Page 113 - Socially Intelligent Agents Creating Relationships with Computers and Robots
P. 113

96                                             Socially Intelligent Agents


                                (a) a set of topics          (b) some speeches about the topic "brandy"
                                    that virtualized-ego (A) has
                                                             (b1)  I used to drink V.S.O.P
                                     brandy
                                               sake
                                      beer                   (b2)   I’m fond of diluted brandy
                                                             (b3)     ... ...
                                                              .        .
                                                              .        .
                                                              .        .



                                           Figure 11.2.  Example of a topics-and-summaries set



                             or “I’m fond of diluted brandy (Figure 11.2(b)) are stored in the memory of the
                             VE and posts about other topics are stored in the same way.


                             3.2     Flow-of-topics representation
                               VEs change topics occasionally by referring to an associative representa-
                             tion set of a flow of topics.  The associative representation proposed for the
                             CoMeMo-Community [5] consists of many-to-many hyperlinks that associate
                             one or more key unit with one or more value unit. The semantics of the associa-
                             tive representationare not definedstrictly. Instead, weleave theinterpretation of
                             the semantics to human association based on our tacit background knowledge.
                               In the case of VE (a) (Figure 11.3), ‘liquor’ is a key unit and ‘brandy’ and
                             ‘beer’ are value units. This associative representation of VE (a) shows a flow
                             of topics from liquor to brandy or beer, and other associative representations
                             such as that for VE (b), shows other flows of topics.
                               Associative representations that show associations of a community member
                             are stored with “topics and summaries” in the memory of VE. One mediator
                             selected by a user among VEs selects the next topic that is associated with
                             the current topic. For instance, the message for changing topics could be; “I
                             associate liquor with brandy. Next, let’s talk about brandy.” We believe that
                             association-based flows of topics make the storytelling of VEs human-like and
                             help users to view VEs as the independently working other selves of community
                             members.

                             3.3     Storytelling by ordered messages
                               In a turn, though all the VEs select messages associated with a topic at
                             the same time, only one VE is selected to speak at a time, which is done by
                             comparing the priorities of selected messages. Each VE generates a priority
                             when it selects a message. The criteria to decide priorities of VEsare as follows:
   108   109   110   111   112   113   114   115   116   117   118