Can humanoid robots, attractive avatars, and other relational agents create the request level of trust and LLC the truthful, perhaps intimate or painful, disclosures often necessary to resolve a dispute or solve a problem? This article will explore these questions. Regardless of whether the reader is convinced that the demise of the human mediator or arbitrator is imminent, one cannot deny that artificial intelligence now has the capability to assume many of the responsibilities currently being performed by alternative dispute resolution (ADAIR) practitioners. Artificial intelligence can be embedded in a variety of physical forms.
This article will focus primarily on robots designed to assemble humans and avatars. Robots can, of course, assume whatever form the designer desires, including human, animal, abstract, or strictly functional (as might be seen in an industrial enterprise). Artificial intelligence, however, does not need to be defined by a physical form. Much of what will be discussed in this article will be relevant to, and include, devices that do not have presence in the physical world. L Avatars, for example, initially were regarded as a "graphic representation of a real person in cyberspace. 2 Virtual worlds such as Second Life, There, and Active Worlds are populated by millions of "residents," that being, individuals who direct their avatars in an essentially limitless number of interactions with other residents in a three-dimensional virtual world. The connection to an actual person once thought necessary is becoming less relevant, and the term "avatar" now includes non-player characters in three-dimensional online games and virtual online salesperson. 6 It is fascinating (and perhaps unsettling) to realize the complexity and seriousness of tasks currently delegated to avatars and robots.
This article will review some of those legations and suggest how the artificial Intelligence developed to complete those assignments may be relevant to dispute resolution and problem solving. Relational agents, which can have a physical presence such as a robot, be embodied In an avatar, or have no detectable form whatsoever and exist only as software, are able to create long term socioeconomic relationships with users built on trust, rapport, and therapeutic goals. 7 Relational agents are interacting with humans in circumstances that have significant consequences in the physical world.
These interactions provide nights as to how robots and avatars can participate productively in dispute resolution processes. Artificial intelligence has two complementary components: the difference between these two components is similar to "the difference between [an] adverb and [a] noun. "9 In other words, a device can either behave intelligently as a result of automated or human-controlled directions, or a device literally can be intelligent in the sense that it requires no external influence to direct its actions. 10 The more readily achievable goal is to create a device that behaves intelligently.
Because we believe that humans are the most intelligent species, it should not be surprising that a significant amount of artificial intelligence research concerning this first goal involves devices that resemble humans-?specifically, robots. Robotics scientists and specialists are creating physical representations of human beings that mimic our movements and even our appearances. Al Robots are being developed that replicate human appearance and movement surprisingly accurately. But simply creating a realistically behaving robot or avatar may not be sufficient.
Will avatars ND robots truly be able to engage humans? Or instead, will the prospect of interacting with lifeless entities feel so unnatural that artificial intelligence devices will not be able to encourage the conversations and disclosures necessary for successful dispute resolution and problem solving? Studies have concluded that persuasive dialogues with computer agents can change attitudes. 12 Results based on interactions in situations other than ADAIR suggest that avatars and robots acting as relational agents also are capable of behaviors that will facilitate dispute resolution and problem solving.
The more difficult, more exciting, and perhaps more troubling goal is the second one. Can we create devices that actually are intelligent and, if so, what role can those devices play in dispute resolution and problem solving processes? Can human mediators and arbitrators be replaced by robots and avatars that not only physically resemble humans, but also act, think, and reason like humans? 13 And to raise a particularly interesting question, can robots, avatars and other relational agents look, move, act, think, and reason even "better" than humans? L.
But What Does "Better" Mean? Better" is a seductive term that demands an assessment and comparative ranking, yet has no apparent objective standards or moral component-?"better" in what sense, according to whose Judgment, and based on what values? When considering potential applications for artificial intelligence devices, one must keep in mind that devices can be created that could result in a loss of human control over both specific, discrete human interactions as well as computer-based programs that support a rapidly increasing share of society's workload. 4 Is this beginning to sound like the beginning of a bad science fiction novel? You wish. In 2009, the Association for the Advancement of Artificial Intelligence met in Sailor, California to debate whether artificial intelligence research should be limited. That location was chosen purposely to evoke the 1975 world-leading biologists' meeting held at that same location to discuss the newly discovered ability to reshape life by trading genetic material between organisms. 1 5 That meeting led to the discontinuation of certain experiments and new guidelines for recombinant DNA that allowed experimentation to continue. 6 At the 2009 conference, scientists looked closely at artificial intelligence systems that communicate empathy to medical patients. This particular artificial intelligence. 17 It is important to note that these are the same types of systems presented later in this article as prime examples of how far artificial intelligence devices have advanced and how valuable these devices will be for ADAIR. One of the scientists' concerns intersects with an implicit theme in this article.
