Date: Fri, 2 Aug 1996 19:19:18 GMT
Reply-To: PSYCHE Discussion Forum <[log in to unmask]>
Sender: PSYCHE Discussion Forum <[log in to unmask]>
From: Aaron Sloman <[log in to unmask]>
Organization: School of Computer Science, University of Birmingham, UK
Subject: Re: Are we always in control? (Was Re: I can't see -- all is dark
This is a very hasty and incomplete response as I am about to leave
for AAAI96 at Portland (returning 10th Aug).
Pete Mandik <[log in to unmask]> wrote on Tue, 30 Jul 1996,
starting with some kind words:
> I find Sloman's proposal for the architecture of self control quite
> plausible. It certainly merits having further details worked out.
Yes. I am aware of many gaps, including the need to replace the
metaphorical talk of a global workspace with something much more
precise. Also the various distinctions: reactive, deliberative, and
meta-management architectures are simplified approximations.
NOTE ADDED 3 Mar 2014: For a more recent elaboration of these distinctions see
Pete then raises questions about the real "you" (or "me"?):
> Recall that Sloman proposes that organisms who enjoy control have three
> layers, the reactive, the deliberative, and the meta-manager. And he
> suggests that you are in control just in case your behavior satisfies the
> goals of the meta-manager.
> Why is it that the *you* in the sense of *doing what you want to do* is
> identified with the meta-manager?
I think that there are two answers here.
The first is that talking about "the you" at all, and trying to identify
it with anything is just a muddle.
(It's not even grammatical, which is a sign that some philosophical urge
is at work that should be controlled.)
We don't have to identify "the *you*" (whatever that may be) with
anything at all to explain the difference between you being in control
of your thoughts and you not being in control of your things. Why should
*you* be anything more, or less, than the whole person? The question is
The second answer is to diagnose the reasons why we are tempted to look
for a special thing within, to be the real "you". We can then hope to
offer a better way of thinking about such things.
I'll elaborate on these two answers.
First another reminder
| ...I'll just summarise by saying that
| meta-management mechanisms are a subset of deliberative
| mechanisms that can to some extent monitor internal
| processes including themselves (like a computer operating
| system), evaluate options, select strategies, and
| make plans for the deliberative system.
(a) Note that I did not claim that there is any *identification* of
anything (you, me, the self, the soul, or whatever) with the
I merely offered the extra layer of architecture as a possible
explanation of how it is possible to make a distinction between being in
control of ones thoughts (i.e. processes in the deliberative
sub-architecture) and being partly out of control of those things.
The basic idea is very simple: if there's nothing in the system that can
monitor deliberative processes, and nothing that can form goals and
plans for the deliberative processes, then there's no way we can make a
distinction between cases where those deliberative processes do and
cases where they do not conform to such goals and plans. I.e. there
would be no distinction between being in control of thoughts and not
being in control of thoughts.
Please do not read any more into this than I have said.
In particular, there is also no presumption of uniqueness of
"personality" in the monitoring system.
Although I did not say this in my original posting, it is obvious
that meta-management *mechanisms* may at different times be
"occupied by" what we may think of as different collections of
"high level software", i.e. different collections of principles,
preferences, beliefs, long term aims, etc.
In pathological cases this will correspond to multiple personality
disorder. (My original message in this thread provoked a private
communication from a therapist supporting the link between the
architecture and MPD).
In simpler, more commonplace, cases the changing "occupancy" of the
meta-management system will correspond to the sorts of (partial) changes
that occur (usually unconciously and non-deliberately) when a person who
is gentle and caring at home becomes dangerous and agressive when
driving a car and hard and calculating when managing her staff at the
Being hypnotised may be yet another case.
How these changes occur is one of the gaps to be filled. It's not just
changes in the meta-management processes, of course.
The explanation of such changes need not refer to any "real you"
deciding which personality should get into the driver's seat in
E.g. such switching between high level meta-management control states
could be one of the more sophisticated functions of part of the purely
(People differ in the extent to which they are aware of such things and
can change them when they decide they need changing.)
In other words, I was not offering a theory that presupposes that
"there's a real unique ME". That's just a romantic myth for which
theologians, poets and teenagers are to blame.
There's lots more to be said about how these different temporarily
dominant *personalities* (or sub-personalities in the more normal cases)
might monitor different subsets of internal events, etc, evaluate
options differently and select different strategies for thinking,
deliberating, deciding, etc. I expect psychologists have already
documented some of the cases.
It's worth stressing that I never intended to claim that the
meta-management system was in TOTAL control of deliberative processes.
As a number of respondents have pointed out to me, and many poets,
composers, mathematicians and others (including Freud) have stressed in
the past, human beings are normally constantly subject to having
unselected thoughts, ideas, reminiscences, etc.
Even when you are busy working hard on a purely intellectual task (e.g.
responding to an email message, designing a program, proving a theorem)
much of the detail simply emerges from the inaccessible murk.
