This chapter modeled emotions using a hierarchical system of heterogeneous components:
Memories of relationships are kept using moving averages for important facts.
The separate FSA for feelings are grouped with nondeterminism, simplifying the model.
Sensations are modeled as fuzzy automata, but are experienced as triggers from multiple sources.
Emotions are similar to fuzzy linguistic variables, although changes are triggered by sensations.
Mannerisms are called by nested probabilistic automata, which are activated when a threshold is crossed.
Moods are modeled as hierarchical collections of states, used to reduce the number of transitions.
From a technological point of view, there are a few problems:
The fuzzy state machines can be difficult to design, because particular combinations of events may cause degeneration.
Customizing the system is a matter of exploiting the flexibility of the specification, and calling user-defined routines.
The system as a whole is somewhat heterogeneous, and despite the advantages, a homogenous systems may prove to be a better trade-off between development and results.
As a whole, the emotional system improved the believability of the characters (mannerisms), and made the changes in emotions more realistic (fuzzy models).
The next chapter examines a concept that was exploited by the design of this emotional system: emergent behaviors. Simple interacting components generate complex patterns in the behaviors. Emergence is particularly important for game AI developers, because it can be so powerful and dangerous at the same time.