Artificial intelligence devices are proliferating and, like it or not, increasingly will become a greater part of dispute resolution and problem solving processes. 18 In our everyday lives we will be forced to live with artificial intelligence devices that realistically mimic human behaviors. 19 These interactions will raise socioeconomic, legal, and ethical issues, and humans will have to think about the consequences of interacting, for instance, with a device that is as intelligent as, and perhaps even more empathetic than, our spouses. 0 Will artificial intelligence devices become even more intelligent than human beings? Some scientists believe that this type of "intelligence explosion" will occur, and that smart machines in turn will develop even ore smart machines until we reach the end of the human era. 21 A reassuringly contradictory point of view, however, is that "[u]until someone finds a way for a computer to prevent anyone from pulling its power plug, ... It will never be completely out of control. "22 The predictions and suggestions in this article do not look quite so far ahead.
This article discusses artificial intelligence devices that exist today, or at least will exist very soon, and suggests how these devices can be integrated into ADAIR processes. Some of the worrisome consequences of using artificial intelligence devices will be addressed, but extensive discussion about the potentially dangerous consequences of employing artificial devices that actually are intelligent goes beyond the scope of this article and must be reserved for another day. But make no mistake.
If one accepts the proposition that parties should have significant control regarding the nature of their ADAIR processes, then parties being encouraged (or forced) to live with artificial intelligence in their everyday lives will become more comfortable and familiar with these devices and eventually will expect ND demand that these devices be included in dispute resolution and problem solving processes. II. What is Necessary for Robots and Avatars to Interact Effectively with Humans? There are many ways to organize a discussion about the contributions that robots and avatars can make to dispute resolution and problem solving.
This article divides that discussion into the two components described in the introductory section. It first addresses the question of how intelligently robots and avatars can behave today in light of scientific advances, and the article then asks whether, and to what degree, a boot or an avatar can be described as intelligent. Although organizing the discussion in this manner certainly is helpful, more must be done at the outset. This article argues that robots and avatars can perform, at least for some purposes, as effectively as human mediators.
To make that case it is necessary to identify the capabilities considered essential for effective human interaction and to then assess the degree to which robots and avatars possess those characteristics. This subsection summarizes a mainstream theory as to what capabilities are essential for human interaction. The article subsequently provides numerous examples of robots and avatars interacting with humans and fulfilling delegated duties. These examples fully capable of effective human interaction.
When considering whether a robot or avatar can act as a surrogate for a human mediator, it is logical to assume that the robot or avatar must be able to replicate human physical and intellectual processes precisely. And if the goal is to provide a substitute for a human mediator that literally is as similar as possible-?in effect a twin for that human-?then this concern is well- founded. But there is an important caveat. Artificial intelligence may not need to mimic human appearance, movement, and cognitive processes in order to achieve desired results.
If the goal is not merely to duplicate the performance of a human mediator but instead to exceed, or even simply supplement, that performance, then it may prove counterproductive to design a robot or avatar that is a mirror image of a human. Artificial intelligence that is embodied in a physical form very different from a human, or that does not assume any form at all but instead exists merely in a "cloud," may be able to engage a human party who refuses to, or is unable to, engage tit another human (at least at this particular point in time).
Variables that include the personalities of the parties, the parties' present physical and emotional circumstances, the relationship among the parties, and the parties' comfort level with technology are among the factors that will determine whether it is most advisable to introduce artificial intelligence into a dispute resolution process in the form of a very realistic humanoid robot. With that caveat in mind, there remains great value in exploring the question of whether a human mediator's place at the proverbial ideation table can be assumed by a humanoid robot.
The most recent generation of robots and avatars has four critical human capabilities: "engagement, emotion, collaboration, and social relationship. "23 First, the article will discuss what is meant by these terms. Later subsections will provide examples of robots demonstrating these capabilities. Engagement refers to the ability to initiate, maintain, and terminate a connection to another individual. 24 As suggested earlier, in order to engage, a device must behave intelligently.
