Modelling Experiments - Natural and Artificial Reasoning - page 58

5.2 Experiment, Inference and Theory Change

Experimentation is one of the key features of science and technology, yet it is often
treated as simply an adjunct of the construction and revision of theories. Similarly, the
impact of evidence in theory acceptance and revision has traditionally been modelled
in terms of one of the logical or statistical rules of inference. In this chapter we will
draw on studies of experimental work in the history of science to develop a model
of experimentation conducted by agents who interact with each other as well as
with the phenomenology of their experiments. This approach to discovery aims
to integrate formal, empirical and ethnographic methods in order to include some
of those features of science that philosophers, historians, and social and cognitive
scientists identify as important for understanding the process of science and the
conduct of scientists.

5.2.1 Experiments and Experimenters

We will develop the view that experiments can be considered as models of the partic-
ular aspects of reality they are designed to investigate. Therefore, in order to model
experiments I must identify and represent in our simulation those features that all
experiments have in common as well as features that researchers have in common
(Gooding and Addis 1999 ). We will treat these features as parameters of a simu-
lation model that enables us to vary some of them to represent different types of
experiment, different actors, and the fact that making inferences about evidence is a
contingent and socially mediated activity. We can then model experiments ranging
from compelling, idealized thought experiments and decisive or crucial experiments
to those that are exploratory, ambiguous or controversial, as are most cutting-edge
research experiments.

When modelling a person and in particular a scientist
I will refer to them as ' actors 'or' agents '.

This is so we will never confuse a real person with a simulated person during this
discussion. I can vary the sensitivity of actors in our model to new data, their re-
ceptivity to the opinions of other actors, their access to experimental resources, and
their contact with other actors.
Recent philosophical work, by Giere ( 2004 ), Cartwright ( 1999 ), and others argue
that to show how theoretical claims actually engage with the world we must move
beyond a purely semantic conception of scientific theories. Experiments are con-
stantly designed and re-designed in the context of implementation and use, so they
can be considered as types of models which mediate between abstract, theoretical