Information Technology Reference
In-Depth Information
finite, fixed set of necessary and sufficient conditions for its application (as assumed
in analytical philosophy), nor by stipulating an exact set of referents (Addis and
Gooding
2004
). Rather, meaning is given by sets of objects and associations that are
invoked when a term is used. Membership of these sets can change (Hesse
1973
;
Kuhn
1974
). In science—as in everyday life—words and phrases often emerge from
concrete situations in which participants jointly work out ways of describing what is
going on (Arrighi and Ferrario
2004
). New terms, symbols or images are created that
acquire meaning through collective use in real situations (Gooding
1990
; Suchman
1987
; Goodwin
1995
). Even after terms or symbols have acquired an established us-
age that does conform to semantic rules, experts must sometimes validate judgements
about correct usage or the validity of an inference. It is the scientific counterpart of
establishing the authenticity of a painting or whether an ornament is made of real gold
(Putnam
1975
). This is the context in which we shall argue that experiments function
as models. These models mediate between the emerging language of description and
the explanation of the changing phenomenology of a domain.
5.5
Experiments as Mediating Models
Scientists use experiments to constrain the variability of a complex world by selecting
certain features and processes and excluding others. The desired level of control
can be difficult to achieve, so experiments develop over time; they have histories.
Sometimes experiments become autonomous of the theories they are meant to test
(Hacking
1983
). During the twentieth century the character of experiment changed,
from bench-top activities conducted within a single space by a very few people, to
being complex arrangements of detecting and measuring machines, data flow, data
analysis, and control procedures for managing the machines, information and large
numbers of people (Galison
1997
; Pickering
1995
). Experiments were absorbed
into distributed cognitive systems (Giere
2004
; Nersessian
2005
). Nevertheless, one
aspect of experiments did not change: their design, use, interpretation and validation
all involve seeking a consensual view through negotiation between individuals and
groups of individuals.
Experiments are social objects constructed
via
epistemic practices. They are not
simply ways to display facts. The way in which an experiment is perceived, no
less than the way its results are interpreted, will vary according to context. Similar
points have been made about models: models are abstractions that select features
of the world and mediate between the abstract concepts of a community and the
dappled, complex world they are theorizing (Cartwright
1999
). Models have histories
showing that their complexity varies according to function and context. Models
are also investigative instruments, and like experimentation, modelling can become
an autonomous activity (Morrison
1999
). As with experiments, there is no unified
method for modelling (de Chadarevian and Hopwood
2004
). Models and experiments
facilitate mutual adjustment of theories in relation to empirical evidence, networks of
theoretical assumptions, and the material exigencies of acquiring data. Experiments