The direction of the eyes, the nod of the dead, hand gestures, body position, the delay before response, the determination of when to interrupt-?these cultural cues have been carefully studied and deconstructed, and, as a result, it increasingly is possible for robots and avatars to connect with humans by using these cues. 25 Emotions play a major role in human behavior and are critical when it comes to initiating and sustaining relationships.
Emotions can create obstacles to problem solving by diverting attention from substantive issues, damaging relationships, or providing opportunities for exploitation. 26 But emotions also can be a valuable asset, providing motivation, enhancing relationships, and making it easier to listen and learn. 27 Researchers are developing computational theories of emotion that allow robots and avatars to interact emotionally with humans, concluding that emotions are closely intertwined with cognitive processing "both as antecedents (emotions affect cognition) and consequences (cognition affects emotions). 28 In order to interact with humans, a robot or avatar must recognize and understand cues such as facial expressions, gestures, and voice intonation and, in turn, convey information about its own emotional state by using appropriately responsive cues. 9 Collaboration is, of course, a term that is near and dear to the hearts of dispute resolves and problem (and possibly other robots and avatars) to achieve a shared goal. 30 Collaboration is a higher level process dependent on engagement, but the relationship is not strictly hierarchical. 1 The progress of the collaboration can affect how engagement behaviors are interpreted because, for example, failure to make eye contact when the collaborators both are focusing on a document will not indicate intent to disengage. 32 Social relationships between humans and robots or avatars to date have been worth-term with an immediate collaborative goal such as shopping or entertainment. 33 But that is changing. A social relationship is an extended engagement that may be necessary, for instance, to address issues that require behavioral modification such as weight loss and substance abuse. 4 Domestic relationship and separation issues, for example, are often a subject of mediation and may require changes in behavior. A social relationship can improve collaboration and thus increase the chances of achieving a desired goal. 35 This brief discussion of these capabilities will make it easier to appreciate and understand the sophistication of the artificial intelligence devices described below. And it certainly helps us understand what will be necessary for an artificial intelligence device to replace a human in a collaborative process.
But again, please keep in mind that these characteristics will not be necessary in every circumstance and, in fact, are not present in all of the following examples. The fact that an artificial device does not have all the qualities necessary for an extended human interaction does not alter the fact that the device still may be able to accomplish a specific goal. And the fact that an artificial device does not replicate a human precisely may lead to more productive human interactions in certain situations.
Ill. The Adverb: Robots Behaving Intelligently In order to determine how behavioral artificial intelligence devices can be integrated into dispute resolution and problem solving processes, it will be helpful to explore how those devices are being used in other contexts. Although some of the current applications are not immediately transferable to ADAIR, they do illustrate the state-of- the-art for artificial intelligence and may suggest potential applications.
One application that certainly deserves close examination is robotic technology. Robotic technology represents a type of artificial intelligence that has intrigued both scientists and the public at large for generations. 36 On the one hand, scientists are driven by intellectual curiosity and professional demands to discover new information and tools that explain and simplify our lives. 37 The public, on the other hand, often is inspired by popular culture that romanticizes the possibilities of the future. 8 Whatever the reason behind the fascination, however, one thing is apparent-?robotic technology already is part of contemporary modern life and it quickly is becoming even more integral. 39 Robots can present a variety of appearances that range from shockingly lifelike to fatalistically hybrid human-mechanical. Carnegie Mellon University Valerie and the Naval Research Laboratories' George, for example, present only a human face on top of a generic, metallic, cylindrical mobile base. 40 In contrast, Gemini closely replicates human appearance and movements. 1 The European Union's CAST robot has a small (allowing for collaboration on assembly tasks) and the Massachusetts Institute of Technology Media Lab's Leonardo closely resembles a small animal with significant expressive capability, particularly in its face. 42 Mel is a penguin designed for "hosting. "43 Mel resembles a stuffed animal and has a movable head, beak, and wings on top of a mobile base. 44 He can guide and inform humans about environments such as stores and museums, and supervise human actions with objects found in those environments. 5 Using algorithms for initiating, maintaining, and terminating conversations, Mel has demonstrated that he can follow a humans face and gaze, and also look and point at shared objects at appropriate times. 46 Mel an nod his own head, recognize human head nods, converse about himself, participate in collaborative demonstrations, locate a human in an office environment, engage the human in a conversation noting where that person is looking and the time that passed since the person last spoke, and respond to human cues signaling a desire to disengage. 