Path: utzoo!utgpu!jarvis.csri.toronto.edu!rutgers!tut.cis.ohio-state.edu!bloom-beacon!ai-lab!miken
From: miken@wheaties.ai.mit.edu (Michael N. Nitabach)
Newsgroups: comp.ai
Subject: Re: Free will and responsibility.
Summary: notion of 'free will' requires new definition if it is
	 to remain a useful concept
Keywords: Behaviorism, materialism, dogma, science
Message-ID: <2586@rice-chex.ai.mit.edu>
Date: 24 May 89 00:20:42 GMT
References: <10333@ihlpb.ATT.COM> <3850@uhccux.uhcc.hawaii.edu> <52019@linus.UUCP> <533@orawest.UUCP>
Organization: MIT AI Lab, Cambridge, MA
Lines: 91

John Sebes states in reply to my recent posting:
| >{more good stuf omitted} If, on the other hand, the regularities in this
| >agent's behavior can only be explained by taking into consideration factors
| >*internal* to the agent, such as its beliefs and desires, then the
| >agent must be said to possess free will.
|Metaphysically, this is not sufficient, since the internal factors could
|be completely determined by prior external factors. Also, the distinction
|between external and internal breaks down if you push it too hard.
|Pragmatically, however, it is fine, and I agree that this is fact just
|the stance that most people, including scientists, take in those moments
|during their daily business such issues become relevant.

John's statement that free will breaks down when internal factors are
completely determined by external factors is absolutely true, *if* one
uses a particular metaphysical definition of free will, namely, the one
apparently held by John.  This is the classical definition of free will
which antonymically opposes the concepts of free will and determinism.  That
is, acts performed with "free will", are just those acts which are not
determined--i.e. those acts that "could have been otherwise."  What I am
proposing is that this definition of free will does not suffice to describe
the concept we could possibly be intending when we use the term.  First, I
would like to show that this definition is not useful.  Second, I will
propose a definition that I believe more closely approximates what we really 
must mean
when we use the term 'free will,' in both ordinary, scientific, and legal
discourse.

It is easy to see the non-utility of the classical definition of free will,
if we consider a situation in which the definition is satisfied, but we
would never intuitively consider that the agent possessed free will.  
Consider a being who possesses
a special neuron which is somehow tuned into a quantum mechanical (and hence
inherently random) process in its interior; depending on the outcome of the
completely non-determined quantum process, the neuron will either fire or
not fire, and thus lead the agent to either perform or not perform some
action.  In this case, use of the classical definition leads to the conclusion
that the agent possessed free will, because the action *even given the
identical state of every sub-atomic particle in the universe just prior to
the action* "could have been otherwise"; i.e was undetermined.  However, it
is clear that this agent would *not* be said to possess free will in
choosing whether or not to perform the action.  It seems to me that the crucial
issue is not whether or not a particular action is physically determined or
not, but rather whether or not the agent can be said to have "controlled"
the action, i.e. whether the cause of the action arose from the beliefs
and desires of the actor, or from unavoidable environmental constraints.
This is basically the definition that I proposed in my last posting: that
an agent possesses free will just to that extent that his actions are
determined by his own beliefs and desires, and not by environmental factors.

A reasonable response to this proposal is that all I have done is to base a
concept which can be rigorously defined (e.g. the classical definition), and
redefined it based on common, unprincipled intuitions about what we mean
when we use the term "free will."  I agree with this characterization of my
strategy here.  However, consider our choices in the matter.  If we choose
to hold the classical view then we must accept the fact that no-one really
possesses free will, and no-one is responsible for their actions.  Therefore,
we must either free all the prisoners in jail, or revise our notion of legal
responsibility such that it does not correspond with our everyday notion of
responsibility (namely, that some act was performed freely).  The other choice
choice is to revise our notion of free will in a way such that it will in 
some way account for our legal and social notions of responsibility, yet not
be defined circularly as just that property which we possess when we commit
acts which we are responsible for.  I submit that my proposed definition can
fulfill these requirements.

Just to summarize:  If we hold the classical definition of free will, then
we must conclude that no-one has free will, and that we are completely
unjustified in attributing blame or praise to anyone for their acts.  This
shows clearly that the classical definition of free will is not useful in
explaining how we *actually do* use the concept, because we regularly do
apportion praise and blame to one another.  The alternative is to hold a
notion of free will that will make the appropriate distinction between
acts which an agent is and is not responsible for.  One example of such a
notion is that free will exists to the extent that an action is caused by
the beliefs and desires of the agent, and not by environmental factors.  In
essence, what I am proposing is that we discard a conceptually clean and
elegant, but useless in describing human behavior, definition of free will,
and replace it with a more "vulgar" notion, but one which can explain many
of the regularities in the ways in which we actually do apportion praise and
blame to the other members of our species.

I am leaving MIT tomorrow for the summer, and I don't know how often I will
be able to get news.  Thus, I would appreciate it if any postings regarding
this topic of free will could be also sent to me directly by e-mail.  I am
somewhat new to the net, so if this is an unreasonable request please ignore
it.  Thanks.

Michael Nitabach                          email: miken@wheaties.ai.mit.edu
Dept. of Brain & Cognitive Sciences
Massachusetts Institute of Technology
Cambridge, MA