Upload
nicolette-merrihew
View
215
Download
0
Tags:
Embed Size (px)
Citation preview
Design and evaluation of a robotic TV-assistant:
balancing personality and control
Drs. B.W. Meerbeek, PDEng. (Philips research)Drs. H.C.M. Hoonhout (Philips research)
Ir. P. Bingley (Philips research)Dr. J.M.B. Terken (TU/e)
First iCat user workshop – 27th March 2006, Eindhoven
First iCat users workshop 27-03-2006
Design and evaluation of a robotic TV-assistant 2 / 15
Overview
• Introduction• Research questions and hypotheses• Designing personality and levels of
control• Experiments in HomeLab• Main findings
First iCat users workshop 27-03-2006
Design and evaluation of a robotic TV-assistant 3 / 15
Background• More robots in domestic environment (UNECE)
– Human-robot interaction more important• Media Equation (Reeves & Nass)
– Social response encouraged if machine:• uses full-sentence text or speech• demonstrates task knowledge as well as reasoning• can act autonomously on behalf of the user
– Human-robot interaction ≈ human-human interaction – People assign personality to robots (a.o. Dautenhahn)
First iCat users workshop 27-03-2006
Design and evaluation of a robotic TV-assistant 4 / 15
Personality and Control• Personality of a machine can help users to form
mental model and to interpret and predict its behaviour (Norman)– What personality do users prefer?
• As machines (personal robots) become more intelligent they can take over more tasks from humans– Who should be in control?
First iCat users workshop 27-03-2006
Design and evaluation of a robotic TV-assistant 5 / 15
Context• iCat, platform for human-robot interaction research
• Application: TV-assistant
• Facial expressions• Head movement• Multi-colour LED’s• Camera (“see”)• Microphone (“hear”)• Touch sensors (“feel”)• Speakers (“speak”)
First iCat users workshop 27-03-2006
Design and evaluation of a robotic TV-assistant 6 / 15
Research questionsPreliminary question:• Is it possible to convey personality with a robot?
And if so,• What personality do users prefer?• What level of control do users prefer?• Is there an interaction between preferences for
personality and for level of control?
First iCat users workshop 27-03-2006
Design and evaluation of a robotic TV-assistant 7 / 15
Expectations
• No main preference for personality or level of control
• Personality should match the level of user control
Low preference
High preference
High preference
Low preference
Introvert, formal conscientious
robotExtravert, agreeable,
friendly robot
Low user control
High user control
• Extrapolating from observations of human-human interaction:
First iCat users workshop 27-03-2006
Design and evaluation of a robotic TV-assistant 8 / 15
Personality design (1)Extraversion
- Agreeableness
+ Conscientiousness
+
Somewhat introvert
Modest
Somewhat shy
Expectantly
Polite
Distant
Formal
Orderly
Serious
Careful
• Facial expression, e.g. eyes move away• Motion, e.g. head to chest• Speech, e.g. low fundamental frequency (pitch)• Linguistic style, e.g. use more formal words
Catherine
First iCat users workshop 27-03-2006
Design and evaluation of a robotic TV-assistant 9 / 15
Personality design (2)
Lizzy
Extraversion +
Agreeableness +
Conscientiousness -
Extravert
Talkative
Enthusiastic
Takes the initiative
Assistantly
Empathic
Warm-hearted
Amicable
Jovial
Impulsive
Somewhat careless
Light-hearted
• Facial expression, e.g. more eyebrow movement• Motion, e.g. faster head movements• Speech, e.g. faster speech rate• Linguistic style, e.g. chitchat
First iCat users workshop 27-03-2006
Design and evaluation of a robotic TV-assistant 10 / 15
Levels of control• Low user control
– System-driven interaction style– Natural language dialogue
• High user control– User-driven interaction style– Command and control
First iCat users workshop 27-03-2006
Design and evaluation of a robotic TV-assistant 11 / 15
Movie• TV-assistant recommends programme to user
Second clip
First clip
Catherine
Lizzy
Low user control
High user control
Movie available on community website:http://www.hitech-projects.com/icat/download.php
play
First iCat users workshop 27-03-2006
Design and evaluation of a robotic TV-assistant 12 / 15
Experiments• Preliminary experiment demonstrated that
personalities were recognized as intended
• Main experiment– Combinations of two personalities and two
levels of control (2x2-design)– 32 participants in experiment of 2 hours– Personality assessment– User preference
First iCat users workshop 27-03-2006
Design and evaluation of a robotic TV-assistant 13 / 15
Summary of main findings• Overall, there seemed to be a preference for Lizzy• In initial interactions, low user control slightly
preferred over high user control• In longer term, preference may shift from low to
high control
• Lizzy seemed most preferred with low user control• Catherine matched better with high user control
than with low user control
• People perceived more control with Lizzy than with Catherine (identical level of objective control)
First iCat users workshop 27-03-2006
Design and evaluation of a robotic TV-assistant 14 / 15
Discussion• Personality perception less clear in main
experiment– Distributed attention
• Limited functionality of TV-assistant• Relatively short interactions • Single user
First iCat users workshop 27-03-2006
Design and evaluation of a robotic TV-assistant 15 / 15
Thanks for your attention
www.research.philips.com/[email protected]
First iCat users workshop 27-03-2006
Design and evaluation of a robotic TV-assistant 16 / 15
Personal robots
•More than 6 million personal robots in 2007
•Both for entertainment and household tasks
Number of robots for domestic use Source: UNECE
x 1
mill
ion
1
0
2
3
4
Entertainment &
Leisure
Vacuum cleaner lawn
mower
Other cleaning
20032007
a robot is a mechanical device which performs automated tasks, either according to direct human supervision, a pre-defined program or, a set of general guidelines, using artificial intelligence techniques.