COMMENT ⊗   VALID 00004 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	ascrib[f83,jmc]		Ascriptionism
C00008 00003	The Ascriptionist Hypothesis
C00010 00004	.require "memo.pub[let,jmc]" source
C00018 ENDMK
C⊗;
ascrib[f83,jmc]		Ascriptionism

	My purpose is to elaborate and defend a new doctrine -
ascriptionism.  Naturally it has precursors, but for the time being we
won't bother with them.

	The doctrine, assumed in (McCarthy 1980), is that entities
have mental and other intentional properties only in the sense that
they can usefully be ascribed them.  This has as a consequence that
much philosophical study of mental qualities, that aimed at finding
the limits of their applicability, is useless pedantry.  This should
not be regarded as an assertion that certain philosophers are pedants
by nature, merely that a certain scientific error has led them
to study unanswerable questions.  The point is that in order to
study a phenomenon like belief, one should study the center of its
domain of applicability, not its boundaries.

	Consider a young child beginning to ascribe beliefs.  He
may say, "Fido thinks the bone is in my left hand."  Why does
the child say such things?  More basically, how does he come to
think such things?  In a very few years you will more and more
observe children equally spontaneously ascribing beliefs to
computer based machines.  The child will say about a teaching
machine, "It doesn't think I finished lesson three".

	The child ascribes beliefs for the same reason that the
physicist ascribes entropy or momentum.  While it is not directly
observable, its use as a theoretical concept helps make sense of phenomena.
The child's ascription of the belief is more stable than his opinion
of how the dog came to believe it or his prediction of the dog's
future behavior.

	The child's habit of ascribing beliefs did not arise through
the acquisition and internal representation of a rule asserting
"X believes Y if and only if  ...".  Therefore, the psychologist
or philosopher might be misguided in looking for a rule of the
form "The child will ascribe a belief of  Y  to  X  if and only
if . . .", although it is logically possible that a psychological
rule could exist without the internal repreentation of a rule.

	However, the evidence is against both.  First of all,
we know that we can create situations in which the child will
be puzzled about whether to ascribe belief.  Moreover, we can
probably create situations in which whether the child ascribes
belief on the basis of given observations will depend on
immediately preceding events that everyone, including the child,
will subsequently admit to be irrelevant to the question of 
whether belief should be ascribed.

	Therefore, we propose to regard belief and many other
mental qualities as essentially approximate.  They are terms
in informal theories, and their usage depends on how the whole
theory is used.

	Does this make them subjective and unsuitable for scientific
study and formalization?  No, not if we use what (McCarthy 1980)
introduces as second order definitions.  The behavior of people
in ascribing belief
and ascriptions of belief that will be useful for
computer programs share certain qualities.  Different people
and different machines will have somewhat different rules.
Therefore, the important cases will be the ones in which
almost everyone agrees - not the cases where there is controversy.

	If study of the boundaries of the concept of belief isn't
likely to be fruitful, what is?

	This doctrine has the consequence that the fruitful study
of mental qualities should emphasize the relations among the
central cases of the different concepts rather than the limits
of applicability of the separate concepts.
The Ascriptionist Hypothesis

	We present a hypothesis or doctrine
that we shall call %2ascriptionism%1.  Briefly the idea is that
a mental quality like belief is best studied as an ascription
that people and computer programs make in order to understand
the behavior of other people and computer programs.  The hypothesis
has the consequence that much of the way mental qualities are
studied by philosophers and psychologists
is unlikely to lead to much understanding.

	Here are the details, which fortunately are few.

	1. The object of much theoretical research about
mental phenomena is to arrive at definitions that agree
with our intuitions and the results of research.
The goal of such research is typically a sentence of the form

	%2P believes  p   if and only if . . .%1 .

	2. 
.require "memo.pub[let,jmc]" source;
.cb ASCRIPTIONISM - A HYPOTHESIS OR DOCTRINE OR DOGMA

.CB "John McCarthy, Stanford University"

	Suppose we want to understand a mental phenomenon such as
belief.  Our science may be philosophy, psychology, physiology or artificial
intelligence.  Thus we may be interested in the abstract nature
of belief, in the structure of human belief, in the representation
of belief in the nervous system, or in making computer programs
that believe statements when this is warranted by their experience.

	It is natural to hope that the result of our study will
be a sentence of the form

"Person  ⊗P  believes  ⊗p  if and only if ⊗E" where  ⊗p  is a sentence
or proposition or whatever we choose to regard as an appropriate
object of belief, and ⊗E is a sentence about the relation between
the person and the world or about the internal structure of the
person depending on what doctrine we are using about belief conditions.

	Indeed most of what is written about belief in philosophy and
theoretical psychology consists in examining one or more candidate
for the above mentioned ⊗E.  However, I want to claim that this
approach of looking for a definition or a characterization is
unfruitful and leads mainly to arguments about distinctions that
turn out to be pedantic and don't contribute to our understanding
of the phenomenon of belief.

	There is another way of studying such phenomena which is
more likely to be fruitful.

	In order to see why a search for an "if and only if"
characterization is unfruitful consider a child learning to
ascribe beliefs to (say) his dog.  "Fido thinks the bone is
in my left hand", the child says.  Surely, the child has not
adopted a definition nor does he pretend to have a characterization.
He has only a very limited notion of what dogs think.  However,
this notion is connected to his experience with how dogs behave.
Why does he say the dog thinks the bone is in his left hand
rather than describing the dog's observations or
predicting what the dog will do?

	The reason is that the concept
of where the dog thinks the bone is covers more of his experience
than either what he remembers about how the situation developed
or his ability to predict what will happen next.  The child has
found that ascribing certain beliefs to dogs and people helps
understand their behavior.  Among other things, the ascription
of belief is required in order to apply a "principle of rationality".
This principle accounts for much behavior by saying, "It does
what it thinks will achieve its goals".  The principle of rationality
is very powerful in explaining human and animal behavior.  Its
use requires the ascription of both beliefs and goals, and it also
requires a way of comparing an observed action with a concept of
it formed before the action takes place.

	While the use of the concept of belief in understanding
behavior requires its linkage with other concepts, no
if-and-only-if characterization of belief is required by the
child or even an adult.  Asking for such a characterization
elicits first puzzlement and then answers that can readily be seen
to be wrong.  There is also no reason to believe that two
people have identical criteria for ascribing belief or even
that a single person has a consistent criterion.  By this I
mean that whether a particular belief is asserted in response
to a question may depend on how the question comes up in ways
that the person will later agree are irrelevant to whether
the belief should be ascribed.

	We can formulate the above hypothesis or doctrine of
"ascriptionism" as follows.

	1. A mental phenomenon like belief should be regarded
as an ascription rather than as an intrinsic property.

	2. Such phenomena are appropriately studied in their
causal relation with other mental phenomena such as wants,
desires and intentions.  Looking for these relations in the
situations that arise frequently will
be more scientifically fruitful than trying to characterize
the phenomena in isolation with the aid of limiting cases.

	3. The principle of rationality and principles associated
with answering questions play an important role in understanding
mental phenomena.

	4. These concepts will be expressed in approximate theories.
(McCarthy 1980) discusses approximate theories, but the general
ideas is that the concepts cannot withstand attempts to give them
complete if-and-only-if definitions.

	5. Both modal and syntactic theories of belief are viable.
For example, there can be no law against studying the relation
between a dog and the English sentence "The bone is under the
bush".