isbn9789526046266

5.1.1 We use mental models to interact with causal systems
Causal vs. logical mental models
It is necessary at this point to differentiate between two influential threads of research on constructs called
mental models.
The term “mental model” received wide recognition after its introduction to cognitive psychology
in the early 1980s. The two main threads of research on mental models are often attributed to the
near-simultaneous publication of two books titled Mental Models, which elaborated in two different ways
on Craik’s (1943) earlier notion of “models of reality”. The articles in the volume edited by Gentner
and Stevens (1983) greatly influenced research into the kinds of mental representations people store
about physical and software systems. The book by Johnson-Laird (1983) is seminal to a body of work
that investigates a certain kind of situation-specific mental representation that people create in working
memory to help them reason about logical problems. Markman and Gentner (2001) term mental models as
described in the Gentner and Stevens book causal mental models and those in Johnson-Laird’s research
tradition logical mental models. Logical mental models are not of interest for my present purposes;
whenever I write about “mental models” in this thesis, I refer to causal mental models. 1
Mental models of systems
People form causal mental models of all manner of things. Schumacher and Czerwinski (1992) note that
while one can have a mental model of a marriage or a social environment, not all topics are equally well
represented on the research agenda. Much of the research on mental models has focused on the way people
interact with and think about complex physical and software systems that involve causal mechanisms (e.g.,
an electrical circuit, a word processor). The mental model is a theoretical construct posited to explain
how people describe the purpose and underlying mechanisms of such systems to themselves and how they
predict future system states. This emphasis stems partially from arbitrary historical reasons that arise
from research tradition 2 and partially from practical reasons: “We would argue that experts in computer
systems are easier to define than experts in marriage.” (Schumacher and Czerwinski, 1992)
Research does exist on causal mental models that is not concerned with humans’ relationships with
technical systems. However, I will limit my discussion to causal mental models of technical systems.
Researchers have further sought to identify the features that mental models of systems have in general,
the characteristics that distinguish between useful and not-so-useful mental models, the transferability
of the knowledge stored in mental models, and the relationships between learning and mental model
construction. I will comment on each of these topics below.
5.1.2 Research has explored the characteristics of mental models
According to Norman’s (1983) seminal description, mental models:
• reflect people’s beliefs about the systems they use and about their own limitations, and include
statements about the degree of uncertainty people feel about different aspects of their knowledge;
• provide parsimonious, simplified explanations of complex phenomena;
• often contain only incomplete, partial descriptions of operations, and may contain huge areas of
uncertainty;
• are ‘unscientific’ and imprecise, and often based on guesswork and naïve assumptions and beliefs,
as well as “superstitious” rules that “seem to work” even if they make no sense;
1 The body – or rather, bodies – of research on mental models make up a complex terminological and conceptual tangle.
As with the word “schema” (Section 4.2), some authors use “mental model” as an umbrella term for all knowledge, a practice
that has been criticized by authors such as Rouse and Morris (1986). There are numerous other more or less idiosyncratic
ways of using the term in the literature, which I will not cover here.
2 The focus on causal systems and devices can be traced back to human–machine interaction studies in the 1960s and
a body of research that investigates process control: the manual control of complex machines and, later, the supervisory
control of increasingly automatic machines (Rouse and Morris, 1986; Wickens, 1996, and references therein).
53