they were perceived and used by other animals, they began to evolve toward a symbolic, meaningful role, so much so that today many of our facial expressions are voluntary, conscious, and deliberate.

Facial and body expressions have evolved because they serve a useful purpose. The signaling of intentions and internal states among animals works to their advantage. The importance of facial expressions was recognized quite early in the study of animal evolution: Charles Darwin devoted an entire book to their study — Expression of the emotions in man and animals. Emotional expressions act as a side-channel of normal communication, outside of and without interference to the spoken language. They offer a commentary, and when this information channel is lost, ambiguity and difficulty in interpretation often results.

Social interaction requires a very different set of behavior than does solitary action. Successful social interaction means cooperation, joint planning, troubleshooting, play, rivalry, competition and comradeship. It means social honesty and probably deception. If animals — including people — are to form functional social groups, they must develop means of communication, of synchronizing actions, of cooperation and, occasionally of deceit. Social interaction was the driving force for the evolutionary development of social signaling devices: facial expressions, colorful plumages, the red-tail of the baboon — all are social signals useful to the communication, interaction, and protection of animals.

There are many controversies surrounding the development of human intelligence. The traditional views link the development of intelligence to language or tools, or perhaps to the need to be flexible and innovative in dealing with a changing, complex environment. There is much to be said in favor of all these possibilities. Recently, however, a new suggestion has emerged: higher forms of intelligence result from the need to handle the problems of social interaction. This is an attractive notion, for social interaction requires numerous talents and abilities, including the ability to let selected participants in a social group know the intentions and beliefs of others. Human intelligence almost certainly did not result from any single factor. It is most likely the result of multiple forces acting over long durations, but social interaction seems like a good candidate to be one of the primary forces.

One of the important aspects of intelligence is the ability to communicate. There are actually two levels of communication that we need to be concerned about: one is internal, between the body and the conscious mind, the other is external, among animals and people. Internal signals are very important for the individual. Thus, the body informs the mind when it is hot or cold, in pain, or ill. The results are subjective feelings of hunger and thirst, sickness, or pleasure. Emotions also serve as internal signals: happiness, pleasure, sadness, and anger. Joy and love, envy and dissatisfaction. Emotions are complex mixtures of biological and mental states, the neurochemistry of the brain interacting with the information processing of events and expectations.

How many times have you been in situations where your mind tells you one thing and your body another? How often have you delayed a decision, saying “I want to think about it,” but really meaning, “I want to see how I feel.”

Some people feel that emotions are a vestige of evolution, a type of “animal” behavior that the human race will eventually outgrow. Well, not really. If you look at the role emotions have played in evolutionary history, it would appear that the more sophisticated the animal, the greater the role played by emotions. We are the most sophisticated of all, and thereby the most emotional.

Emotions speak to our cognitive minds, sometimes telling our conscious selves things we would rather not know. Emotions also trigger body changes and facial expressions that signal others. Moreover, because other people in our social circle are apt to share similar backgrounds, the same basic knowledge, and the same biology, we can assume that things that make us happy or angry are apt to make them happy or angry as well. Taking the point of view of another is aided if only we could read the mind of the other, and this is where external signals come in.

How do we tell what others are thinking? We do so through a variety of techniques, with varying degrees of accuracy. Facial expressions, gesture, and body position act as cues to a person’s internal states. We often call these things “body language,” the name indicating the communicative role. Body language makes visible another’s internal state. The blush of the cheeks, the grimace, the frown and the smile all act as readily perceivable external signals of a person’s internal state, making visible to observers what would otherwise be difficult or impossible to determine.

Of course, facial expressions can deceive. First of all, they are subtle, a rich interplay of complex musculature, facial features, and coloration. Body language is even more complex. Not every one can read the signals, and for that matter, not all scientists are convinced that the signals are there. When someone crosses their legs, are they sending a subtle message or are they simply trying to get comfortable?

