HASTI: a DCog modeling framework
DCog treats human-computer systems as joint cognitive systems:
specifically distributed computational systems. Theories of cognitive
support postulate why certain computational rearrangements of a joint
cognitive system are advantageous to the user within that system.
In order to proceed with analyzing or designing cognitive support,
therefore, some method of analyzing and modeling joint cognitive systems
is required. Cleary, this will frequently involve some modeling of the
human and computer and their interactions. Thus if tools researchers are
to analyze and design a broad variety of different types of cognitive
support, they will need some fairly capable modeling techniques for
modeling joint human-computer cognition. Thus one requirement is to
select from the literature appropriate models and techniques that could
be merged into a comprehensive toolkit for tools researchers to use.
The first requirement is selecting appropriate and generalized modeling
content.
This requirement could partly be fulfilled by collecting together
some models of cognition that are commonly found in the HCI and
cognitive science literature. One could look to, for instance,
"unified" cognitive models such as
SOAR, and perhaps add some key features if required. However
this tactic falls short of requirements on several counts. The
most serious problems are that they are too detailed and tedious to
use for lightweight and broad-brush analysis, and they are not organized
to make it plain how cognitive support principles (the support factors
of RODS) can be applied. Most cognitive models
are performance models, that is, they predict or explain facets of
human performance (speed or timing of action, errors and slips, etc.).
What is needed to analyze support is a different type of model: one
that explains how cognitive processes are helpfully changed.
What is required, then, is not onlyh a collection of appropriate models
and techniques, but a collection that is integrated and packaged in
a way that exposes the ways of supporting cognition by reengineering
it.
This chapter works towards fulfilling these requirements by providing a
DCog modeling framework called "HASTI". HASTI is built by integrating
several different models and modeling frameworks published in HCI and
cognitive science. The crucial aspect of HASTI is the way in which the
different facets of cognition are decomposed in order to expose cognitive
reengineering possibilities. The term "HASTI" is an acronym formed
from the five ways in which the models are decomposed. For instance,
the "H" stands for "hardware" part of the framework; this part of the
model aligns well with prior work on "architectures" for cognition:
relatively low-level cognitive structures that are invariant across
tasks and individuals. For instance, the limitations of "short term"
memory are often modeled as a limited memory component such as a small
set of registers. For analyzing cognitive support, the key aspect of the
hardware part of the model is that it is not homogenous: some parts are
faster or slower, for instance. This suggests that using certain parts
of the cognitive "hardware" may be better than relying on other parts.
Such differences feed into the analysis of cognitive reengineering
possibilities. The remaining four decomposition dimensions of HASTI
delineate other different aspects of cognition. Between them, one can
trace (at a high level) many different ways of rearranging cognition.
The result of this chapter is a high-level, integrated framework for
modeling joint cognitive systems. Undesired details have been suppressed
in order to make it suitable for speedy, broad-brush analysis in a variety
of situations and problem domains. Because it is derived from basic
science modeling, it serves to mediate access to the science base for
non-specialists such as software engineering researchers. The framework
can be used to analyze existing systems, and (with the help of theories
from following chapters) to understand the cognitive support offered in
them, and to determine further possibilities for support.
|