There are many models of mind, and many different exemplars of agent architectures.
Some models of mind map onto computational designs and some agent architectures
are capable of supporting different models of mind. Many agent architectures
are competency-based designs related to tasks in specific domains (e.g.
COG). The more general frameworks (e.g. ACT-R, AIS, SOAR) map across tasks
and domains. A number of models for synthetic minds are based on analyses
and observations of human minds. These types of agent architectures are
capable of performing certain behaviour and cognitive competencies associated
with a functioning mind. There is a problem with many of these approaches
when they are applied to the design of a mind analogous in type to the
human mind - there is no core to mind in any of these theories or designs
other than an information processing architecture. As any specific architecture
is applied to different domains, the information processing content (knowledge
and behaviors) of the architecture changes wholesale. From the perspective
of developing intelligent computational systems this is more than acceptable.
From the perspective of developing functioning (human-like) minds this
is problematic - these models are in effect emotionally autistic.
If mind is an ongoing characteristic of an entity of a certain level of
complexity and a mind is capable of moving through many different control
states, from where do the control patterns that stabilize a mind as an
ongoing (developing) personality emanate? Our current work on this theme
presents an emotion-based core for mind. This work draws on evidence from
neuroscience, philosophy and psychology. As an agent monitors its interactions
within itself and relates these to tasks in its external environment, the
impetus for change within itself (i.e. a need to learn) is manifested as
an unwanted combination of emotions. Such a control state can lead to the
generation of internal processes requiring the agent to modify its behavior
or processes in some way. The modification of an agent's internal environment
is then described in terms of an emotion motivated mapping between its
internal and external environments. Cognition and underlying processes
are used to navigate the agent-oriented internal environment of emotion.
It is suggested that personality traits are a manifestation of this emotion
core. Personality becomes an emergent property of the cognitive architecture
and its (pre-)disposition to concentrate on certain tasks and favour specific
instances of control states. Personality traits affect and influence the
different categories of cognitive and animated behavior. Moods arise from
the interaction of current temporally-global niche roles (the favouring
of certain aspects of emotion space) and temporally-local drives that reflect
the current focus of the deliberative processing as perceived by the reflective
layer. Temporally-global drives are those associated with the agent's overall
purpose related to its current, possible and desired niche spaces. Temporally-local
drives are related to ephemeral states or events within the agent's environment
or itself. The (high-level) niche-seeking drives (or dispositions) together
with the more orthodox control states bind the theoretical model together
and allow a synthetic agent to become complete and exhibit a (non-shallow)
personality.