The study of human emotions, of course, is a complex topic, one that has occupied psychologists for years. We are learning a lot about the neurochemistry of emotions. We know that there is a complex interplay between the neurological state of arousal and the cognitive interpretation of the state. One long-standing argument is over the ordering: which comes first, the emotional state or the interpretation? To simplify the complex issues, one side of the argument, those who think the emotional state comes first says that, essentially, you first notice your body’s emotional state and then interpret it: “Oh, oh. My heart is pounding. I feel tense. I’m sweating — I must be afraid.” The other side says, no, the interpretation leads to the state: “I have to give a really important presentation tomorrow. I’m not ready. I can imagine everyone looking at me, their eyes registering skepticism. I’m afraid. There, see: my heart is pounding. I feel tense. I’m sweating.” Both sides of the debate have merit, which means that the true story is probably a combination of these two differing views. To the onlooker, however, it doesn’t matter how the emotional state arouse: the facial expressions and other body signs are external signs of the resultant state.

To speak only of the body signaling the conscious mind is a great simplification. Nonetheless, simplification is a useful starting point, useful and scientifically reasonable. Moreover, the one person’s interpretation of another’s mental state really does depend upon this information. Of course the visible signs of emotion are often ambiguous, incomplete, and misleading, but so too are our judgments: our interpretations of another’s feelings and beliefs are not very accurate and often are misleading.

Animal Deceit

The fact that facial expressions and spoken language can be used to communicate internal states is useful, both for conveying accurate information and for the deliberate deception of others. I can appear happy when in fact I am not, feign sadness for an event that pleases me, or disguise my knowledge in a variety of ways. The study of deceit in animals is a powerful way to get clues about their level of intelligence, clues to the working of evolution. As a result, there is an ever-increasing number of studies concentrating upon the evolution of deceit, the false communication of knowledge, intentions, and actions. Deceit seems a natural property of higher animals, and not just to protect the young from predators. Chimpanzees, for example, use deceit to avoid sharing food, to curry favor with higher-ranked animals, and to obtain sexual favors behind the backs of disapproving higher-ranked males. These kinds of deceits may very well require more intelligence and cleverness than truthful communication.

Deceit is a necessary part of civilized life &mdash not the evil deceit where one person seeks to benefit at the expense of another but the polite deceit of social interaction. The polite “thank you” when someone presents an unwanted gift, or the polite acknowledgements when someone else’s well intentioned efforts go awry at your expense. Social interactions require falsehoods to maintain themselves, and little benefit would result were every negative thought transmitted to others. Mind reading may seem like a desirable trait, but it would more often backfire, causing grief where none was intended or desired. All societies and cultures have developed social codes that govern interactions and mask true feelings and beliefs under a cloud of cordiality or, at least, civility. The elaborate honorifics and “speech acts” of society exist for good reason.

Lies and deceits have their place in the world. Social interaction would be less pleasant if the truth were always told. Casting blame on the other person is an excellent administrative policy. When I need to make an unpopular decision, I sometimes blame my administration — “I’m sorry, but my Dean won’t let me do that” — checking first with my Dean, of course, to make sure that my stand is understood and approved of. In turn, I advise the people who work for me to use the same tactic, to tell others that they are sorry, but their boss won’t allow it. In fact, it works in both directions. When my Dean wants to resist an order from above, he can call the department chairs that report to him and get us all to allow him to say, “sorry, but my departments won’t agree.”

Social interaction is complex. Human beings are the most social of all animals, and we have evolved elaborate schemes for interactions, schemes that allow us to co-exist and cooperate with friends and to resist the pressures of enemies. The act of deceit is complex, and most animals are incapable of it, not because they are more honest but because their brains are inadequate. Only the most sophisticated of beings can lie and cheat, and get away with it.

There is, by now, a large research literature on the abilities of monkeys and other animals to deceive. One of the marks of evolutionary development is the evolution of “social artifacts,” the ability to use social interactions and strategies for cooperation and for deceit, all to the betterment of the social group. It was a relatively late development in evolutionary history and of all the animals that do practice social cooperation, the human is superior at forming tight social bonds as well as in using the social deceits necessary for smooth social structures.

Only the most advanced of primates, the chimp, seems capable of true deceit. Other species sometimes try, but it is a bit too much for their minds to manage. Look at this description of an African monkey, the vervet, trying to deceive:

“A male vervet, Kitui, gave leopard alarms when challenged by a rival male, causing the other male to flee up a tree.”See that is real deception: to get rid of his rival, Kitui called the monkey equivalent of “Fire! Fire!” The problem is, the monkey couldn’t quite pull it off.

