View Chapter

Chapter 72 — Social Robotics

Cynthia Breazeal, Kerstin Dautenhahn and Takayuki Kanda

This chapter surveys some of the principal research trends in Social Robotics and its application to human–robot interaction (HRI). Social (or Sociable) robots are designed to interact with people in a natural, interpersonal manner – often to achieve positive outcomes in diverse applications such as education, health, quality of life, entertainment, communication, and tasks requiring collaborative teamwork. The long-term goal of creating social robots that are competent and capable partners for people is quite a challenging task. They will need to be able to communicate naturally with people using both verbal and nonverbal signals. They will need to engage us not only on a cognitive level, but on an emotional level as well in order to provide effective social and task-related support to people. They will need a wide range of socialcognitive skills and a theory of other minds to understand human behavior, and to be intuitively understood by people. A deep understanding of human intelligence and behavior across multiple dimensions (i. e., cognitive, affective, physical, social, etc.) is necessary in order to design robots that can successfully play a beneficial role in the daily lives of people. This requires a multidisciplinary approach where the design of social robot technologies and methodologies are informed by robotics, artificial intelligence, psychology, neuroscience, human factors, design, anthropology, and more.

Overview of Autom: A robotic health coach for weight management

Author  Cynthia Breazeal

Video ID : 558

This video presents an overview of Autom, a robot designed to serve as a personal coach for weight management during a longitudinal study. Fifteen robots were deployed over a period of two months and were compared to two other conditions: A computer coach with the same dialog (but no physical or social embodiment) and a paper log (standard of care). The primary question the study addressed was long-term usage and engagement as that is the most critical to keeping weight off. The hypothesis (verified by the longitudinal study) is that the physical-social embodiment makes a positive difference in people's sustained engagement, perception of their working alliance, and social support provided by the robot (than the other two interventions). People were more engaged with the robot than the other two interventions, and the emotional bond was notable in the robot modality and much less so in the other two interventions.

Nonverbal envelope displays to support turn-taking behavior

Author  Cynthia Breazeal

Video ID : 559

This video is a demonstration of Kismet's envelope displays to regulate turn-taking during a "conversation". In this video, Kismet is "speaking" with one person, but also acknowledges the presence of a second person. The robot is not communicating an actual language, so this video is more reminiscent of speaking with a pre-linguistic child. The nonverbal turn-taking behavior is what is being highlighted.

Learning how to be a learning companion for children

Author  Cynthia Breazeal

Video ID : 560

This video demonstration describes a project whereby we train a policy via learning-by-demonstration for a social robot to serve as a learning companion for young children during free-form educational play. Training data was captured during a Wizard-of-Oz paradigm where the robot played the color-mixing game app with 183 children. Once the model was trained on this data, we did a human-participant study with 85 children to compare the behavior and efficacy of the autonomous robot versus a Wizard-of-Oz-controlled robot. We also compared the children's behavior to just playing the game app without a robot learning companion. We found that the presence of the robot learning companion resulted in deeper exploration of the subject matter of the app (color mixing) and more behaviors targeted to this activity (e.g., there was more random tapping of the app when the robot was not present). The autonomous robot's behavior was not statistically different from the Wizard-of-Oz-controlled robot.

Social learning applied to task execution

Author  Cynthia Breazeal

Video ID : 562

This is a video demonstration of the Leonardo robot integrating learning via tutelage, self motivated learning and preference learning to perform a tangram-like task. First the robot learns a policy for how to operate a remote-control box to reveal key shapes needed for the next task, integrating self-motivated exploration with tutelage. The human can shape what the robot learns through a variety of social means. Once Leo has learned a policy, the robot begins the tangram task, which is to make a sailboat figure out of the colored blocks on the virtual workspace. During this interaction, the person has a preference for which block colors to use (yellow and blue), which he conveys through nonverbal means. The robot learns this preference rule from observing these nonverbal cues. During the task, the robot needs blocks of a certain shape and color and which are not readily available on the workspace, but can be accessed by operating the remote-control box to reveal those shapes. Leo evokes those recently learned policies to access those shapes to achieve the goal of making the sailboat figure.

Mental-state inference to support human-robot collaboration

Author  Cynthia Breazeal

Video ID : 563

In this video, the Leonardo robot infers mental states from the observable behavior of two human collaborators in order to assist them in achieving their respective goals. The robot engages in a simulation-theory-inspired approach to make these inferences and to plan the appropriate actions to achieve the task goals. Each person wants a different food item (chips or cookies), locked in one of two larger boxes. The robot can operate a remote control interface to open two smaller boxes, one containing chips and the other cookies. The task is inspired by the Sally-Anne false-belief task, where the humans have diverging beliefs caused by a manipulation witnessed by only one of the participants. The robot must keep track of its own beliefs, in addition to inferring the beliefs of the human collaborators, as well as infer their respective goals, to offer the correct assistance.

A learning companion robot to foster pre-K vocabulary learning

Author  Cynthia Breazeal

Video ID : 564

This video summarizes a study where a learning-companion robot engages children in a storytelling game over repeated encounters over two months. The learning objective is for pre-K children to learn targeted vocabulary words which the robot introduces in its stories. In each session, the robot first tells a story and then invites the child to tell a story. A storyscape app on a tablet computer facilitates the narration of the story. While the child tells his or her story, the robot behaves as an engaged listener. Two conditions were investigated where the robot either matched the complexity of its stories to the child's language level, or does not. Results show that children successfully learn target vocabulary with the robot in general, and more words are learned when the complexity of the robot's stories matches the language ability of the child.

Influence of response time

Author  Takayuki Kanda

Video ID : 806

This video illustrates the importance of response time in interactions with a social robot. In the first part of the study, it was revealed that it is hard to wait for more than two seconds. In the second part of the study, a technique to use a "conversational filler" is developed, which moderates the frustrations of waiting too long.

A scene of deictic interaction

Author  Takayuki Kanda

Video ID : 807

This video illustrates the "deictic interaction" in which the robot and a user interact using pointing gestures and verbal-reference terms. The robot has a capability to understand the user's deictic interaction recognizing both the pointing gesture and the reference term. In addition, there is a 'facilitation' mechanism (e.g., the robot engages in real-time joint attention), which makes the interaction smooth and natural.

An example of a social robot in a museum

Author  Takayuki Kanda

Video ID : 808

This video shows a scene of interaction between a social robot in a science museum and museum visitors. The science museum would be one of the appropriate places for such a robot, because a novel robot would attract visitors' attention to the robot, which would also contribute to the purpose of the museum, i.e., to help visitors better understand science. Further, a social robot can redirect visitors' attention to museum exhibits by explaining museum exhibits.

An example of repeated, long-term interaction

Author  Takayuki Kanda

Video ID : 809

This video shows examples of repeated interactions between a robot in a shopping mall and mall visitors. The robot was designed for repeated long-term interaction. It identified visitors using RFID tags and gradually exhibits friendly behaviors over time.