Chapter 4: RODS: A Cognitive Support Explanation Framework
Why is a tool useful? One aspect of usefulness is being able to support
or assist cognition. Tools researchers want to explain and predict
how this support is generated. We would like to be able to explain how
tools (in general) can support cognition; we would like to be able to
explicitly state and compare claims about how particular tools
support cognition. In order to build the theories or claims of how
artifacts support cognition, it is necessary to have a capable theoretical
backdrop onto which suitable theories and claims can be painted. Although
there are plenty of possible schools available from which to draw upon,
it is still necessary to carefully select and draw out exactly how
to make theories and claims of cognitive support. Prior work has not made
it clear how to do this in a general way that can be applied to more
than a few select examples of cognitive support identified in the survey
in chapter 3.
This chapter introduces a new overarching framework for discussing and
analysing cognitive support. The framework is built from prior work in
a field of cognitive science called distributed cognition (DCog).
DCog is an eminently suitable starting point for the systematic study of
cognitive support. It treats cognition as a process distributed between
humans and artifacts. Doing so means that it can be used to unfold
the reasons for how the artifacts assist in the cognitive processes. In
contrast, non-distributed cognitive modeling techniques normally treat
external influences in a parenthetical and ad hoc manner.
To progress with DCog, however, the ways of expressing and explaining how
artifacts cause the support needs to be addressed. The key step in doing
so gives the resulting framework its name. The framework is called RODS,
and is an acronym generated from the four foundational elements within
the framework. They are called support factors: fundamental
ways in which cognitive support is offered to the user. Each support
factor identifies a principle for explaining a class
of similar types of cognitive support. Because the framework adopts
DCog, the support factors are all computational principles of
how performance is affected. This is because joint human-computer
cognition is treated as a distributed computational process. Each
support factor, therefore, identifies a distinct principle for why
certain rearrangements of computation yield computational advantages.
For instance, the "D" in RODS stands for "distribution", and relies
on the well known principle that distributing computational processes
between multiple computers can reduce the computational load placed on
the individual computers, and can allow computations that would not be
possible for an individual computer in the network. For example, moving
a large database from the local machine and onto a separate server can
reduce the computational load on the local machine, and simultaneously
make it possible to work with a bigger database. Some new overheads are
involved, but that is the price for the computational assistance. In a
similar way, various computational aspects can be offloaded from a human
worker. For example, various contents of memory can be offloaded onto
external memories: appointments, to do lists, bookmarks, and so on.
The internal cognitive load can be reduced and harder tasks can be
tackled because the effective memory increases. Some overhead is
generated, but that is the price paid for the cognitive support.
Although the above example is simple and familiar, it illustrates
the basic way of explaining any type of cognitive support within RODS.
It turns out that using just the four RODS support factors, many different
prior theories can be integrated and used to explain a very wide variety
of cognitive support. This integration is real progress since a great
variety of support types can be explained uniformly within a single
framework. To date, theorizing about cognitive support has been done
in a relatively piecemeal fashion.
The chapter describes the four types of support factors in RODS
and relates them to prior research on problem solving, automation,
and external representation use. It also discusses the fundamental
convictions underlying DCog that makes the framework possible, and
introduces a way for making claims about how a specific tool can be
argued to support cognition. In addition, techniques for modeling joint
human-computer systems are also discussed. These techniques form a
key part of the overall RODS framework. They introduce principled ways
for allowing DCog analyses to be generalized across multiple different
instantiations of DCog systems. In that way, DCog analyses are not
limited to speak about specific DCog systems. In addition, a way of
encapsulating and abstracting overheads is introduced. This makes
it possible to model joint cognitive systems without getting bogged
down in many of the details that are unimportant for the analysis of
cognitive support. Overall, the RODS framework provides a principled,
comprehensive, generalizable, and simplified theoretically driven
toolkit for analyzing and explaining cognitive support within joint
human-computer systems. It is a key foundation for doing any kind of
work in designing or analyzing cognition in human-computer systems.
|