1. The Responses of People to Virtual Humans in an Immersive Virtual Environment
Maia Garau, Mel Slater, David-Paul Pertaub, Sharif Razzaque
Presented by Eugene Khokhlov
Eugene

2. The Responses of People to Virtual Humans
An experiment investigating responses of people to virtual humans.
Responses investigated are:
Presence (a state of being present; current existence)
Copresence (sensing being close enough to be perceived)
Heart rate & electrodermal activity (EDA)
Eugene

3. Clarifications Agents = virtual humans
Hypothesis: the higher agent’s responsiveness (virtual human) would lead to participants feel like agents are humans.
VR Environment for this paper is the library settings where students are studying at the table.
Eugene

4. Experimental Design
Condition 1 (Static): All agents (virtual humans) were static, frozen in a reading pose.
Eugene

5. Experimental Design
Condition 2 (Moving): The agents were animated. Movements included fidgeting, turning pages, and looking around. Non responsive to the participant.
Eugene

6. Experimental Design
Condition 3 (Responsive): Condition 2 plus responsiveness to participant’s location in the space. Upon approach change posture, involve in gazing/staring behavior.
4 zones
Intimate (Agent is right next to participant)
Personal (Agent not that far from participant)
Social-consultative (Between personal & Public)
Public (Quite a hike)
Eugene

7. Zones Explained
If participant is in the “public” zone, the agent would move as in condition 2.
As participant moves through each zone toward the agent, the agent’s behavior would change as follows: gaze would increase, posture would become more upright, frequency of gazing and posture would increase.
Eugene

8. Condition 3 (Responsive)
Eugene

9. Condition 4 (Talking)
Condition 4 (Talking): the approached agent would briefly speak to the participant.
Language is not recognizable, but tonality suggests a question, pause, then another question.
After a few seconds an agent would say “OK” and turn back to the table.
Eugene

10. Participants $7.50 per hour for 1 hr experiment. Condition Male Female
1. Static 7 4
2. Moving 6
3. Responsive
4. Talking 5
Eugene

11. Apparatus
2 Rooms
1st – reception room where participants were greeted and completed questionnaires
2nd – a laboratory containing the Cave and the PC for monitoring participants’ physiological responses
Eugene

12. Physiological monitoring
Participants were fitted with EKG sensors on their torso and EDA sensors on their non-dominant hand.
Cables were secured in a pack strapped to the participant’s back and linked to the PC via serial cable.
Eugene

13. Environment
Eugene

14. Agents & Virtual Environment
Agents (Virtual Humans) three male and two female agents who are sitting at the library table, studying.
Additional behaviors such as blinking, leaning forward to read a book were implemented.
Eugene

15. Task
Observe your surroundings and we will ask you a few questions about what you experienced.
Four minutes to explore the room.
No mention was made of the characters
Post questioner was given.
Eugene

16. Post Questionnaire
Copresence – extend to which the participants had the sense of being with other people.
Participant behavior – to which extent participants reported that they altered their behavior in response to the agents.
Agent awareness – the extent to which the participants perceived that the agents were aware of them.
Eugene

17. Electrodermal Activity EDA
Studies show that EDA increases when a person approaches another person.
This study wanted to see if the same is true for the person approaching an agent (virtual human).
Eugene

18. Result Copresence Results
Eugene

19. Questionnaire Results
Participants who experienced the “talking” agents (condition 4) were significantly more likely to want to interact with the agents than in any other condition
Participants who experienced the responsive and talking agents (condition 3 and 4) were more likely to perceive the agents as being aware of them than in “static” (1) or “moving” (2) conditions.
Eugene

20. Copresence significance
Condition 3 & 4 (responsive and talking) are significantly different from 1 (Static) & 2 (moving),
but Condition 2 (moving) is not significantly different from the static 1 condition.
Eugene

21. Computer usage
The greater the degree of computer usage, the more the agents were responded to as a computer interface rather than human like.
Eugene

22. Participant Behavior
In condition 4 (talking) there is a significant difference on how participants reported that their behavior was actually affected by agent presence.
Avoidance, or talking back…
Eugene

23. Perceived agent awareness
Extent to which the participants perceive the agents to be aware of them in various ways. (3 & 4) are significantly higher than 1 &2.
Eugene

24. Physiological Measure
Data was not available for all participants, because the equipment did not always function correctly.
EDA & heart rate is significantly higher for condition 3 (responsive)
Heart rate increase diminishes with computer use.
Eugene

25. Physiological Measure
Eugene

26. Discussion
Sense of personal contact was significantly higher in condition 3 & 4 as described earlier.
In conditions 1 (static) & 2 (moving) the fact that the agents did not respond made some participants feel “invisible” and “ghostlike,” and unable to engage in two-way interaction of any form.
Eugene

27. Discussion
In spite of this, several people mentioned their surprise that despite their rationalization of the agents as being computer-driven, they nonetheless responded to them on some level as people.
Eugene

28. Conclusion
Participants who encountered the visually responsive agents in condition 3 experienced a significantly higher sense of personal contact with the agents. The effect diminished with experienced computer users.
Eugene

29. References
Maia Garau, Mel Slater, David-Paul Pertaub, Sharif Razzaque The Responses of People to Virtual Humans in an Immersive Virtual Environment Presence, Vol 14, No 1, February 2005,
Eugene