A Survey of Socially Interactive Robots

1
A Survey of Socially Interactive Robots

• Terrance Fong, Illah Nourbakhsh, Kerstin
  Dautenhahn

Presentation by Dan Hartmann
2
Context - History

• The first work in social robotics involved
  stigmergy as a model for behavior in insect
  colonies
• Stigmergy was described to explain how social
  insect societies produce complex behavior
  patterns, from individuals performing simple ones.

3
Context - Societies

• Insect societies are anonymous, homogenous
  groups.
• Many animals form individual societies, where
  each member forms relationships and social
  networks

4
Context Breazeals Four Classes of Social
Robots

• Socially Evocative
• Relies on the human tendency to anthropomorphize
• Social Interface
• Provides a natural interface by employing
  human-like social cues

• Socially Receptive
• Socially Passive but benefits from interaction
• e.g. learning from demonstration
• Sociable
• Pro-actively engages with humans to satisfy
  internal social aims

5
Context Three More Suggested Classes

• Socially Situated
• Surrounded by a social environment, they must be
  able to distinguish between social agents and
  other objects.

• Socially Embedded
• Physically connected to a social environment
  requiring at least rudimentary social concepts,
  such as taking turns.

• Socially Intelligent
• Show aspects of human style social intelligence,
  based on deep models of human cognition.

6
Context Papers Scope

• This paper focuses on "peer-to-peer" human-robot
  interaction
• The underlying assumption is that humans prefer
  to interact with machines in the same way that
  they interact with other people

• Robots with human social characteristics
  including

7
Methodology Design Issues

• Natural human-robot interaction
• manifest believable behavior, establish
  appropriate social expectations
• Real-time performance
• Must operate at human interaction rates

• Readable social cues
• Must send signals to the human to provide social
  feedback.

8
Methodology Embodiment

• That which establishes a basis for structural
  coupling by creating the potential for mutual
  perturbation between system and environment"
• - Authors definition

9
Methodology - Embodiment

• Morphology
• Appearance biases interaction (e.g. a robot dog
  will be treated differently than an
  anthropomorphic robot)

• Design Considerations
• Needs enough humanness
• for user comfort
• Needs enough robot-ness to prevent false
  expectations of the robot's capabilities

10
Methodology - Embodiment

• Anthropomorphic
• Many argue that to interact socially with people
  a robot should resemble a human
• Caricatured
• Realism is not necessarily needed for
  believability.
• Functional
• The embodiment should reflect the tasks it must
  perform.

• Zoomorphic
• Most common are "pet" type robots
• Human-creature relationships are simpler than
  human-human relationships
• Easier to avoid the "uncanny valley in previous
  slide

11
Methodology Human-Oriented Perception

• To interact with humans in the real world, social
  robots must perceive the world the same way that
  humans do
• In particular, they must be able to track human
  features and interpret human communication
• Similar perception may require similar sensing

12
Methodology Human-Oriented Perception

• Each of these tasks is mentioned and references
  papers for in depth work.
• People Tracking
• Speech Recognition
• Gesture Recognition
• Facial Perception

13
Topics That I Am Skipping for Time and Relevance

• Dialogue
• Personality
• Emotion

• User Modeling
• Socially Situated Learning
• Intentionality

14
Discussion Attitudes Towards Robots

• Khan describes a survey to investigate peoples
  attitudes towards intelligent service robots. Two
  significant findings were
• A robot with machine-like appearance, serious
  personality, and round shape is preferred
• Verbal communication using a human-like voice is
  highly desired.

15
Discussion Field Studies

• Scheeff et al. conducted two studies to observe
  how a range of people interact with a
  creature-like social robot.
• Children were observed to be more engaged than
  adults.
• A friendly robot personality was reported to have
  prompted qualitatively better interactionthan an
  angry personality.

16
Discussion One Last Point for Perspective

• Paraphrasing Wood,the authors say
• Humans and robots must be able to coordinate
  their actions so that they interact productively
  with each other. It is not appropriate (or even
  necessary) to make the robot as socially
  competent as possible. Rather, it is more
  important that the robot be compatible with the
  humans needs, that it matches application
  requirements that it be understandable and
  believable, and that it provide the interactional
  support the human expects.