That's why I included "or do not contradict" when I wrote:
Conjecture: being in control of thoughts, or attention,
involves being in a state where the deliberative
processes and other internally monitored processes
conform to, or do not contradict, goals and plans of
the meta-management system.
Don't look for any "you" in there.
(b) The second task is to diagnose the causes of the demand for
something to `identify with' "the you" .
I suspect one source is social pressure to assign blame or praise for
purposes of social control (understood in the broadest possible sense,
including what you'd like your friends and loved ones to do to you).
This can function BOTH via mechanisms that we are not necessarily
conscious of (including reinforcement learning systems in the reactive
sub-architecture) AND ALSO via changes within meta-management processes.
For the latter to work, the meta-management processes must already have
some (genetically determined or otherwise inculcated) preference for
taking account of the judgements of others, including judgements about
how one should think, how one should reason, how one should assess
alternative ways of deliberating, i.e. judgements about the very
material the meta-management system is concerned with.
(A curious feature of such an architecture is the way it can also
treat information about its own judgements in the same way as it
treats information about the judgements of others. When people rebuke
themselves, blame themselves, exhort themselves, etc. these *social*
control mechanisms may perhaps be triggered -- to good or bad effect,
depending on the case).
Insofar as part of the architecture is not amenable to such influences
we start making (spurious) distinctions between the "real" you who can
be praised or blamed and the other bits.
There are probably other influences also.
Anyhow, instead of all this we should allow that the really
real you has different components: i.e. you have a complex
(Kant: a good man is a man with a good will. Pity about
the nether regions.)
> We can imagine an alternative architecture
Is this a different *architecture* ???
> whereby I do not identify with
> the goals of the meta-manager, but instead with the edicts of one of the
> lower layers.
I don't know what "identify with" means here.
I can think of a possible interpretation. E.g. you try
becoming a certain sort of hippy or whatever.
But it's still just meta-management choosing strategies
for the deliberative system: "OK now -- Just give in to all
those urges from now on. Eat, drink and be merry for tomorrow
(Or have a hang over.)
> ..What might such an alternative be like?
Note: it's not an alternative. It's a particular form of
> ...Well, maybe the
> goals of the meta-manager would be relayed to me in the form of auditory
> hallucinations, voices of the gods, ("Pete don't forget to pick up a
> six-pack at the store") which I obey unless there's some more urgent
> buisness to attend to, like fighting off a wild boar, or something.
Why would you call these ghostly goals the goals of the meta-manager?
The meta-management mechanisms are *defined* as those that allow
monitoring, evaluating and controlling of deliberative processes.
If some other part of the architecture is able to generate voices
"heard" by the meta-manager or visions "seen" by the meta-manager
that's an interesting possibility, but you need to be careful how
you describe it.
It was an important feature of the architecture I described that
reactive mechanisms could disturb, interrupt, divert, deliberative
processes (including meta-management processes).
(Think of those cases where a horrible jingle that you have heard,
e.g. a silly advertising tune, just sticks in your brain and keeps
> So anyways, it seems that one of the many details that will need to be
> worked out in a theory of self control is to say a bit more about what
> the *sense* of self is and why it gets glued to one layer and not the other.
You may find it useful to read an old paper by Marvin Minsky :
`Matter Mind and Models' originally published about 30 years ago.
It is now available online
[Link corrected 2 Mar 2014]
He makes a point which is close to a point made independently by P.F.
Strawson (and no doubt many more) namely that insofar as we need to
include ourselves (and other people) in our models of the world we need
models with a "bipartite" structure, where one part is mechanical,
physical, geometrical, etc. and another part is concerned with the
information acquired, stored, processed, used, i.e. social and
psychological experiences. (My loose paraphrase.)
The same would be true of any intelligent robot.
Similar distinctions can be made about the information that nodes
in a computer network have about the network, e.g. there's information
about the various physical nodes and links and their locations etc. and
there's information about the types of software, services provided,
information stored, etc. at various locations.
Each node in the network may have a model of the whole network,
including itself, and may include a similar division in its model
If some of the nodes have such information and other lesser nodes do
not, then those that do may include, in their models of themselves, the
fact that they have models of themselves, and perhaps also something
about the contents of those models. The recursion need not be viciously
Since all such information is *necessarily* an oversimplification of
reality our models of ourselves will be oversimple in various ways: both
our models of ourselves as physical and our models of ourselves as
things with minds.
No doubt there are strong cultural differences.
Anyhow, I suggest that instead of trying to `identify with' a part of
yourself you simply live with the fact that you are very complex, and
although you can distinguish cases when you are in control of
your thought processes and cases where you are not, most of what goes
on in your mind just goes on.
Aaron Sloman, ( http://www.cs.bham.ac.uk/~axs )
School of Computer Science, The University of Birmingham, B15 2TT, England
EMAIL [log in to unmask]
Phone: +44-121-414-4775 (Sec 3711) Fax: +44-121-414-4281
Back to: Top of message | Previous page | Main PSYCHE-D page