At 7:24 5-5-96, Keith Sutherland wrote:
Bernie Baars' appeal to our everyday conscious experience (and without a single reference to "folk psychology"!) was a welcome breath of fresh air, ...
Hear hear! ...
But does Baars' concept of the self work? ... the emphasis is very much on a plurality of self-systems. Now all this makes good analytic sense, but our experience is (almost always) of a single self -- "me", not "we" -- albeit with different (social, motivational, emotional etc.) aspects. So we are back with Schroedinger's problem of the one and the many.
How about the following sketch for a solution? These systems have common access to several of the body's internal resources, let us suppose: they can control motor function, have access to the first veto on linguistic output or the earliest clues of comprehension of linguistic input, etc.. But they all have access to the individual's memory of itself, ie its egocentric longterm recall of 'what things were like then'. This is a continually updated store of impressions (with all kinds of extra information added to them as commentaries, beliefs, etc.; no doubt very complex and dynamic, with other unconscious processes continually updating it, repairing important inconsistencies in it, making it conform more to our desires, etc.: but all that matters here is that it be unique.) Call it our narrative memory. One important special aspect of the representational language or code used by the narrative memory is a term which plays the role of the "essential indexical"; that is, it stands for th individual itself in the systems's view of its own history. ('Now' is the time of the EI, 'me' is the agent of the EI, 'here' is the location of the EI, etc. The need for such a special term can be demonstrated on purely logical grounds, by the way.)
Many of these accesses have lockouts, so that only one of the various competing self-systems can have control at once. (I dont know exactly how a lockout is switched, though this will be a very important aspect of how the overall system behaves, and may itself be in the charge of lockoutable processes. One of the functions of pain, for example, may be as means of defeating otherwise secure lockouts to force attention on an injury.) The architecture is designed so that many of the 'lockouts' are usually coordinated, so that when a self-process has control of one it tends to get most of the others (though this is by no means guaranteed and often fails to be completely comprehensive, eg "when my left hand doesnt know what my right hand is doing"). This convergence of command therefore might be thought of a special master key which is passed between the self-processes and which gives the possessor the unique key to control of resources. One special such control is the 'right' (in this imagined quarreling family of selfs) to place itself at the EI location when changes are made to the narrative memory.
Now, since these processes all share a single narrative memory, and since each, when in control, is the locus of the "I" in the systems awareness of itself, the system will always have a single sense of its own identity which is continuous through time; moreover, this "I" will usually know that it is in control of itself and of its body; and this will not be an illusion, let me emphasise, but indeed be literally correct.
Some such model would seem to answer Schroedinger's qualms yet be consistent with Baars intuitions.
(These ideas are not new, by the way. Minsky, in particular, has described similar constructions in far more detail.)