Towards Sociable Robots

Cynthia BreazealMIT Media Lab

Robotic Presence Group

DARPA/NSF Study on Human-Robot Ineraction

Outline Brief survey of human-robot interaction

research domains Design issues as applied towards sociable

robots Ingredients/Grand challenges of sociable

robots

DARPA/NSF Study on Human-Robot Ineraction

Humans in Hazardous environments

Robots as critical systems Pre-curser missions to characterize

planet Set up infrastructure for astronauts Assistance on Mars (carry tools, etc.) Life support on Mars

Morphology Vehicular/mobile/carry payload Humanoid/space shuttle/use

astronaut tools Balance of control

Tele-operation, VR Supervised from within Supervised face-to-face

JSC

JPL

DARPA/NSF Study on Human-Robot Ineraction

Human supervised teams Single human commanding many robots Balance of control

Heterogeneous vs homogenous teams Hierarchical? Autocratic? Democratic? Load balancing and task allocation Cooperative & distributed conrol (USC) Scaling to (very) large numbers

Interface issues Usability of software for tactical scenarios

(GaTech) Communication/interaction

DARPA—Mixed initiative control of autonoma teams

UCB

DARPA/NSF Study on Human-Robot Ineraction

Robotic-augmentation of humans

Robot as an “extension” of your body Enhance ability of a surgeon Robotic prosthetic, wheelchair Robotic exoskeleton

Interface issues Brain-machine interface (CalTech, Duke) Bio-mimetic robots

DARPA—Bio:Info:Micro

Intuitive Surgical MIT iBOT

DARPA/NSF Study on Human-Robot Ineraction

Robots in the human environment

Robots in the home, office, etc… Domestic assistants, healthcare,“smart” appliances,

entertainment, education… Robots that are a part of your everyday life

Autonomy in human social environment On the job learning Untrained users of different gender, age, culture, etc. Long-term relationship

Sony NEC

DARPA/NSF Study on Human-Robot Ineraction

Human-Robot Relationships

Smart tool (surgical robots) Complete independence (vacuum cleaner robot) Extension of you (robotic prosthetic) Commander/troops Pet owner/pet Master/servant Peer/colleague

DARPA/NSF Study on Human-Robot Ineraction

A Question of Interface (HCI)

Intuitive, natural interface for untrained users (Reeves&Nass) Humans are experts in social interaction

Humans automatically and unconsciously respond socially and naturally to technology.

If technology adheres to human social expectations, people find interaction enjoyable and feel empowered and competent

Holds for specialists and lay people

How does age, gender, culture, etc. impact this? How to measure, evaluate?

DARPA/NSF Study on Human-Robot Ineraction

What is a sociable robot?

To build robots that have a “living” presence, that people can interact with, communicate with, understand, and teach in human terms. Robots that support human social intelligence Robots that are socially intelligent themselves Self-motivated, pro-active creature, not appliance

Process of synthesis and iteration to come to a deeper scientific understanding of human sociability

DARPA/NSF Study on Human-Robot Ineraction

Design issues for sociable robots:Morphology

Match morphology to the task and environment Humanoid form

Send and receive human social cues in similar modalities Gaze direction Gesture Vocalizations/speech Facial expressions

Human engineered environment tailored to human morphology

CMU NEC Sony Sony

DARPA/NSF Study on Human-Robot Ineraction

Design issues for sociable robots:Appearance

Establish suitable social expectations Organic human faces difficult to achieve Balance familiarity vs “too plug-compatible” Biases interaction (dog-like, human-like, etc) Implied abilities (physical, cognitive, etc.)

Mechanical “cartoon” Anthropomorphic but creaturish Appeal and to portray youth

KSRP Waseda MIT SUT

DARPA/NSF Study on Human-Robot Ineraction

Design issues for sociable robots:Personality

Should the robot have a designed personality? Compatible with person’s personality, culture…(HCI) Encourages creature rather than tool-like interactions Encourage infant-caregiver interactions

Portrays youth and curiosity Simplest human-style interaction between human and robot Study in social development---humans must engage robot socially Benefits robot’s perceptual/behavioral limitations Benefits learning scenarios

DARPA/NSF Study on Human-Robot Ineraction

Design Issues for Sociable RobotsPsychological, etc. modeling

Science can guide design of perceptual, motivational (“drives” & “emotions”), cognitive, behavioral, and motor systems.

