Preliminary Results from the Questionnaire Study

1
Preliminary Results from theQuestionnaire Study

• AtGentive _at_ Oxford
• 18.05.2006

2
The Characters
3 facial expressions
Five characters (from left to right) Dino,
William, Julie, Philippe, and Onty
3
Study Design

• 12 scales
• Affect space (emotional valence, arousal,
  dominance)
• Usability space, ISO 9241-11 (effectiveness,effici
  ency, satisfaction)
• Attention space (conspicuousness,
  distractiveness,interestingness)
• Role space (sex, pertinence, believability)
• 5 agents 3 facial expressions
• Respondents are instructed and the Web-based
  questionnaire administered face-to-face
• Our aim is to collect 90 responses (30 for each
  between-subjects group)

within-subjects
between-subjects
4
Preliminary Results

• 45 respondents (16 female, 29 male)
• 14-16 per group (expression)
• Age from 21 to 53 years (mean 28 yrs)
• Agent character has a statistically significant
  effect on several scales¹
• 45 evaluations per each avatar
• Results still somewhat preliminary
• Especially concerning potential sex differences
• Facial expression has already some significant
  effects¹
• emotional valence
• moderates experienced dominance and characters
  conspicuousness
• Only 14 to 16 respondents per facial expression
• More data is required and will be collected

¹ In this case, statistical significance means
plt0.05 or less
5
Affect Space Valence and Arousal
Aroused
Arousal
Calm
-
Emotional valence

6
Affect Space Valence and Dominance
L Negative agent J Positive agent K Neutral
agent Female respondents Male respondents
Self
Dominance
Agent
-
Emotional valence

7
Usability Space Efficiency and Satisfaction
Female respondents Male respondents All
respondents
Satisfaction
Efficiency
8
Attention Space Conspicuousness and
Interestingness
Interestingness
Conspicuousness
9
Role Space Pertinence and Believability
Believability
Pertinence
10
Preference
Least pleasant
Most pleasant

? 6 4 1 1 2
? 3 5 1 3 3
? 9 5 0 2 0
18 14 2 6 5

? 1 3 5 3 2
? 4 1 5 2 3
? 1 1 7 2 5
6 5 17 7 10
11
Discussion

• Facial expressions of a virtual character
  influence social human-computer interaction
• E.g. a character can signal that it wants to take
  control
• People expect to learn more efficiently with
  virtual characters
• Better results lead to satisfaction (?)
• Interesting characters draw attention
• but conspicuousness does not necessarily lead
  to interest (e.g., Character draws negative
  attention)
• Pertinent characters are believable
• Human-like, pertinent characters as teachers
• Non-human characters are conspicuous and
  interesting
• Motivators and attention guides
• Some characters commented to be more suitable for
  children
• Support for having different characters for
  AtgentSchool and AtgentNet

12
Thank you!