Information Technology Reference
In-Depth Information
If we have no current observations to draw upon, then we must make all of our judgments
from previous experience. If we have both previous experience and current perceptions based
upon observational data, our judgment must be based on both.
The human decision making model that was devised based on IQ tests looks very
promising. IQ tests present, not all, but some facts, and there are many possible
solutions to the problem presented. We are required to make the best and most
practical judgment. We first start by looking at the features of the facts. By extracting
the features of the facts given, we should be able to pinpoint to the right class of
hypothesis as our best judgment based on previous experience. The problem remains
of extracting only the right set of features, since they help us tune into the right
hypothesis.
Having put forward the right hypothesis to account for the facts given, we proceed
to assess the hypothesis against a purpose. Since the criteria for IQ is simplicity, the
nature of the hypothesis should help us to select the 'correct' criteria of simplicity
that will then be evaluated to determine the acceptability of a given hypothesis.
Since IQ tests reflect how human judgments are made given limited facts, then
the framework of feature-extraction and criteria-selection technique that closely re-
semble pattern recognition techniques is a good candidate for human intelligence. I
make no claim that this approach
is
the model of human intelligence. However, in the
absence of a standard technique of human decision-making process, this approach
using the solving of IQ tests seems to give acceptable results.
One of the attractions of the taxonomic approach is the symmetrical nature of the
model produced. The reverse process of ranking and selecting concepts for retroduc-
tion is employed in the selection of criteria for induction to create an equilibrium.
Each of these may employ learning as described in Chap. 11.
Hypotheses are used to make predictions and in our case, to predict the next
number of a given sequence. We ran our model against 85 samples taken from
Eysenck (
1974a
,
b
). He also provided a graph to determine IQ value. Interpreting
the results in terms of such scoring system, our system has an IQ ranging between
132 and 143 points depending on learning technique engaged.
References
Addis TR (1985) Designing knowledge-based systems. Originally Kogan Page, now Chapman &
Hall, Published October. Hardback: ISBN 0 85038 859 7. Soft back: ISBN 1 85091 251 3
Eysenck HJ (1974a) Check your own I.Q. Penguin Books, London
Eysenck HJ (1974b) Know your own I.Q. Penguin Books, London
Wason PC, Johnson-Laird PN (1968) Thinking and reasoning, penguin modern psychology UPS
11. Penguin Book, New York
Zakaria MS (1994) A model of machine intelligence based on the pragmatic approach. PhD Thesis,
Computer science department, University of Reading March