Match to human counterparts Find same things salient Perceive behaviorally relevant cues Recognizable behavior, expressions Predictable, understandable behavior, etc.

Creature-like autonomy, robustness, flexibility, adaptability

Understand natural systems (animals, people)

DARPA/NSF Study on Human-Robot Ineraction

Wolfe’s model, VGS2.0

Provides locus of attention to organize behavior Human and robot both find stimuli interesting Gaze direction is feedback cue to human

inh

ibit

rese

t

Frame Grabber

Eye Motor Control

Top down,task-driveninfluences

w w w w

skin tone habituationmotioncolor

attention

DARPA/NSF Study on Human-Robot Ineraction

Examples

5.856Overall

3.08face

5.08hand

6.58pink cupskin toned and movement

5.08black&white cowmovement only

6.08green cylinder

6.58multi-colored block

change in visual

behavior,

face reaction,

body posture

motion across

centerline,

shaking,

bringing object close

8.58yellow dinosaurcolor and movement

commonly read cues

commonly used cues

average time (s)

presentationsstimulusstimulus category

Matched to human Readable Understandable

DARPA/NSF Study on Human-Robot Ineraction

Design issues for sociable robots:Managing Interaction Complexity

Robots have limited perceptual, cognitive, behavioral abilities compared to people Imbalance in social sophistication between human and robot Tightly coupled and contingent interactions Mutually regulated Regulate interaction between robot and human

Role of “emotions” and “drives” with expressive feedback Modulate intensity of interaction Turn-taking with para-linguistic envelope displays Entrainment

DARPA/NSF Study on Human-Robot Ineraction

Video of turn-taking with envelope displays

DARPA/NSF Study on Human-Robot Ineraction

Entrainment and Regulation

Naive subjects Age from 25 to 28 All young professionals. No prior experience with

Kismet Video recorded

Turn-taking performance 82% “clean” turn transitions 11% interruptions 7% delays followed by

prompting Evidence for entrainment

Use shorter phrases Wait longer for response Wait for multiple phrases

37+8:06 – 8:43end @ 8:43

447:18 – 8:02

216:54 – 7:15

76:43 – 6:50start @ 6:43subject 2

37+17:30 – 18:07end @ 18:07

7016:20 – 17:25

1915:56 – 16:15

2115:37 – 15:54

1315:20 – 15:33start @ 15:20subject 1

subject 3

80+9:20 – 10:40end @ 10:40

458:25 – 9:10

587:18 – 8:16

186:58 – 7:16

536:00 – 6:53

245:30 – 5:54

155:08 – 5:23

104:52 – 4:58start @ 4:52

seconds between

disturbances

time stamp (min:sec)

DARPA/NSF Study on Human-Robot Ineraction

Need a multi-disciplinary community!

HCISCIENCE

ROBOTICS & AI

• Guide design of robot

• Understand human side

• Advance scientific understanding of both

• Human-centered design

• Measurements and Evaluation

• Usability• Teach-ability• Variation with

gender, age, culture, etc.

• Robotic design• Real-world autonomy• Task performance• Perception, Decision making,

Knowledge, Learning, Emotion, Personality, etc.

DARPA/NSF Study on Human-Robot Ineraction

Ingredients & challenges of sociable robots Life-like behavior

Real-world Autonomy Believability Commonsense

Human aware Perceiving people

Speech, gesture, expression,etc. Understanding people

ToM, empathy, story-based, BDI, etc. Being understood

Self understanding ToM, autobiographic memory, etc.

Readable Adhere to human ToM of robot

Socially situated learning Tutelage, imitation, goal emulation, training,

etc. Evaluation criteria Human psychology Ethics