If a tree casts a shadow is it telling the time?

Russ Abbott
Department of Computer Science
California State University, Los Angeles
Los Angeles, Ca, USA
Russ.Abbott@GMail.com

Abstract – Physical processes are computations only when completely automated and involves no humans. [Emphasis ad-
we use them to externalize thought. When suitably formu- ded.]
lated—with nature as an oracle—agent-based computing In other words, Google’s computers are reading your email
in an open environment represents the current consensus —but no human beings are. That most people find this re-
view of how to think about the way much of the world assuring illustrates the intuition that it’s what goes on in
works. the mind of a human being that matters.
Keywords: Thought tools, computation, thought, ideas, One might object that if a computer is reading one’s email
models, unconventional computation, agents, agent-based and storing its contents in a Google database, a person
computing. might be able to read it later. That’s quite true, and the fact
that only Google computers (and not Google employees)
1 Introduction are reading one’s email does not guarantee one’s privacy.
In the preface to the first edition of the International The point we wish to emphasize, however, is that if no per-
Journal of Unconventional Computation, the editorial son ever reads the original or stored copy of one’s mes-
board (IJUC Editorial Board, 2004) welcomes papers in sages, then most people will justifiably feel comfortable
“information processing based on physics, chemistry and that their privacy has not been violated.
biology.” But it is left undefined what it means to say
(a) that a physical, chemical, or biological system is doing After all, all email is read—in one sense of another—by a
information processing or (b) that information processing number of computers as it passes from sender to receiver.
is “based on” physics, chemistry, or biology. In this paper No one has ever worried about that. The moment of viola-
we explore the implications these questions tion occurs when some living being becomes consciously
aware of one’s personal information.
This paper has been inspired by the questions:
But, one might argue, the kind of reading that occurs when
• What is computation? a computer transmits a message along a communication
• How can computation be distinguished from other channel is qualitatively different from the kind of reading
that occurs when a Google computer determines which ads
processes in nature?
to place next to a message. The former kind of reading
Our primary conclusion will be that physical processes are treats messages as character strings. No meaning is extrac-
considered computation when we treat them as external- ted. The kind or reading that Google computers do extracts
ized thought. In other words, it’s all in the mind. (or attempts to extract) meaning so that related ads can be
We will also argue that unconventional computation as displayed.
well as other models of computation ranging from ubiquit- This raises the question of what we understand by the term
ous computing and interactive computing to systems of meaning. That’s clearly a larger topic than we can settle
systems are all variants of agent-based computing. We will here. But our short answer—and the reason we believe
agree (although on uncontroversial grounds) with Wegner that most people find Google’s answer reassuring—is that
that this model of computation is more expressive than our intuitive sense of meaning has something to do with an
Turing machines. idea or thought forming in a mind.1
Is Google reading my email? At this stage in the development of technology, most
That’s the first question in the Google Gmail help center people don’t believe it makes sense to say that an idea has
(Google, 2006). This question arises because Gmail places formed in the mind of a computer—or even to say that a
ads next to email messages, and the selection of ads is computer has a mind at all. We may speak informally and
based on the contents of the messages. Google’s answer to say something like “the computer is doing this because it
this question has varied over time. On March 13, 2006, the thinks that.” But when we say these sorts of things, we are
posted answer was as follows. deliberately speaking metaphorically; we don’t really think
Google computers scan the text of Gmail messages in order to of computers as actually “thinking.” Until we start to think
filter spam and detect viruses, just as all major webmail ser- of computers as having human-like minds—minds that
vices do. Google also uses this scanning technology to deliver have subjective experience, minds in which ideas can form
targeted text ads and other related information. The process is
1
This clearly is different from the formal semantics sense
in which meaning refers to a mapping from an expression to a model.

