%126
\newpage
\begin{center}
{\bf\Large A New Direction in Systems Analysis -- 
From Computation with Measurements to
Computation with Perceptions}

\medskip
L.A. Zadeh


\vspace*{.25in}
Abstract
\end{center}

In one form or another, decision processes play a pivotal role
in systems analysis.  Decisions are based on information.  More
often than not, decision-relevant information is a mixture of
measurements and perceptions.

It is a long-standing tradition in science to deal with
perceptions by converting them into measurements.  As is true
of every tradition, a time comes when the underlying assumptions
cease to be beyond question.

A thesis advanced in our work is that closer analysis leads
to the conclusion that in most fields of science -- and especially
in systems analysis -- conversion of perceptions into measurements
is, in many cases, infeasible, unrealistic or counterproductive.
The alternative is to develop a machinery for computation with
perceptions which exploits the vast computational power of modern
computers.  In essence, this is the aim of the computational theory
of perceptions (CTP).  Somewhat paradoxically, the source of
inspiration for this theory is the remarkable human capability to
perform a wide variety of physical and mental tasks without any
measurements and any computations.  Underlying this capability is
the brain's crucial ability to manipulate perceptions -- perceptions
of time, distance, direction, speed, force, shape, color, similarity,
likelihood, intent and truth, among others.

The point of departure in the computational theory of perceptions
-- the first stage in the reasoning process -- is conversion of
perceptions into propositions expressed in a natural language,
with a proposition viewed as a carrier of information which provides
an answer to a question.  A key idea in CTP is that the meaning of
a proposition may be represented as a generalized constraint on a
variable.  This idea forms the basis for what is called constraint-
centered semantics of natural languages (CSNL).

The second stage in CTP involves translation of propositions in the
initial data set into the constraint language GCL, resulting in a
collection of antecedent constraints which constitute the initial
constraint set ICS.  The third stage involves goal-directed
propagation of initial constraints augmented with decision-relevant
constraints induced by an external knowledgebase.  The goal is a
terminal constraint set in GCL which upon retranslation -- the
fourth and last stage -- yields the end result of the reasoning
process.

Existing theories, especially probability theory, decision analysis
and systems analysis, lack the capability to operate on information
which is perception-based rather than measurement- based.  The
primary objective of the computational theory of perceptions is to
add such capability to existing theories and thereby enhance their
ability to deal with real world problems in an environment of
imprecision, uncertainty and partial truth.