Getting the Point Across: Exploring the Effects of Dynamic Virtual Humans in an Interactive Museum Exhibit on User Perceptions

IEEE Transactions on Visualization and Computer Graphics (2014), pp. 636-643, doi: 10.1109/TVCG.2014.26
Experiments
Tasks
Findings

Motivating goal: Can the visual appearance of a virtual human be dynamically selected based on user specific characteristics to maximize the effectiveness of the interaction? Specific user study goal: Understand how the relationship between the virtual human's body-mass index (BMI) and the user's BMI affect user perceptions and confidence on the information provided to them in a virtual human experience.

Variables

Independent Variables
  • virtual human bmi - average (24) vs obese (31)
  • gender - male vs female
  • user bmi - value computed from sensor (height and weight) readings. users classified as either average (below 30) or obese (above 30)
System Info

Displays
Input Hardware
  • button - custom-made 6 button pad
  • microsoft kinect - used to detect number of users and to verify bmi calculations
  • height sensor - senix toughsonic tspc-30s1 located at 7.5' above the ground
  • weight scale - arlyn scale 320d located on the middle of the exhibit.

Participant Info

Participants were museum visitors at the Museum of Science and Industry in Tampa, FL. 209 were female. Sample is varied in terms of age and background.

Total # Age Range Gender Balance
333 12 - 90

Users had to interact with a set of virtual humans and learn about public health topics.

Interaction and Environment

Interface

Users interacted with three virtual humans during the experience. The experience was divided in four stages: greeting, instruction, recall and survey. 1) During the greeting stage, users interacted with the first virtual human. During this stage, users inputted their gender and selected from a list of three possible public health topics to learn from (asthma, melanoma and living a healthy lifestyle). 2) During the instruction stage, users interacted with a second virtual human (Dr. Blackwell). Users learned from the selected topic from Dr. Blackwell. Dr. Blackwell was selected to match the user's selected gender. The virtual doctor could either have average BMI (24) or obese BMI (33) 3) During the recall stage, users interacted with the third virtual human, a virtual patient. Users gave advice to the virtual patient based on the information received during the instruction stage. The virtual patient was selected to match the user's selected gender. The virtual patient could either have average BMI (24) or obese BMI (33). 4) During the survey stage, users interacted again with the first virtual human and filled in a survey on their perceptions of the knowledge received from the experience. Users interacted with the virtual humans by selecting from a list of possible statements to say to the virtual humans. The virtual humans answered with pre-recorded speeches.

Dimensionality Scale Density Visual Realism
3D Small Low Low
Metrics

  • user task perception - subjective answers to the questions "how knowledgeable was dr. blackwell about [asthma|melanoma|living a healthy lifestyle]?" and "how well will your patient follow your advice about [dealing with asthma|how to prevent melanoma|living a healthy lifestyle|
  1. There was a significant interaction between virtual human bmi, gender, and user bmi on user task perception for a interactive experience of the ve task.

    Responses from users to the question: “How knowledgeable was Dr. Blackwell about the topics discussed in the lesson?” were analyzed using a factorial ANOVA. This ANOVA included three factors: 1) BMI concordance status between user and virtual doctor, 2) u

    Specificity: Neither

  2. There was a significant direct effect of gender on user task perception for a interactive experience of the ve task.

    There was a statistically significant difference (p=0.001) between genders in the answer for the question "How knowledgeable was Dr. Blackwell about the topics discussed in the lesson". Female users reported on average perceiving the virtual doctor as mor

    Specificity: Neither