7 Humans respond when Mel tracks their faces, returning Meld's gaze, and they nod more frequently at Mel when he recognizes their nods. 48 Numerous examples from various disciplines and professions demonstrate how robots can be used. If the health sciences, for instance, find it productive to use boots when a patient's life, or at least his or her health and well-being, literally may be at risk, then certainly there is a role for robots in ADAIR. Psychologists, for example, are using this form of artificial intelligence to achieve fairly sophisticated interactions with young patients suffering developmental disorders such as autism. 9 In this particular application, intelligent technology is embedded in a social robot, an electronic device with humanoid or other "creature-like" characteristics. 50 These robots are programmed to interact with children in a manner that replicates human interaction "by exchanging a variety of social cues, such as gaze direction, facial expression, and visualization. "51 Notably, social robots have been able to elicit desirable behavior from autistic children that those children typically do not demonstrate in their daily lives. 2 Not only have many of these children interacted directly with the robot to a greater degree than they have interacted with humans, they also have relied on the robot to facilitate interaction with third parties. 53 Thus, through the use of robotic technology, psychologists increasingly are able to achieve hermeneutic results that otherwise would be difficult to obtain. 54 If robots can elicit positive and desirable responses in this therapeutic context, then certainly one can imagine dispute resolution or problem solving circumstances where a social robot might encourage a productive response even though traditional attempts have failed.
Similarly, there is increasing interest in using social robots to fulfill the healthcare needs of an aging population. 55 The objective in this case is to create a robot that not only serves a utilitarian purpose, but also provides a "hedonistic" experience. 6 The fact that robots can both provide this relatively high level of social experience, and also be perceived as something more than a piece of equipment, suggests that robots may be able to collect information from a party frankly too frustrated to communicate directly with other humans.
One team of robotics is fine-tuning a robotic caretaker to work with the elderly in their homes, providing services and companionship that will enable aging people to retain greater independence for a programmed to fit the specific needs of its owner, such as assisting a visually impaired owner with navigation around the house or reminding a cognitively impaired owner to take medication, while simultaneously providing basic social interaction. 8 As can be the case with other robotic technology applications, the advent of social robots to care for the elderly also eases the strain on a limited labor pool. 59 The percentage of older adults in the United States is rapidly increasing and will more than double between the years 2010 and 2030. 60 Although older adults may suffer cognitive impairments as they grow older, many retain the ability to engage in face-to-face conversations. Because face-to-face conversation is multimode (verbal, nonverbal, and vertebral behavior), allows for repetition and clarification, and has mechanisms to help focus participants' attention, individuals with impairments still may be able to communicate face-to-face using methods that remain available. 62 Believing it is necessary to establish social and emotional ties in order to motivate older adults over extended periods of time, researchers created an avatar relational agent to interact with older adults (aged 62 to 84) every day for two months in an effort to encourage those adults to exercise by walking. Although the researchers would have liked to use unconstrained speech input, concern about speech recognition software and natural language understanding led the researchers to offer on-screen multiple choice responses, dynamically updated throughout the conversation. 64 Users were given touch-screen monitors and results were compared with individuals given only pedometers and educational pamphlets. 5 The vast majority of respondents interacting with the relational agent looked forward to those interactions (75%), found the system easy to use, and perhaps most importantly, registered a statistically significant increase in recorded pedometer steps as compared to the control group. 66 Reporting they would like to continue interacting with Laura (the avatar),67 users indicated that although the relationship initially was strange, by the end of the two month period they liked, trusted, and even cared for Laura. 8 Several users even reported that they felt that Laura also liked and cared about them. 69 Note, however, that in a second study the researchers found that although the dialogues were written to provide significant variability in each says interaction, most participants found the conversations repetitive at some point and consequently, many lost their motivation to follow the relational agent's advice. 70 One study participant remarked, for example, that "it would be great if Laura could just change her clothes sometime. 71 The researchers then designed a virtual laboratory to further explore long-term human-virtual agent relationships, and their first study evaluated how the perception of agent repetition impacts adherence to a health behavior modification program. 72 This study, which involved only twenty-four artisans and produced admittedly preliminary results, concluded that there is a negative effect as dialogue variability declines. 73 Participants' performance relative to their walking goals decreased significantly over time when perceptions of repetition increased.
