Path: utzoo!attcan!uunet!know!sdd.hp.com!ucsd!sdcc6!beowulf!schraudo
From: schraudo@beowulf.ucsd.edu (Nici Schraudolph)
Newsgroups: comp.ai.philosophy
Subject: Re: Thoughts on emergence
Message-ID: <schraudo.655858621@beowulf>
Date: 13 Oct 90 22:57:01 GMT
References: <62500@iuvax.cs.indiana.edu>
Sender: news@sdcc6.ucsd.edu
Lines: 64
Nntp-Posting-Host: beowulf.ucsd.edu

I like David Chalmer's attempt to define "emergence", but I am bothered
by his inclusion of subjectivity ("interesting") and teleology ("unan-
ticipated") in the definition.  I believe that this can of worms can be
avoided by using a different ansatz due to Paul Churchland:

"A property P specified by its embedding theory T1 is emergent with respect
to the properties of an ostensibly reducing theory T2 just in case

1. P has real instances,

2. P is co-occurrent with some property or complex
   feature recognized in T2, but nevertheless

3. P cannot be reduced to any property postulated by
   or definable within T2."

(Paul Churchland, "Reduction, qualia, and the direct introspection of brain
states", Journal of Philosophy 82:8-28, 1985)

Basically, 1) restricts us to actual phenomena, 2) ensures the two levels
of description talk about the same phenomenon, and 3) is the beef of it.
Note that this definition is similar to Chalmer's (2) and consistent with
Reed's "emergence in simple systems" - in other words, it makes almost every
interesting property of a composite system emergent.  Now let's throw away
the trivial cases:

	An emergent property P is non-trivial iff it is not deducible from
	the property or complex feature in T2 that it is co-occurrent with.

Note that the failure to deduce P implies that either T1 or its reduction
to T2 is incomplete.  The reference to theories thus nicely formalizes the
subjectivity introduced by Chalmer: P is unanticipated by an observer O pre-
cisely to the extent that it is not deducible in O's theoretical framework.
This seems to be a formalization of Chalmer's (6) - or is there a nuance
that I have missed Dave?

To illustrate all this, the property "is_thrashing" of a virtual memory
(VM) computer system, defined by its embedding theoretical framework
("Paging Science"), is emergent with respect to the theory of non-VM
computing since

1) virtual memory systems can thrash,

2) thrashing is co-occurrent with certain combinations of properties
   (viz. memory access patterns) that are recognizable in a non-VM
   framework, but

3) there is no concept corresponding to thrashing in a non-VM framework.

This emergence may be non-trivial to the average computer user (who would
be at a loss as to why the system is crawling at a snail's pace), but to
the extent that Paging Science is a well-developed and well-grounded (in
non-VM computing) science it will be trivial to the Paging Scientist.

Not surprisingly, the people that like to talk about emergence are typically
found in fields where the either the theory itself or its grounding in a
potentially reducing low-level theory is underdeveloped, resulting in an 
abundance of non-trivial emergent phenomena.

-- 
Nicol N. Schraudolph, C-014                      "Big Science, hallelujah.
University of California, San Diego               Big Science, yodellayheehoo."
La Jolla, CA 92093-0114                                     - Laurie Anderson.
                          nici%cs@ucsd.{edu,bitnet,uucp}