“However, to reinforce his point, Kitui descended from his own tree and walked across the open ground toward his rival, still calling the equivalent of “Run for the trees.” … (This is like) a human three-year old who with crumbs all over his face denies having raided the cookie jar”

Quotations and the evidence about primate behavior from Jolly, A. (1991). Thinking like a vervet. Review of Cheney & Seyfarth (1990), “How monkeys see the world: Inside the mind of another species”. Science, 251,574-575.

Also see: Cheney, D. L., & Seyfarth, R. M. (1990). How monkeys see the world: Inside the mind of another species. Chicago: University of Chicago Press.

Notice that the vervet monkey is intelligent enough to give a false signal in order to scare away a rival, but not intelligent enough to act out the entire behavior. It doesn’t really matter for vervets because the rival male isn’t sophisticated enough to see through the ruse, to realize that Kitui’s behavior indicates that the call is a fake. Young children will make the same type of error as Kitui, but parents are quite capable of seeing through the deception. So, for this kind of intelligent behavior, a vervet is acting somewhat like a young child.

Notice what Kitui would need to realize in order to do the ruse correctly. Kitui would have to know not only that the leopard alarm would cause the rival to flee, but that his own behavior must be in accord with the falsely announced state. The animal needs to know not only what a signal means, but how its signal and its own behavior will be interpreted by others, and then to understand that one can contradict the other and that others can perceive this contradiction.

A chimp wouldn’t be fooled by Kitui. Thus, a chimp has been seen using his fingers to readjust his mouth in order to hide a grin before turning to bluff a rival. This behavior shows that the chimp is aware of his own facial expression, aware that it is visible to rivals, and probably aware of how the rival will interpret the expression. I have seen the same behavior in adult humans.

Artificial Devices and Artificial Evolution

As we construct artificial devices with ever more power, even more intelligence, perhaps we will have to make them mimic natural evolution. Technology slowly evolves, not in the same way as the natural evolution of life, but through the artificial evolution of design. But in many ways, the evolution of machines is driven by the same pressures as the evolution of life: modifications that enhance performance and allow the organism or machine to survive and to compete in the world will survive, those that do not will disappear. Slowly, designers will add signals and warnings, self-assessments and communication devices, providing the artificial equivalents of emotions, facial expressions, and social interaction.

Natural evolution combined with cultural conventions determines the nature and interpretation of facial expressions of people. Machines pose an interesting problem, for they are artificial devices, manufactured by people. It is the rare machine that works entirely alone, isolated from other machines and from people. Machines have to be started, stopped, monitored, adjusted, and maintained by people. Many require considerable control by humans. Machines are social devices, for their manufacture results from interaction with people. As a result, some of the same pressures that gave rise to facial and emotional expressions in animals apply to machines as well, except here, the signals have to be designed, deliberately constructed and provided on the machine.

The lights and sounds of an automobile play a role analogous to the facial expressions of animals, communicating the internal state of an auto to others in its social group.

With animals and people we saw that there were two different forms of signals: internal and external. The same is true for machines, except here, we have to readjust our idea of the basic “unit” of analysis. For a person or animal I distinguished between the body and the mind. Internal signals informed the conscious mind of the subconscious information: body states such as hunger, fatigue, and comfort as well as mental states such as fear, joy, or anger. With a machine, there is no such thing as the conscious mind &mdash but there is the user.

Machines can be regarded as symbiotic units consisting of a machine and a person. Thus, a copying machine forms a functional unit only when combined with a person: person+copying machine. Similarly, the automobile by itself is non functional: the critical unit is driver+car. The person lies in relationship to the machine somewhat like the role of consciousness to the body: a supervising element that watches over and maintains the system, even as the system — human or machine — sometimes operates relatively autonomously. This is a dangerous metaphor to pursue deeply, for it fails in all sorts of ways, but it does have useful characteristics with regard to the way by which the machine ought to interact with people.

