(previously copyrighted)
My Extreme Markup Languages 2006 presentation was titled
"Metaprogramming, Ontologies and Still Nobody's Home". The important
points from that presentation / audience panel will be presented in
this group posting, starting with 'awareness'.

Awareness what is it? What's it good for? Here we discuss awareness in
a machine. We are not talking about awareness as it is in a human, nor
even in any known animal. The awareness discussed here is
metaphorical, it is not human awareness. Many people use the term
awareness as a synonym for consciousness. No consciousness, as
experienced by humans, is discussed here. What is discussed here is
not offered as a model of consciousness in humans.

I have presented conference papers on computer software technologies
which implement some of the prime systems needed to support the
performing of metaphorical / analogical processing. I have also
presented material which mentions information sources (publications)
which describe the general belief ((Jay Ingram, Theatre of the Mind))
that much of what humans do, in terms of mental processing, on a daily
basis occurs, by and large, in the pre-conscious (some people,
incorrectly, use the term subconscious to refer to the same mental
processing) and only now and again do the RESULTS of all this myriad
of processing pop into ("appear" in) consciousness. See also Jeff
Hawkins book "On Intelligence"; and "Scripts, Plans, Goals and
Understanding", and other books by Roger Schank, which discuss his
technology of "Reminding".

Awareness, the type that does not include consciousness, is very
valuable in humans. An example of such non-conscious awareness is that
which occurs in the human visual system. Detailed vision only occurs
in a rather small cone of coverage oriented directly where the eye is
pointing and extending only a rather few degrees to each side of
straight ahead. The eye performs not consciously controlled saccades,
flitting from one direction of gaze to another. This is necessary
because the eye responds to change, if an image can be kept constant
on the retina or unchanging it will automatically disappear. There are
"rods" in the _periphery_ of the eye which are not colour sensitive
nor are these in large enough numbers to achieve an image but they are
quite sensitive to light. If something moves in the input field out in
this (visual) periphery area we "automatically" (and quickly) turn our
head and look directly in that general direction. This is a common
phenomena in all humans. It is important that the reader understand
that at no time was there any kind of "image" experienced from out in
this peripheral area, nor was there any kind of "light", nor "flash",
nor "pulse" (of light) _experienced_! The only thing that happens is a
sudden "need" to turn one's head and point one's face in a certain
direction. It is simply a feeling, a somewhat pressing one. [Jay
Ingram in his recent book, and also Antonio DeMassio in his books talk
about many such phenomena in humans.] One's head usually just darts to
one side and a great deal of attention is placed upon gaining
(conscious) information about what exists and what is happening in the
new direction we now face. A similar thing may happen based on our
hearing. Some sound occurrence, or cessation, has occurred and our
attention is strongly focussed in the direction of that change or event.

While we never consciously "see" any image or light from the periphery
we do sometimes seem to consciously "hear" the sound that grabbed our
attention. The almost inaudible snap of a twig in the forest (which
seems to happen more frequently after the sun has gone down) catches
our attention.

Okay, okay what does all this have to do with computers and
computational metaphor. Well, now comes the discussion which makes the
connections. We started off saying "the awareness discussed here is
metaphorical." We saw how humans daily use obviously complex
processing (in their brain) to detect things of interest "in our
visual realm even though we don't "see" them. A similar thing happens
with hearing. The result is that a noticeable "tap on the shoulder" is
delivered to consciousness, which then dutifully provides us with a
(conscious) "experience" (of something). In the case of vision we
discussed no visual experience has yet occurred , all we experience is
the need to use "full attention" to quickly get (new) information
about surroundings, particularly via vision. We also experience that
our head is turning, but there is no memory of having consciously
initiating that head turning, or any reason to do so. We do
experience a strong _sense_ of the _direction_ we should be attending to.

Simply put, motion of an object causes a change in the light pattern
of the (relative) field of vision it occurs in. Theatre marquee lights
cause the sensation of apparent motion in this way. The processing
which performs this "detection" is far more than a simple (UN*X style)
"dif" of two rasters. That all of this processing "completes" below
the level of consciousness is of interest.There is a constant
maelstrom of "mental processing" going on "below the level of
consciousness" (a qualitative and subjective METAPHOR). "We" (our
"conscious selves") which "experiences" is mostly ignorant and
clued-out, initiate a strong _conscious_ attention but only as a
result of "being tapped on the shoulder and handed a cue card" [by the
"society of processes" which ceaselessly toil "beneath" our "awareness".]