These observations certainly are not surprising and serve as reminders that, as with human-to-human interactions, variability is a productive (and even essential) attribute for engagement. One cannot expect parties involved in a problem solving process to continue to be engaged with a relational agent that falls he most sophisticated relational agent will have a diminished capacity to provide conversational, emotional, tonal, facial, and physical responses as compared to a human. Consequently, it is particularly important to ensure that a relational agent does not fall into a discouragingly predictable pattern.
Avatars have been used successfully in other health care contexts. Two empathetic middle-aged avatar discharge nurses, one African-American and one Caucasian, were created to help hospital patients with low health literacy read and follow directions. 74 Understanding the value of multiple modalities for communicating health care information, the virtual nurses were given the ability to hold and point at an image of each patient's After Hospital Care Booklet (CHAP), providing verbal explanations while the patient followed along in a paper copy with explicit instructions as to when to turn the page. 5 The nurses spoke with the patients once a day every day, used a short "open book" quiz format to confirm patients' understanding, and alerted human nurses to intervene if a patient failed a quiz a second time, even after the virtual nurse guided the patient to where the correct answers could be found in the CHAP. 6 The system thus offered an intuitive conversational agent interface, redundant modalities for communicating medical information (screen images, printed text, and synthetic speech), and comprehensive checks. 7 Recognizing the importance of caring, empathy, and good "bedside manner," the nurses' informational dialogue was augmented with relational dialogue and relational behavior. 78 The nurses (who traveled around the hospital on a mobile kiosk), addressed patients by name, began every interaction with a social chat, used appropriate humor, offered feedback at every empathetic opportunity, and referred to information discussed in previous interactions in an attempt to foster continuity. 9 Forty-nine patients aged 20 to 75 found the system very easy to use after less than a minute of training, reported high satisfaction, expressed few reservations about receiving medical information from an avatar, and stated that they would follow the nurses' directions. 80 In a second study, seventy-four percent of hospital patients stated that they actually would prefer to receive discharge directions from the virtual nurse rather than a doctor or a human nurse. 1 Patients reported that they did not receive enough time and attention from either the doctors or hospital nurses and very much appreciated that fact that the avatar nurses would spend whatever time was necessary to ensure that the patients understood the directions. 82 The hospital patients, who typically are entirely submissive and completely dependent on hospital staff, felt empowered and less helpless because they understood relevant medical information that allowed them to be more actively involved in their own health care. 83 Empowered?
Less helpless? More actively involved in the resolution of their problem? Mediators often work long and hard to assist parties to achieve these results. In fact, "[I]n a transformation approach, empowerment and recognition are the two most important effects that mediation can produce, and achieving them is its most important objective. "84 If avatars can help achieve results like this when a patient's life literally may be at risk, then it frankly is absurd to claim that avatars have no role to play in dispute resolution or problem solving.
A medical research team at Carnegie Mellon University has demonstrated that artificial, robotic intelligence can accomplish tasks that can perform minimally invasive surgical procedures without significant sorption of internal organs that a human surgeon simply cannot replicate. 86 Controlled with a Joystick and designed with multiple Joints that adjust automatically to maneuver through the intricate pathways of the human body, the robot mimics the natural movements and biological structure of a live being-?in this case, a snake-?to accomplish its goal while reducing the risks and complications associated with traditional surgery. 7 Granted, although the application of snake-like mobility to a dispute resolution process may not be immediately apparent, this example illustrates hat in certain situations robotic devices can accomplish what humans cannot. The fact, however, that this robot can be controlled so precisely confirms that the facial expressions and movements of a human-like robot also can be controlled to replicate those of a human too very precise degree. The United States Armed Forces are well aware of robots' potential applications.
Robots can be dispatched, for example, into areas too dangerous for human personnel. 88 The same research team that created the surgical snake was enlisted to design and build a robot paramedic that can initiate diagnosis of a wounded soldier's condition before human paramedics are able to remove him safely from the battlefield. 89 The fact that the same research team that designed the snake was involved in this wartime application illustrates the flexibility and adaptability of robotic technology.
