by David Levy
The participants were set to work on a problem-solving task commonly employed in experimental psychology, a task known as the Desert Survival Problem.* When the participants first attacked the problem, they would try to solve it by themselves, creating their own ranking for the survival items. They then went into another room, one at a time, where they worked on the task in collaboration with their assigned computer. They all exchanged information with their computer about each of the twelve survival items, and, if they wished, the participants could then change their initial rankings. Once the human participants had interacted with their computer, they would be sent into a third room, where they wrote out their final rankings and responded to questions about their interaction with their computer, questions such as “How similar was the computer’s approach to your own approach in evaluating the twelve items?” and “How helpful were the computer’s suggestions?”
The results of this experiment revealed a lot about how people perceive team relationships. When the humans believed that they were on the same team as the computer, they assessed the computer as being more like themselves relative to how much like themselves the participants thought the computers to be when the participants worked alone. These “teamed” participants also thought that their “teammate” computer had adopted a problem-solving style more similar to their own and that their computer agreed more completely with their own ranking of the items. Another tendency was for the teamed participants to believe that the information given to them by the computer was more relevant and helpful and that it was presented in a friendlier manner compared to the participants who did not believe they were members of a human-computer team—all this despite the fact that the information was identical and was presented in an identical manner in both cases. Other indications of relationship building between the human participants and their computers were that the teamed participants tried harder to reach an agreement with their computer on the rankings and were more receptive to their teammate’s suggestions and influences.
One of the most important conclusions of this study was to confirm the work of earlier psychologists who “have long been excited by how little it takes to make people feel part of a team, and by how much is gained when they do.” Reeves and Nass had extended this earlier research by showing that feelings of being part of a team are powerful enough to affect people’s interactions with computers, once they believe that their own success depends also on the success of the computer.
This research was groundbreaking work at that time, but even more remarkable than the ease with which their goal was accomplished was what the experimenters learned when they tested two simple computer personalities, each designed into a program that collaborated with a human user on the Desert Survival Problem. One of these computer personalities was “dominant,” using strong language in its assertions and commands, displaying a high level of confidence when communicating with the human test subjects, and leading off the dialogues with its human collaborators. The other computer personality was “submissive,” using weaker language, in which assertions were replaced by suggestions and commands with questions, and inviting or allowing the human collaborator to start each dialogue. It was found that those humans who themselves had more dominant personalities* enjoyed interacting with the dominant computer more than they did with the submissive one, while those with a more submissive personality preferred interacting with the submissive computer. Furthermore, not only did the human subjects prefer to interact with a computer similar in personality to their own, but they also experienced a greater satisfaction in their own performance on the problem-solving task when collaborating with the similar computer. These results led to the conclusion that not only do humans prefer to interact with other humans of similar personality, but they also prefer to interact with computers that have similar (virtual) personalities to their own.
Other experiments conducted by Nass and his group confirmed that humanlike behavior by a computer enhances the user’s experience of the interaction and makes the computer more likable. One example of this phenomenon is the ability of computers to increase users’ liking of them by means of flattery, by matching the users in personality, and through the use of humor, which has been found to lead to assessments of them as being more likable, competent, and cooperative than computers that do not exhibit any humor. Another example came from highly expressive teaching programs that were found to increase students’ feelings of trust in the programs because the students perceived them as helpful, believable, and concerned.
Designing Robot Personalities
Designing a robot with an appealing personality is an obvious goal, one that would allow you to go into the robot shop and choose from a range of personalities, just as you will be able to choose from a range of heights, looks, and other physical characteristics. One interesting question is whether it will be necessary to program robots to exhibit some sort of personality friction for us to feel satisfied by our relationships with them and to feel that those relationships are genuine. Certainly it would be a very boring relationship indeed in which the robot always performed in exactly the manner expected of it by its relationship partner, forever agreeing with everything that was said to it, always carrying out its human’s wishes to the letter and in precisely the desired manner. A Stepford wife. Perfection. No, that would not be perfection, because, paradoxically, a “perfect” relationship requires some imperfections of each partner to create occasional surprises. Surprises add a spark to a relationship, and it might therefore prove necessary to program robots with a varying level of imperfection in order to maximize their owner’s relationship satisfaction. Many people have relatively stable personalities and would therefore probably appreciate robots whose own personality and behavior exhibited some, but not a huge amount of, perturbation. This variable factor in the stability of a robot’s personality and emotional makeup is yet another of the characteristics that can be specified when ordering a robot and that can be modified by its owner after purchase. So whether it is mild friction that you prefer or blazing arguments on a regular basis, your robot’s “friction” parameter can be adjusted according to your wishes. Your robot will be programmed to recognize and measure friction when it is there, by the nature of your conversation with it and the tone of your voice, and to increase or decrease the level of friction according to your preferences.
