The Understanding of Meaningful Events in Gesture-Based Interaction - Intelligent Video Event Analysis and Understanding - page 13

an ancient and natural trait of human beings, applies consequently to computer sys-
tems, interfaces and computer-generated output as well. The human operator tends
thus to decipher and interpret the response of the system in accordance to previ-
ous knowledge gained through social interactions. The deciphering takes place at
various levels including affect, behavior and cognition.
The Media equation of Reeves and Naas [39] shows that humans respond to
computers as if they were social entities. There are many cases when humans treat
computers by attributing them several human characteristics and even adjust their
behavior as they would towards other people (with regards to politeness for exam-
ple). When analyzing this behavior, two different aspects need to be considered: first
of all, the developments in anthropomorphic interfaces and, second, the tendency of
humans to anthropomorphize the objects they interact with.
With regards to the first aspect, there is a considerable interest in developing
interfaces that expose human-like attributes not only in the form of intelligent re-
sponses and dialogues but that also incorporate visual human characteristics: virtual
characters and actors [5, 6, 11, 14, 53, 62]. Cassell [6] looks at such embodied con-
versational agent interfaces as:

“specifically conversational in their behaviors and specifically humanlike in the way
they use their bodies in conversation. That is, they may be defined as having the same
properties as humans in face-to-face conversation
...
Embodied conversational agents represent a new slant on the argument about whether
it is wise to anthropomorphize the interface.”

Gratch et al. [14] for example explore the rapports that arise from continuous
positive feedback between dialogue partners and show how a virtual character can
induce stronger effects than face-to-face communications of human partners. Wag-
ner et al. [53] investigate how realistic such virtual characters need to be in order
for the communication and task completion to be effective and engaging. However,
even simple virtual actors produce sometimes strong emotions, implication and af-
fect. A simplified and well-known example of a virtual character acting as an in-
terface agent is the Microsoft Office Assistant 2 which was introduced in order to
help and assist users with tips to enhance production: Links the cat may seem like a
pleasant companion while working (even without the tips) while Rover, the dog, is
amusing when searching content in the owner's computer.
The same principles apply even more prominently for human-robot interactions
for which the embodiment is present at a much higher level [21, 22, 37]. For exam-
ple, Kanda et al. [21] explored the possibility that robots could establish relation-
ships with children in order to serve as learning partners, similar to other children.
Commercially available smart toys exist such as Sony AIBO 3 which is able to move
around, look for toys, play and communicate with the owner. Honda's humanoid
robot ASIMO 4 is capable to interpret postures and gestures and move independently

2

http://en.wikipedia.org/wiki/Office Assistant

3

http://support.sony-europe.com/aibo/

4

http://world.honda.com/ASIMO/