Databases Reference
In-Depth Information
Hobbs [11] uses first order logic to describe his causal complexes. Pearl [25]
develops probabilistic causal networks of directed graphs (DAGs).
The causal complexes explicitly considered by Hobbs and Pearl have a
required structure that may be overly restrictive for commonsense causal un-
derstanding, namely:
•
If
all
of the events in the causal complex appropriately happen, then the
effect will occur
•
There is nothing in the causal complex that is irrelevant to the effect
These requirements are probably too precise and extensive to be realized
in a commonsense world. Sometimes, only some of the events need to happen.
For example,
Someone may be able to save more money:
•
If their taxes are lowered or
•
If they earn more money.
Either even may lead to greater savings. However,
Neither may result in increased savings if they also have to pay a large
divorce settlement.
So, if all of the events happen, the effect may happen. If some of the events
happen, the effect may happen. In the commonsense world, we rarely whether
all of the events are in a complex are necessary. For eample,
A man may want to attract the attention of a woman. He may do
a large number of things (e.g., hair, clothes, learn to dance, etc.). If
he does attract the woman, he may never know which things were
relevant and which were not
An issue is how to distinguish between what is in a complex and what is
not. Another issue is how to distinguish between the things that deserve to be
called “causes” and those that do not. Hobbs suggests that a consideration of
causal complexes can be divided into:
•
Distinguishing what events are in a causal complex from those outside of
it. [16, 22, 25, 35, 38]
•
Within a causal complex, recognizing the events that should be identified
as causes from those that are not [31].
Nested granularity may be applied to causal complexes. A complex may
be several larger grained elements. In turn, each of the larger grained elements
may be a complex of more fine grained elements. Recursively, in turn, these
elements may be made up still finer grained elements. In general, people are
more successful in applying commonsense reasoning to a few large grain sized
events than the many fine grained elements that might make up a complex.
A question concerning complexes is: To what extent can we increase the
causal grain size and still have useful causal information? Conversely, can