One important consideration for robot programmers when planning a robot’s personality and behavior will be how best to cope with different cultures. Just think of the courting rituals and the chaperone phenomena in some Latin countries, the Chinese tendency not to be too physically demonstrative in public and the contrasting lack of inhibitions displayed in some other countries, and the tradition of arranged marriages in certain cultures—a tradition that ought to present no problem for robots, because the parents of the human bride or groom will simply make all the choices in the robot shop as to its physical appearance and other characteristics, rather than leave these decisions to their offspring. Whatever the social norms of the prospective owners and their culture, a robot will be able to satisfy them. Similarly with religion, the details and intensity of which can be chosen and changed at will—whether you’re looking for an atheist, an occasional churchgoer, or a devout member of any religion, you have only to specify your wishes when placing your order at the robot shop. The key here will be ensuring that the robot has a flexible personality. It will most likely leave the factory with a set of personality traits, some standard and others chosen by the customer, but a robot will be able to set any or all of these traits aside as required, to allow the robot itself to adapt to the personality needs of its owner.
The example of the dominant and submissive problem-solving programs devised by Nass and his team suggests that creating artificial personalities will probably not be an immensely difficult task for robot scientists. Likewise, the creation of blue eyes, a sexy voice, or whatever other physical characteristics turn you on, are all within the bounds of
today’s technology. And if what turned you on when you purchased your robot ten years ago no longer turns you on today, the adaptability of your robot and the capability of changing any of its essential characteristics will ensure that it retains your interest and devotion. When robots are able to exhibit the whole gamut of human personality and physical characteristics, their emotional appeal to humans will have reached a critical level in terms of attracting us, inducing us to fall in love with them, seducing us in the widest sense of the word. We will recognize in these robots the same personality characteristics we notice when we are in the process of falling in love with a human. If someone finds a sexy voice in their partner a real turn-on, they are likely to do so if a similar voice is programmed into a robot. If it’s blue eyes that one is after, simply select a blue-eyed robot when you make your choice. If it’s a particular personality trait, your robot will come with that trait ready-made, or it will learn the trait as it discovers its importance to you.
While much of the development work on the hardware for new robot technologies is being carried out in Japan, the West is not lagging behind in the research effort into software for the robots’ emotions and personality.* One reason for the Japanese bias toward hardware is because the Japanese government is determined to employ robots in the future to assist with the massive task of taking care of their aging population, a task for which the hardware must be totally reliable and robust. Another motivation for the Japanese investment in robot hardware research is that it will be the Japanese consumer-electronics conglomerates that will reap the greatest commercial benefits when robots are on sale to the public in high-volume quantities.
These world leaders in robotics, Japan and the United States, have somewhat different approaches and goals. The United States produces and uses far fewer robots than does Japan, because the United States is more reliant on less expensive immigrant labor. According to the latest industry figures in 2006, the United States had only 68 robots in manufacturing industries for every 10,000 human manufacturing workers, whereas Japan had 329 per 10,000. But an even greater distinction lies in the cultural differences between Japan and the United States and how these differences transfer to the different perceptions of the people in these countries to the prospect of our future with robots.
In an article in USA Today,† Kevin Maney summarizes these differences:
U.S. labs and companies generally approach robots as tools. The Japanese approach them as beings. That explains a lot about robot projects coming out of Japan.
A more detailed explanation of these cultural differences was given by the Economist magazine,‡ in an article entitled “Better Than People,” which explained “why the Japanese want their robots to act more like humans.” The article focuses on how these cultural differences affect robotics development in Japan. The reasons are partly economic—the huge growth predicted for the sale of service robots (to $10 billion) by the year 2015—but also cultural.
It seems that plenty of Japanese really like dealing with robots. Few Japanese have the fear of robots that seems to haunt Westerners. In Western books and movies, robots are often a threat, either because they are manipulated by sinister forces or because something goes horribly wrong with them. By contrast, most Japanese view robots as friendly and benign. Robots like people and can do good. The Japanese are well aware of this cultural divide, and commentators devote lots of attention to explaining it. The two most favored theories, which are assumed to reinforce each other, involve religion and popular culture.
Religion plays a role because Shintoism “is infused with animism: it does not make clear distinctions between inanimate things and organic beings.” For this reason the attitude in Japan is to question not why the Japanese like robots but why many Westerners view robots as some kind of threat. And this somewhat benevolent attitude toward robots has been enhanced by their popularity, both in newspaper and magazine cartoons and in films, ever since the launch of Japan’s robot cartoon character Tetsuwan Atomu in 1951.