So what!? It is on the one hand cheering (for "computer people") that
in fact the greater part of our functional intellect is pre-conscious
(aka (incorrectly) "unconscious", or "below the level of (conscious)
awareness.")

Nobody yet knows what consciousness IS. So we can't program computers
to "do" consciousness, we can only program computers to perform
(non-random) things which WE already understand. Consciousness is not
required for most of the intelligent behaviour we do every day. This
gives hope that the processing "below the level of consciousness" can
be modelled in a computer, since the spectre of consciousness need not
be included there.

A first-order view, a single "alpha-level model" of awareness (the
kind that does not include consciousness (like the
un"see"n-visual-objects mentioned earlier)) is "detection" (possibly
with associated action triggered by it).

This is exactly the kind of awareness which a (military or business)
_Situation Awareness_ system has / uses. It is an unconscious
DETECTION of stuff, with subsequent "reporting" of it and also
possibly associated triggered responses. Some database systems have
"triggers" and within the sense under discussion have a primitive (if
not confined) awareness. Databases and situation-awareness systems
contain knowledge but do not contain "knowing". These systems do not
"know what knowledge they have", in knee-jerk stimulus-response
fashion they can "report" the results of a query, which smites them
soundly on the head from outside their system, but they have no
"knowing" of their own content. That is because they are not situated.
[Situatedness will be one of the threads of this group's discussion
across time.]

Is an exploding stick of dynamite an example of a stimulus response
system? (nope) One doubts that there is any mental processing going on
to cause a leg to "kick" (or "jerk") when the doctor whacks it with
his little hammer. The "jerk" or "twitch" is initiated "involuntarily"
by systems "below the level of consciousness" (some think way far
below), no thinking, analyzing, or even consciousness is even required
for the twitch response. We all find it convenient that our
conscious-self experiences 'that the leg is twitching' and often we
also "feel" (experience) the thunk (impact) of the hammer.

I presented at Extreme Markup Languages 2006 a "starter model of
awareness". Awareness, I said, may be implemented by having a planning
program monitor the (set of) activation record of a system and perform
plan detection with the contents it finds there.

(For example, see: A Formal Theory of Plan Recognition and its
Implementation,
Henry A. Kautz.;
Using Plan Recognition in Human-Computer Collaboration,
Neal Lesh, Charles Rich and Candace L. Sidner;
Decision-Theoretic Approach to Plan Recognition,
Wenji Mao, Jonathan Gratch)

This is to say that a program "watches" what actions a system performs
/ emits and also the effects produced by such actions and (attempts
to) determine which goals the system might be pursuing by (attempting
to) determine which plans might be being used by the system.
Individual actions and sequences of same, as located via the
activation record, are members of various plans. The awareness so
obtained , then, consists of (hypothesized and recognized) goals,
plans and activities / actions. The actions convey "how knowledge"
and the plans and goals provide "what knowledge". In SHRDLU-esque
fashion the latter can be used to convey "why knowledge".

In this way a system can not only simply detect what actions the
overall system is executing it can use inference to "hypothesize"
which plans and goals "are being pursued" [yet another spatial
metaphor]. This means that a plan-detecting-planner (program) can
provide a _modicum_ of awareness as we believe it to be in humans sans
the "experience" aspect.

By including metaphorical-plans (like ride a motorcycle across country
instead of a horse) in this capability the system is able to free
itself to some (important) extent from the concreteness and
circumscription of standard literal-planners. Conceptual metaphors
(black hole, event horizon, gravity well, yaddayadda) and
metaphorical-planning is another thread of this group.

Other threads are contexts, what they are, how they are made and how
they define a situation. Contexts provide a viewpoint. Conceptual
metaphors, from the point of view of George Lakoff, myself, Gentner /
Forbus and others.

Other topics covered in my Extreme Markup Languages 2006 presentation
were Posted Agenda, Blackboards and an SVG program that displays a
picture of an analog clock with the seconds hand sweeping out the
seconds and the other hands show local time.

It is one thing to have a collection or library of executable actions
but it is another thing entirely to know you have that collection and
what it means that you can do (using those actions). Reflection
(including SVG reflection) and Annotation are yet other threads that
will be covered in this group.