The robot also can be used to "assess [a soldier's] injuries as he's being carried to a safe location," thereby enabling the paramedics to concentrate on transporting the patient while helping them avoid further casualties. 90 The ability to make a diagnostic assessment, obviously, is an invaluable example of artificial intelligence. Perhaps not surprisingly, the military also has used robots to conduct operations and inflict injury on opposing forces. 1 One such robot is the unmanned ground vehicle (SUB), a device controlled remotely and-?like the robotic-?used to perform duties that would be significantly more dangerous for a human soldier to fulfill. 92 In fact, American forces currently use an estimated six thousand Gives in the Middle East and according to one report, "the military goal is to have approximately 30% of the Army comprised of robotic forces by approximately 2020. "93 To combat the ethical concerns prompted by such a vision, scientists developing the technology intend to maintain the "chain of command" between robots who gather information and humans who act upon it. 4 The ethical concerns raised by artificial intelligence are complex and deserve their own dedicated discussion. Clearly, the ethical concerns will be dramatically increased when discussing artificial intelligence that is intelligent, the second form of artificial intelligence described in the introduction. But even when the discussion is limited to devices that only behave intelligently and must rely on external direction, one still just be vigilant and monitor the ways in which the device is being controlled.
The armed forces also have high hopes for the use of robotic insects to conduct reconnaissance and emergency rescue missions. 95 Unlike the robots described above, however, the robotic insects being developed actually are more appropriately understood as cowboys-?part animal, part machine. 96 Because "beetles and other flying insects are masters of flight control," research scientists have decided to integrate the innate biological abilities of these creatures with artificial intelligence genealogy would allow infiltration and observation of hostile territories with little risk of detection. 8 Furthermore, if a cyber beetle were intercepted, the ramifications would be significantly less severe than if a human operative were captured. 99 While military and medical applications might be an obvious step in the march of technology, one might be surprised to learn the speed at which robotic technology is being applied in the commercial sector. 100 Robotic farmhands, for instance, have been designed to combat "a lack of labor availability in a sector reliant on intense rusts of tough, seasonal work. 101 Farmers can produce crops more efficiently and economically because these devices eliminate human error and increase the rate at which difficult tasks can be accomplished. 102 Similarly, robots can be used in the construction industry to perform tasks that are extremely dangerous for human workers to perform, "such as inspecting high-rises or underwater bridge piers. "103 Because these robotic technology applications eliminate risks associated with manual labor, they likely will reduce costs for consumers when widely adopted. 04 Perhaps even more interesting, however, is the growing number of robots within the home. 105 Machines such as the robot Room, "an autonomous and mobile vacuum cleaner robot that is affordable, has effective utility, and is a commercially successful product," are the tip of the iceberg for the typical household of the near future. 106 The value of domestic robots is being recognized at an exponential rate: a 2002 survey conducted by the United Nations determined that "the number of domestic and service robots more than tripled [over the previous year], nearly outstripping heir industrial counterparts. 107 Regardless of the preceding paragraphs, one still may not be able to imagine how robots can be integrated into dispute resolution and problem solving processes. Specifically, it may be difficult to believe that living, breathing human parties will be able interact with robots as comfortably and easily as they interact with other humans. But as humans become more accustomed to automated interactions within their homes, they also will become more comfortable interacting with robots outside of their homes in a variety of contexts.
Furthermore, arties will come to expect the convenience and efficiency robots can provide. 108 IV. Appearances Matter In an effort to make interactions with robots and other forms of artificial intelligence feel more natural and comfortable, many scientists now are focusing on device design and mechanics. 109 These developers believe that the more realistic and lifelike a social robot appears and behaves, the more easily it will be able to establish "rapport" with human beings and the more likely it will be able to achieve the desired result. 110 The need to create robots that will be accepted raises unique challenges.
Some scientists are focusing on the robots themselves-?exploring ways to simulate human biological structures and physiological systems in an effort to create more physically intelligent robots. All Others are focusing on the humans-?analyzing and applying social psychology theories to understand the ways in which humans learn, as well as the ways in which they adopt and interact with new technology. 112 Both lines of research require scientists to consider the context in which a given robot is to be used, the purpose for which the robot is intended, and the target user for which