Just as animals have two levels of signals, internal and external, there are two levels of interaction of this person+machine unit, one internal, the other external. Internal signals in a person tells of body states, but in the unit of person+machine, it tells the person the internal states of the machine. A machine signaling a person isn’t really the same as a body conveying internal information to consciousness. The machine is performing a form of social communication between it and an outside agency, the person who is using or maintaining it. So the signals between machine and human have to be a combination of external, social signals and internal ones: Internal to the unit of person+machine, but external in the sense that the machine and the person are separate entities.

One of the special kinds of signals that this relationship requires is feedback about the operation itself. It is difficult to use a machine that does not provide feedback to the user. Mechanical devices tend to do this through their construction. A pair of scissors feels firm or sloppy: its blades snip-snap through the air with a pleasant sound, or scrape, moved only with great force. Or they might wobble, providing a sense of insecurity. A good knife provides feedback through its balance and feel as it cuts. Mechanical devices are often visible and audible, conveying considerable information about their operation, even to those who know nothing of mechanics. The designers do not have to provide feedback to the users: the very nature of the machine guarantees that.

Not so with electronic devices. Electronic devices work quietly and smoothly, invisible and inaudible. At most, one might get a hum, buzz, or cracking sound resulting from components that vibrate with changes in magnetic fields or from heating and cooling. These sounds, however, are peripheral to the operation and seldom convey useful information. But more important to the user is that electronic systems deal with information, not mechanical movements. Information is a commodity that exists conceptually, not physically. It occupies no space, makes no sound.

One of the reasons that modern technology is so difficult to use is because of this silent, invisible operation. The video cassette recorder, the digital watch, and the microwave oven — none are inherently complicated. The problem is their lack of communication. They fail to interact gracefully. They demand attention and services, but without reciprocating, without providing sufficient background and context. Little or no feedback. There are many reasons to need feedback about the state of a system, reasons dealing with our own need for knowledge and reassurance. This kind of feedback is essential in normal social intercourse. The spoken “hmmm” or the nodding of the head by the listener to a conversation assures the speaker that the message is being received. The feel of the screw’s resistance to my turning of the screwdriver provides useful feedback about the success of my operations. Feedback is a necessary part of all interaction, whether with people or technology, but it is more absent than present in today’s information-based technology.

If our information-based technologies are to become socialized members of society, interacting with and supporting the activities of people, then they have to be able to interact with us people on our terms, not on theirs.

Our most modern technologies are social isolates. Today’s technology provides us with ever-more complex machines, devices that can work at a distance or through non-mechanical components. Humans are often unaware of their presence, unaware of their internal states. The modern information processing machine fits the stereotype of an anti-social, technological nerd: it works efficiently, quietly and autonomously, and it prefers to avoid interactions with the people around it.

Just as it is valuable for humans to know of their own internal states, the better to manage their own existence, designers of mechanical devices need to signal the internal states of their machines, the better to keep them maintained and functioning. Hunger and thirst of animals translate to energy supply of machines, perhaps specified as the fuel level or the state of the battery. Is the machinery too hot, too cold? Lubricated properly? In appropriate adjustment? These are the things that the person maintaining the machine needs to know, information provided naturally by one’s own body, but that must be provided artificially for our artifacts.

This internal information is provided in a number of ways. Sometimes it is not given explicitly, but rather is stated as rules (“lubricate every six months”), sometimes it is assumed that the users will notice and repair deficiencies as they appear (“tighten connections as needed”). More complex machines require indicators of their internal states, and these are provided by lights and gauges, by instrument panels. The instrument panel of an engine shows critical aspects of its internal state, thereby allowing the driver to control it safely and efficiently, perhaps much as the hard-driving coach carefully monitors the emotional responses of the players, attempting to push them hard enough to do some good, but not so hard as to be destructive.

