User-Centered Systems Design: A Brief History - Foundations for Designing User-Centered Systems - page 40

2.2.3 Cognitive Modeling and Programmable User Models
A cognitive model is an approximation of how people reason. The goal of a
cognitive model is to explain scientifically very basic cognitive processes, explain
how these processes interact, account for errors and breakdowns in these pro-
cesses, and derive predictions about how those reasoning processes will proceed
under different conditions.
Cognitive modeling is a method developed from early work in the late 1950s
when psychologists realized that computational processes may be a good analog of
human reasoning processes: like humans, computers take input in the form of
symbols, require memory for information storage, and manipulate those symbols
with algorithms to produce output. It was therefore proposed not only that human
reasoning could be an inspiration for thinking about computational processes but
also that computers may be a good way for us to simulate human reasoning and
therefore derive deeper understandings of how humans think (Newell et al. 1960 ;
Newell and Simon 1972 ).
Cognitive models in the late 1960s, the 1970s, and the 1980s focused on how
people solved problems symbolically: humans take input in and form symbols,
require memory for information storage, and use algorithms to manipulate those
symbols to produce output. The models were usually limited to one task (or one
type of task) and usually simulated reasoning in terms of what was going on in the
user's mind. They addressed human information processing but did not address
how information is taken in from the external world, how actions are performed in
the world, and the ways in which real world settings impact the pace at which
those processes take place. Each model was essentially a micro-theory of how
some part of behavior occurred, and it was independent of other micro-theories.
Over time, the need to integrate the micro-theories increased, which led to the idea
of unified theories of cognition (UTCs; Newell 1990 ).
These theories are implemented as cognitive architectures available as com-
puter simulations that constrain how models (based on task knowledge) can per-
form tasks in psychologically plausible ways. So, for example, often when humans
perform two tasks simultaneously, the performance on one is affected by the
performance on the other. Cognitive models are essentially programs written in a
specific language to run on particular cognitive architectures. The models can
perform complex tasks including perception, learning, reasoning, problem solving,
remembering, decision making, proprioception (how people manage their bodies
in space), and ambulation (how people move around physical spaces).
There has long been an overlap between cognitive modeling and human-
computer interaction. Drawing on these developments in psychological theory and
in simulation modeling, design researchers started investigating the possibility of
building models of how people reason and problem solve when using complex
interfaces, so that predictions about the pros and cons of different interface and
information representation choices could be tested prior to investing in any
interface or interaction development (e.g., Pew and Mavor 2007 ). Models force the