COMMENT ⊗   VALID 00003 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	common[s88,jmc]		What is common sense?
C00010 00003	Informal example of a common sense information situation
C00011 ENDMK
C⊗;
common[s88,jmc]		What is common sense?

	The following is an attempt at a somewhat technical
characterization.  Thus it may not entirely coincide with
ordinary usage, although it is inspired by ordinary usage.

	Common sense is the ability to decide what to do in
certain kinds of information situation, which we will call the {\it common
sense information situation}.  Thus we have made almost everything
depend on our idea of the common sense information situation.

	The common sense information situation is characterized
by incomplete information and by open-endedness about what
information is relevant to achieving the goals.  This should be
contrasted with the information situation in the usual kind
of formalized theory.  The usual kinds of formalized theories
include those of probability and statistics, operations research,
and mathematical physics.  A scientist working in those domains
works informally to create a mathematical model and then works
formally within the model.  In probability and statistics,
there is a sample space.  In operations research a mathematical
model is formed for the particular phenomena being investigated
and optimal action sought.  Physics works with differential
equations.  When we propose to formalize common sense, we must
formalize the stage that precedes the creation of a mathematical
model.  Indeed the model may never be created.  It isn't that we
object to mathematical models.  Creating one is useful, especially
when it captures all the phenomena of importance.  However, the
information is usually not available initially, and in those
problems that require common sense all the way through, a mathematical
model of the conventional kind may never become available.

	The difference is illustrated partially by the following
quotation from Bertrand Russell.

``All philosophers, of every school, imagine that causation is one
of the fundamental axioms or postulates of science, yet, oddly enough,
in advanced sciences such as gravitational astronomy, the word `cause'
never occurs ... The law of causality, I believe, like much that passes
muster among philosophers, is a relic of a bygone age, surviving, like the
monarchy, only because it is erroneously supposed to do no harm $ldots$.''
--- B. Russell, On the Notion of Cause, Proceedings of the Aristotelian 
 Society, 13 (1913), pp. 1-26.
Text actually reproduced from:
P. Suppes, ``A Probabilistic Account of Causation,''
 Acta Philosophica Fennica, Fasc. XXIV, North Holland, 1970.
-------
	Indeed the notion of causation plays no role in the differential
equations of gravitational astronomy.  However, consider the recent
theory that the periodic great extinctions are caused by comets
striking the earth in swarms, and these swarms of comets are caused
by a distant binary companion of the sun in a highly eccentric orbit
occasionally coming into the Oort cloud of comets and perturbing some
of them into entering the inner solar system.  This theory is based
on gravitational astronomy.  However, no actual orbits are calculated,
because the information that would be required to do so isn't available.
Instead the partial information situation forces the theorist to
formulate his theory in terms of the notions of causation that Russell
imagined had been abandoned.

	We aren't able to fully characterize the common sense information
situation, but here's one feature.  We want to predict the future
or at least the consequences of our actions.  However, we know neither
the initial conditions nor the laws of motion well enough to give
precise predictions even to the extent that we believe the situation
is determinate.  Usually not even a probabilistic model is available.

	Actually almost all present AI work aimed at formalizing common
sense knowledge doesn't fully take into account the open ended
character of the common sense information system.  It assumes, usually
implicitly, a model, usually of a very narrow character.  The blocks
world axioms are a case in point.  The formalizations of the Yale
shooting problem are another.  We will argue that these formalizations
are advancing the understanding of common sense reasoning even though
they will need to be transformed in order to achieve the {\it elaboration
tolerance} needed to be useful.  Another example is the current work
on qualitative physics.  Its models are just as schematic as those
using differential equations.  Nevertheless, the qualitative physics
calculations are close to those that will be made by future fully
elaboration tolerant systems in common sense informations situations
in which mathematical physics won't ever be usable for lack of knowledge
of initial conditions and lack of computational ability.


ACHIEVING ELABORATION TOLERANCE THROUGH FORMALIZING CONTEXT

	This is my candidate
Informal example of a common sense information situation

We are traveling from A to B.
We are prepared to learn

1. The flights

2. The usual road to the airport is blocked off.