The instrument panel of the automobile is an internal communication within the driver+car unit. It is mostly self-centered, communicating information from the machine to the driver. Most instrument panels are like that: lights to tell whether a device is turned on, meters and other indicators to tell of the current state, buzzers and alarms to tell when something is wrong and needs immediate attention. No social protocols, no etiquette. No checking to see whether the person is busy at some other activity, usually not even a check to see if other alarms or warnings are also active. As a result, when there are serious difficulties, all the alarms and warnings scream in their self-centered way, the simultaneous array of lights and sounds impeding intelligent actions by the operators of the system. In places that have large control panels, such as industrial control rooms, commercial airplanes, and even the hospital operating room, the first act of the human operators is to shut off the alarms so that they can concentrate upon the problem. Unfortunately, the machines have no way of learning from the experience — you can’t spank them and send them to bed, nor is there the equivalent of a note to the parent. As a result, when trouble next strikes, the same rowdy behavior reappears.

Social issues are even more serious when we consider socially interacting units of machines. The prototypical example is driving, where the unit of driver+car interacts with large numbers of other similar units. Here is where we have a need for external signals, signal that communicate with the other units sharing the road, signals that allow others to know just what actions are being performed, and in many cases, what actions are intended. There are times when it is necessary to know what is on another’s mind.

The same technology that makes modern transportation so efficient would kill us were not certain rules adopted and followed. Thus, vehicles are re¬stricted to certain locations. Similar directions of travel are put into the same corridors with some separation between those going in other direc¬tions. For safety, order and regulation is essential, even in societies that normally shun order and regulation.

With automo¬biles, we use traffic lights and signs to indicate who may go and who may stay, what can be done, or who has precedence over another. And we use turn signals and brake lights to tell others of our actions and intentions. In the case of brake lights, we signal actions as we carry them out: in the case of turn sig¬nals, we signal our intentions before we actually commit them into action. In either case, we allow others to know our future actions so that they can ensure that theirs do not conflict.

The brake lights of the automobile serve no function for the operator. Rather, they are a way to communicate with other drivers. The brake light means that the brakes are applied, which the other driver interprets to mean that the car is slowing down. Moreover, the other driver will usually search for a reason, some explanation of the brake lights. This means that the lights can serve a valuable communication purpose. On a long, normally non-congested highway, if the car in front of you applies its brake lights, it usually signals some unexpected obstruction or danger on the road, and it is usually wise to slow up and be more alert.

And here is where intentions come in. Social interaction is enhanced when the participants know not only what is happening at the moment, but what will happen. Of all the signals of the automobile, only the turn signals announce intentions.

Intentions are tricky, for they play many roles in social interaction, some obvious and necessary, some subtle and devious. In games we signal intentions in order to deceive. Of course, our opponent does the same, and interprets our signals knowing full well they may be deceptive. Thus starts the elaborate ruse and counter-ruse, where we try to determine how other people are reading our minds. Suppose that in a game I want to kick the ball to the right. I could first pretend to kick the ball to the left, but I know that they will know that this is a pretense. But if I pretend to kick to the right, they will expect me to kick to the left, unless they know that I will think that, in which case they may realize I really intend to kick to the right. So I decide to fake a kick to the right, and then run. Except that when the actual time comes, things may happen so fast that none of the plans can be executed. In games and war we signal intentions falsely more often than truly.

Imagine doing this in traffic, signaling a left turn, hoping that this will open up a hole in traffic that will let me dart to the right.

I once got a driver’s license in Mexico City, where aggression was the rule. But even here, intentions had to be signaled honestly. Above all, it was essential to avoid eye-contact with other drivers. In the traffic circles of the city, the trick was to avoid letting the other drivers see that you had seen them. Once the other drivers knew that you knew they were there, they would proceed at high speed around the circle, completely ignoring your presence, because they knew that you knew that they were there, so they expected you to stop or slow down. And you had to, or be killed.

On the other hand, if you could manage to avoid letting them see you see them, you could proceed with impunity, because now it was their responsibility to avoid you. If you collided, it couldn’t have been your fault, because after all, you didn’t see them.

Most places in the United States don’t let you get away with such games: in my community in southern California, for example, fault and blame are mechanically assigned according to strict orders of precedence: rules of the road determine whose responsibility it is to avoid accidents. Thus, at intersections, the automobile on the right has the right of way, and all the eye-contact in the world won’t change that.

