Information Technology Reference
In-Depth Information
Cone
CIRCLE-POINT
POINT-CIRCLE
Pyramid
POINT-SQUARE
SQUARE POINT
TRIANG
CIRCLE
SQUAR
Fig. 7.2
Venn diagram illustrating unique projections (e.g.,
circle
for sphere and
square
for pyramid)
and the shared phenomenology of hypothesized objects (e.g., sphere and cone can both have a
circle
projection; cone and pyramid can both have a
triangle
projection)
This process of negotiating hypotheses about objects compatible with the ob-
served projections can be specified in terms of our analogy to variable or
irrational
sets (Chap. 5, Addis and Gooding
2004
). Possibilities for observation depend on
which experiment is selected. But this depends in turn on which hypothesis is most
believed and whether doing an experiment seems more likely to reduce an actor's
uncertainty about the world than would consulting another actor. So the process is
far more complex than can be captured by applying a single inference rule to a set
of observations.
This process of negotiating hypotheses about objects compatible with the ob-
served projections can be specified in terms of our analogy to variable or
irrational
sets (Chap. 5, Addis and Gooding
2004
). Possibilities for observation depend on
which experiment is selected. But this depends in turn on which hypothesis is most
believed and whether doing an experiment seems more likely to reduce an actor's
uncertainty about the world than would consulting another actor. So the process is
far more complex than can be captured by applying a single inference rule to a set
of observations.
7.2
Why Inference Can't be Modelled
We remarked above that the order in which observations are encountered would affect
the pattern of belief-revision. This process is also influenced by the order in which an
agent consults other agents. Different patterns of observation and consultation will
produce different states in each of the agents. The order in which such events occur
is not predictable. An iterative approach—a
simulation
—can capture this aspect of
science in a way that static, structural or semantic models do not.
