From MEXICA to MEXICA-Impro: The Evolution of a Computer Model for Plot Generation - Computational Creativity Research: Towards Creative Machines - page 271

former are the same for all agents; however, the latter might change between partic-
ipants in a collective narrative. Thus, if agents' pre and post conditions are different
then the requirements to perform an action might vary between them as well as the
consequences of it. As a result, each participant will generate different Contextual-
Structures from the same sequence of actions. This situation opens interesting ques-
tions about how this characteristic influences the interaction between the narrative's
contributors and how it shapes the features of the final product.
In the second option, the number of actions that each agent includes in its knowl-
edge base might be different. That implies that during plot generation one or both of
them might have to deal with unknown situations. So, collaborators must be prepared
to affront these conditions.
In the third option, the agents' set of Previous Stories might be dissimilar. They can
diverge in two aspects: the narratives' themes or their number. Regarding themes, one
can provide each agent with a set of completely different tales, e.g. a set of physical
violent Previous Stories and a set of love and tacky Previous Stories. As a result, their
Contextual-Structures contain different information that might or might not comple-
ment each other. With respect to their number, the amount of Contextual-Structures
depends on the quantity of Previous Stories. Thus, given a sequence of actions, one
agent might have more choices to progress a narrative than the other. Of course,
all these alternatives can be combined. So, MEXICA-impro has a great potential to
explore different aspects of the interaction between collaborative creative agents.
So far, we have only tested the model employing two different sets of Previous
Stories: violent narratives versus love narratives (see Pérez y Pérez [ 4 ]). The work
concentrates on analysing how the agents' knowledge-bases are affected when the
new plots generated by MEXICA-impro are incorporated as part of the Previous
Stories. To analyse the results we employ what we refer to as a Knowledge-Map (K-
Map) (see Fig. 13.3 ). K-Maps allow comparing the number and features of existing
KB-Structures in memory through time. Elements in the map are grouped by its
characteristics. The vertical axis indicates the number of tensions and the horizontal

EMOTIONAL LINKS
012345678 9 0 1 2 3 4 5

0
1
2
3
4
5
6
7
8
9
10

Fig. 13.3 A knowledge map