In Mexico, there were other ways of signaling intentions. Thus, if two cars were approaching a narrow, one-lane bridge from different directions, the car that first flashed its lights thereby announced that it was coming through, so the second car had better yield. The flashing headlight was to be interpreted as, “I got here first, so keep out of my way.” As long as everyone understands such signals they work fine.

The problem is that other cultures completely reverse the meaning of the signals. In Mexico, one wins by aggression. In Britain, one wins by politeness and consideration. So in Britain, in a similar situation, the car that flashes its lights first is signaling, “I see you, please go ahead and I will wait.” Imagine what would happen when a Mexican driver encounters a British driver.

Drivers of automobiles get pretty good at reading the intentions of others. Brake lights and turn signals offer one method, these a formal, mechanized set of signals. Eye contact another. Headlights also serve a valuable communication purpose, with the flashing flights conveying many different messages, depending upon the circumstances. Horn blasts and hand signals are also used. Basically, any part of the automobile that other drivers will recognize as being under control of the driver can be used to signal something.

Notice that not all signals are truthful. Deceit exists on the highways just as much as in other social endeavors. The impatient driver can try a variety of tactics to gain open roadway, although the flashing of lights or blaring of the horn is the most common, most direct method. Some signals are ambiguous or confusing, such as the turn signal that continues for block after block, either signaling the eventual desire to turn or simply a remnant from a previous turn.

Turn signals are peculiar devices, neither human nor artificial. They are really not a way for the automobile to communicate with people. Turn signals are simply an aid to normal human-human communication: instead of shouting or pointing, one simply flips a small lever here, resulting in visible flashing lights there, outside the automobile. Even so, turn signals are an important start toward the graceful interaction of people and machines.

Graceful Interaction of People and Machines

Human social interaction has developed a rich assortment of methods to ensure social harmony. Every culture has developed means of maintaining politeness and courtesy, of communicating needs without offending. If machines are to interact gracefully, they too must follow these conventions.

Designers of machines usually provide the critical signals of the machine’s internal state for they know that maintenance is essential to operation. But then they often stop, failing to take into account the needs of the user of the device. As a result, the machines are still stuck in the a-social world of isolated devices. Worse, they have no manners. If machines operate in isolation with no need for interaction with people or other machines, then the lack of social graces and feedback about their internal states can be excused. But when machines are intended to operate with people, then the lack of socialization can lead to difficulties. Think of the telephone, continually intruding upon conversations, insensitive to the ongoing activities, forcing interruptions through its demanding ring whether the time is convenient or not. So it is with most machines, shunning interaction except to demand attention. We call such behavior in people “spoiled,” “arrogant,” or “insensitive,” but somehow we have accepted it from our machines.

Social cooperation requires more than letting others know your actions and intentions. It is also necessary to know how the others have received your communication: did they understand? Do they approve? Will they abide by it? When I talk with someone, I need to know how they are responding. Are they interested or bored, do they understand or are they confused? When people do joint activities, they need to agree upon the division of activities in advance in order that they can synchronize and coordinate their efforts, not conflict. And during the activities, knowledge of the other’s actions are important, if only to know that the other person is still interacting.

It really is essential to get some feedback, if only the “hmmm” form the other person. Otherwise how do I know they are attending? How do I know whether or not they are even alive?

We do have machines that are showing some of the first, early signs of social graces. Some can guide expectations, and even question actions. Thus, in the word processing system I am using to write this essay, if I try to do a complex operation, I might be warned:

“The current action cannot be undone: Do you wish to proceed?”

Or if I try to move a file from one location in the computer storage system to another, I am sometimes warned:

An item with the same name already exists in this location. Do you want to replace it with the one you’re moving?”

These are early signs of social maturity: polite, meaningful concern about the possible effects of the operation I requested be performed. Yes, some artificial devices show the early vestiges social responsibility, displaying their internal states for others to see, sometimes assessing the impacts of their own actions and warning others of them. The interaction is primitive, however, and often not as effective as one might expect from human colleagues. The subtleties and richness of natural emotions and natural attentiveness to social interaction are missing. But even so, the first glimpses of artificial systems that exhibit cooperative, social behavior are appearing.