Robot Chromosomes
A huge step forward on the path to creating robots with humanlike personalities and emotions has recently been taken by Jong-Hwan Kim* and his team at the Robot Intelligence Technology Laboratory in Daejeon, South Korea, who have been working on the development of successive versions of a robot called HanSaRam. In a 2005 conference paper, “The Origin of Artificial Species,” Kim and his colleagues describe the artificial chromosomes they have developed for robots.
The basis of Kim’s idea is that the entire collection of a robot’s artificial chromosomes will contain all the information about the robot that corresponds to the information stored in our DNA. Thus Kim’s programmed genetic makeup is modeled on human DNA, although instead of being a complex double-helix shape as in a human chromosome, each artificial chromosome is equivalent to a single strand of genetic makeup. In humans the principal functions of genetic makeup are reproduction and evolution, but in robots the makeup can also be used for representing the personality of the robot and can be electronically transferred to other robots.
Kim’s approach to robot personality was inspired by the evolutionary biologist Richard Dawkins, whose book The Selfish Gene asserts that, “We and other animals are machines created by our genes.” Kim draws a parallel between humans and humanoids by proposing that the essence of the origins of an artificial species such as humanoids must be the genetic code for that species. His paper presents the novel concept of the artificial chromosome, which Kim describes as the essence for defining the personality of a robot and the enabler for a robot to pass on its traits to its next generation, just as in human genetic inheritance. Thus the artificial chromosome creates a simulation of evolution for its artificial species.
If we think in terms of the essence of the creatures, we must consider this the origin of artificial species. That essence is a computer code, which determines a robot’s propensity to “feel” happy, sad, angry, sleepy, hungry, or afraid.7
Continuing the parallel between humans and humanoids still further, Kim suggests that the main functions of a robot’s genetic code are reproduction and evolution and that the code should be designed to represent all the traits and personality components of these artificial creatures. Thus his artificial chromosomes, being a set of computerized representations of a DNA-like code, will enable robots to think, feel, reason, express desire or intention, and could ultimately empower them to reproduce,* to pass on their traits to their offspring, and to evolve as a distinct species.
Kim’s team has designed fourteen robot chromosomes in all, six of which are related to the robots’ motivation, three to their homeostasis,† and four to their emotions. These chromosomes dictate how robots should respond to various stimuli: avoiding unpleasantness, achieving intimacy and control, satisfying curiosity and greed, preventing boredom, as well as engendering feelings of happiness, sadness, anger, and fear and creating states of fatigue, hunger, drowsiness, and so on, all of which will combine to imbue the robot with “life.” Kim’s robots will be able to react emotionally to their environment, to learn and make reasoned decisions based on their individual personalities.
For ease of development and testing, Kim’s simulated chromosomes have been programmed into a simulated creature—a software robot called Rity, living in a virtual world—that can perceive forty-seven different types of stimuli and is able to respond with seventy-seven different behaviors. As determined by their genetic codes, no two Rity robots react in the same way to their surroundings. Some become bored with their human handlers while others, because they have a different personality, pant and express their “happiness” at the sight of their humans. It’s all in their genes! One of the next steps by Kim and his team will be to create the equivalent of the human X and Y chromosomes, conferring on robots their own version of sexual characteristics, including lust. Thus if male and female robots like each other, “they could have their own children.”
Kim readily admits one of the principal messages of
the movie I, Robot—namely, that the feasibility of giving robots their own personalities and emotions might make them a danger to humanity. To counter this he suggests employing artificial chromosomes “to design brilliant but mild-tempered and submissive robots,” which is one way to ensure that we do not become enslaved by our creations as they evolve. Given this elementary precaution, by the time “malebots” and “fembots” are available for general consumption the market will be ready for them.
The Ten Factors as Applied to Human-Robot Relationships
We saw in chapter 2 how common it is for people to develop strong feelings of affection, including love, for their pet animals. And in chapter 3 we examined the same phenomenon as it relates to virtual pets such as the Tamagotchi. Now we come to examine the ten principal factors that cause humans to fall in love with humans, as discussed in chapter 1. Let us consider which of these factors might also be important in causing humans to fall in love with robots.
At the outset we should recall the importance of proximity and hence repeated exposure as major factors that contribute to placing people in a situation in which falling in love becomes more likely. In the case of a robot, both proximity and repeated exposure are easy to achieve, subject to the robot’s cost. Simply buy a robot and take it home and both of these criteria are instantly satisfied.
In chapter 1 we also discussed Byrne’s law, which shows that we are more inclined to like someone when we feel good. The empathetic robot, able to determine what makes a particular human feel good, will therefore have a head start in its attempts to seduce. The robot will do its best to create “feel-good” situations, perhaps by playing one of its human’s favorite songs or by switching on the TV when its human’s favorite baseball team is playing, and then it will exhibit virtual feelings that mirror those of the human, whether they be feelings of enjoyment when hearing a particular song or cheering on a baseball team.
