Download Etiquette and Efficacy in Animated Pedagogical Agents

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts
no text concepts found
Transcript
Etiquette and Efficacy in
Animated Pedagogical Agents:
The Role of Stereotypes
Kristen N. Moreno,1 Natalie K.
Person,2 Amy B. Adcock,1 Richard N.
Van Eck,1 G. Tanner Jackson,1
Johanna C. Marineau1
1University
of Memphis
2Rhodes College
Agents in Learning Technology
• Computerized tutoring systems are
becoming more widely used in many
settings
• Recent enhancements include the addition
of animated agents as tutors
• Agents can be programmed to behave in
accordance with social expectations
• How do agents affect learning?
Etiquette in ITSs
• Failure to conform to social expectations:
breach of etiquette
• Reeves & Nass (1996): people apply rules
of human interaction to computers
• Rules of etiquette facilitate social
interactions among humans
• Could affect human-computer interactions
AutoTutor
• An intelligent tutoring system developed by
Tutoring Research Group at University of
Memphis
• Tutors college students in physics and
computer literacy
• The agent is intended to facilitate humancomputer interactions
Agents’ Effects on HCI
• Agents do not facilitate all types of learning
• Agents could potentially annoy or distract
learners
• Important to understand the factors that
determine the effects of agents on humancomputer interaction
Agent Characteristics
• May vary in race, age, gender, or other
demographics
• Expectations for their behavior may vary
with these characteristics
• Expectations may be based on stereotypes
The Role of Stereotypes
• Mental device that simplifies social
environment
• Stereotypes offer information about
members of certain social groups
• But information is often wrong
• Still, people use them frequently (even if
often unintentionally)
Stereotyping Agents?
• Do people’s stereotypic expectations about
agents affect pedagogical efficacy?
• Failure to conform to stereotypic
expectations could be a breach of etiquette
• Conforming too closely to stereotype could
also constitute a breach of etiquette
Our Research Questions
• Do people stereotype agents?
• Do stereotypes of agents affect pedagogical
efficacy?
• Participants formed impressions of agents
who were apparently members of certain
social groups
• Agent delivered tutorial on blood pressure
• Assessed stereotypic perceptions, learning
Participants and Design
• Participants: 39 Introductory Psychology
University of Memphis students (69%
female)
• Design: 2x2x2 between-subjects
– Agent ethnicity: African-American or Caucasian
– Agent sex: male or female
– Participant ethnicity
Agents
• 4 agents created using Poser 4
• Different agents to manipulate sex
• To manipulate ethnicity, changed several
physical features across agents (within
genders)
• Recorded live voices
Procedure
• All materials presented on computer by
Macromedia Authorware
• Participants wore headphones (groups of 1
to 10 people)
• Agent first gave navigation instructions
• Stereotype ratings
Procedure
• Blood pressure tutorial (self-paced)
• 18 multiple choice questions on blood
pressure
• Demographics and manipulation checks
Excerpt from Tutorial
Now that you know what blood pressure is, let’s talk
about what the numbers, such as 110 over 70,
actually mean. The top number is the systolic
pressure. This is the peak or maximum blood
pressure that occurs as the left ventricle of the
heart pumps blood into the aorta, leaving the heart.
This contraction of the heart is called systole. The
normal systolic pressure in a healthy adult is 100
to 140 millimeters of mercury. The bottom number
is the diastolic pressure. This is the pressure
exerted against the aorta when the left ventricle of
the heart relaxes.
Stereotyping Indices
• Participants rated agents on traits that were
relevant to stereotypes of their groups, as
indicated by pilot test
• Scales for African-American males,
Caucasian males, African-American
females, Caucasian females
• Alphas ranged from .73 to .89
Results: Stereotyping
5
Stereotyping
4.5
4
Male Agent
Female Agent
3.5
3
2.5
2
AfricanAmerican
Male
Caucasian
Male
AfricanAmerican
Female
Stereotyping Index
Caucasian
Female
Learning Index
• Proportion of correct answers on 18 item
test
• Range was .44 to 1.00, M = .81
Results: Learning
0.9
Learning
0.8
African-American P
Caucasian P
0.7
0.6
0.5
Male
Female
Agent Sex
Ancillary Indices
• Participants rated agents on pleasantness of
voice, trustworthiness, likeability,
interestingness, knowledgeability, and
teaching skills
• Caucasians found male agent to be less
interesting than others
• None of these factors correlated with
learning
Stereotyping Agents
• Participants stereotyped animated agents
based on sex
• Did not stereotype based on ethnicity
• Further evidence that people see computers
as social actors and apply the human
interaction rules to them
Pedagogical Efficacy
• Agents differed in pedagogical efficacy
depending on sex
• TRG is interested in creating maximally
effective and enjoyable agents
• Important to discover the factors that
account for differences in pedagogical
efficacy
Role of Etiquette
• Conformity or nonconformity to stereotypic
expectations could constitute a breach of
etiquette
• Some participants addressed this directly
Future Directions
• Although agents differ in pedagogical
efficacy, the characteristics that contribute
to these differences have yet to be clearly
established
• Currently examining whether stereotypic
expectations about domain knowledge
interact with demographic characteristics of
agents to affect learning
Conclusions
• Limited evidence that people stereotype
agents as they do humans
• Sex of the agent appears to affect learning
in at least some cases
• Rules of etiquette may determine an optimal
level of conformity to stereotypes