Perhaps the simplest form of social cooperation among artificial systems is the “handshaking” protocols of communicating systems, invisible to the normal human user, but essential nonetheless. “Handshaking,” is, of course a human custom with a long cultural evolution. Today, shaking hands is part of the ritual by which people meeting for social or business purposes introduce themselves and get set for conversation or business. With machines, the term “handshaking” has been reserved for the initial steps of a communication protocol in which all the devices determine that they are connected properly, that their messages are in synchronization, and that they are directed to the correct recipients and in the correct format. It is easy to eavesdrop on these proceedings by listening to the first stages of a telephone connection of computer to computer or facsimile machine to facsimile machine. Better yet, listen as computer tries to talk to fax, or even to a person. Then the handshaking fails, as the automatic system tries this protocol and that before quitting and gracelessly hanging up the telephone.

It is a sign of our technological era, of course, that elaborate handshaking protocols and other social niceties have been developed to handle the interactions among machines, but that no such civility seems to have become standard for the interactions between machines and people.

Human emotions, facial expressions, and social interaction have evolved over millions of years. We have had time to do things slowly, to work things out with care. Even so, there are occasional mismatches as people fail to understand one another, fail to cooperate.

What will happen with our machines? In principle, artificial evolution can proceed much more rapidly than can biological evolution. Artificial evolution can take advantage of knowledge and experience, but so far there is little evidence of attention to these.

I fear that the rush to autonomous machines is proceeding too rapidly. Our machines are barely social now: they are still at an early stage of development, still primarily self-centered, still focused on their own needs and not those of their operators. What will happen when they are given more power, more authority? How can we shape the evolution of machines so that they become more humane, more in line with human needs and values?

Mind reading is an essential activity for social communication. If I am to interact with machines in a friendly, useful manner, then I need to be able to do the equivalence of reading the mind of the machine. Machines don’t have minds, but they do have internal states. More and more, they are able to have goals and plans, expectations and even desires. The interactions will be smoother and more friendly if I can know these things. Tell me, machine, is there enough tape to record that two hour show? If I move my computer file from one place to another, will I destroy something important? How about remembering that action so that tomorrow, when I no longer remember where I put it, you can help me? And, please, tell me what you are doing now.

The vervet monkey, the baboon, the chimpanzee, the ape, the human may all have developed deception to a great degree of sophistication, but I am not yet ready for this aspect of evolution to be copied by my technology. Let’s learn from the lessons of technology, but this does not mean that we must copy faithfully everything that we note. But some deception, some deceit might be valuable, just as sit can aid and smooth human interaction.

When I tell my computer system to get rid of some old files, maybe the system should lie, politely of course. A white lie, for my own benefit. I would think that it had destroyed my files, but it could secretly keep them in computer purgatory — neither active and available for use nor destroyed, just waiting around for a final decision. Never to be destroyed, at least not until there was no choice. When the system truly runs out of storage room, then let it destroy the oldest things from purgatory until there is enough room to continue. What good is this white lie? One day, when I suddenly realize that I need the item I had asked to be destroyed, why guess what, my friendly computer can smile and produce it. Smile? Why not?

A smile is an interesting concept. Could we copy it in our technology? We read one another’s faces and body language to understand their emotional state, their intentions, and their responses to our actions and words. Facial expressions are an interesting, accidental evolutionary development. An accident that life has capitalized on. A social accident. Little indicator lights scattered all over the face so the other person can read your mind, well, maybe not your mind, but perhaps your true feelings and beliefs. Sometimes even before you yourself are aware of them.

Facial expressions are a lot richer in information content than a few lights or sounds or meters. They result from a very complex mix of muscles, blood vessels, and body structures, controlled both subconsciously by neuro-chemical processes and consciously, deliberately and willfully. As a result, facial expressions are rich and varied. They reflect many subtle variations of mood. The blush that affects facial color, plus body position, and the sound of the voice, all give subtle indications of the underlying mood. Would that our machines were so sophisticated?

Just as people need to communicate acts, intentions, and emotional states, to give continual feedback and evidence of expected actions and outcomes, so too will machines have to interact more fully, more completely to provide the same kind of information. Will we have to repeat the whole ensemble of human emotional and facial expressions in our artificial devices? Yes, I think so. Technology recapitulates phylogeny.