Abbott – 10/17/2008 If a tree casts a shadow … 1/13

—then most people will feel comfortable with Google’s This is not intended to be mystical or profound—just a
reply that its computers, but no human beings, are reading statement of a brute fact: an idea is something that occurs
one’s email. only in someone’s mind. The ideas in this paper exist only
in the mind of the author and the minds of the readers as
To come the author and readers are thinking them. These ideas
Section 2 continues the discussion of thoughts and ideas
don’t exist, on the paper or on the computer screens on
and introduces the notion of thought tools. Section 3 offers
which these words appear. They don’t exist in the com-
a brief history of thought tools. Section 4 considers how
puter memory in which these words are stored. Just as the
computation might be defined. Section 5 switches gears
moment at which an invasion of privacy occurs is when
and discusses the agent-based computing paradigm as
some being-with-a-mind learns something personal about
more than just an approach to programming and modeling
us, an idea exists only when someone is thinking it.4
but as common to many of the new ways we think about
the world. Section 6 draws a grand conclusion. We go to such lengths to make this point because our fun-
damental position in this paper is that computations, like
2 Thinking and thought tools ideas, are also mental events. But computations are mental
events that we have figured out how to externalize in such
If a tree grows in a forest, but no one counts its rings is it
a way that we can use physical processes to perform them.
counting years? Is it performing an unconventional compu-
tation? If a tree grows in a forest but no one knows it’s When a tree grows rings, it just grows rings. But when we
there, is it instantiating the idea of a tree? These questions use that tree-ring growth as a way to count years, i.e., to
have the same sort of answers as does Bishop Berkeley’s help us work with ideas such as the idea of a year, then we
famous question: if a tree falls in a forest with no one can say that the tree has performed a computation—for the
around to hear it, does it make a sound? purposes of this conference, an unconventional one.
The question is not as difficult as it seems—although our When a computer runs is it computing? Our answer is the
answer is not Berkeley’s answer.2 The issue is not whether same. A computer is computing only when it is understood
the physical events occur. Of course they do. But if there is to be performing some externalized mental activity.
no one around to hear the sound that they make there will
be no subjective experience of a sound.
Disciplined thought
In the view presented here, computing is the use of physic-
If a tree falls in a forest, it generates (what we call) sound al processes to help us think in a disciplined manner. The
waves whether someone is there to hear them or not. Those qualifier in a disciplined manner is intended to distinguish
sound waves are physical events. If powerful enough, they between (a) thinking which, when understood broadly may
can damage things in their path, and they can shatter ob- refer to whatever arises in consciousness (or perhaps even
jects that vibrate in sympathy. All this will happen whether unconsciousness) and (b) thinking which, when understood
or not there are any witnesses. in a more narrow sense refers to coherent, organized
But if no one is there to hear the sound, if no being has a ideation, i.e., the creation, review, and analysis of mental
subjective experience of the sound, then no sound will be content that can be formulated as an organized body of
heard. No subjective experience will occur in any being ideas.5
capable of having a subjective sound experience. Besides what would traditionally be called ideas, we also
The same holds for ideas. Like the subjective experience of include as within the realm of disciplined thought such
a sound, the idea of a tree exists only as a subjective exper- mental phenomena as (mental) images, (mental) melodies,
ience. If no one has that subjective experience, then a tree (mental) harmonies, etc.
without anyone knowing about it will not be instantiating A brief history of the internalization and then
the idea of a tree. the externalization of thought
Even if one grants that the idea of a tree is exactly the right One may trace one thread through the history of discip-
way to describe that particular aspect of nature, that idea lined thought as a process first of the internalization and
exists only as an idea, and it exists only in the mind of then the externalization of thought.
someone who is thinking it. Ideas exist only within the Initially we (humanity) looked outward to answer questions
realm of mental events, i.e., as subjective experience. In about how to make sense of the world. Imagine the diffi-
saying this we are taking an explicitly anti-Platonic stance:
there is no realm outside the mind in which ideas exist on 4
This position requires some care in formulation. If an
their own.3 idea exists only when someone is thinking it, what does it mean to say that
two people have or had the same idea. We believe that these issues can be
worked out.
2
Berkeley’s answer is that it does make a sound because
5
It is unlikely that any body of knowledge can be formulated completely on
God, who is always everywhere, hears it. a conceptual level. No idea is completely independent of experience. After
all, the term idea itself is understood to refer to a mental event which is an
3
We are not taking a stand on nominalism vs. realism. aspect of subjective experience of a living being. The very act of thinking
Although we believe that our (human) ideas about how nature should be is both a subjective experience and a physical event. So there will be in-
described are not arbitrary and that entities other than the elementary definable primitives in any discipline. But for the most part, the kind of
particles do exist (see Abbott, 2006), that is not at issue here. In this paper, thinking to which we are referring is thinking that would be considered
we are making a point that should be obvious on its face: an idea is a men- disciplined, rigorous, and intellectually honest, i.e., the kind of thinking
tal event. As such it exists only when someone is thinking it. that is valued in an academic environment.

Abbott – 10/17/2008 If a tree casts a shadow … 2/13

culty early man faced. What should one think about the value. Thus was born the externalization of thought. The
world? What does it consist of? How should one think fundamental issues of externalization are the following.
about how it works? What frameworks should one use to
organize one’s experience? These are the fundamental • How can thought—which, after all, exists only in
questions of metaphysics. Even asking such explicitly for- the mind—be represented outside the mind in forms
mulated questions was initially beyond our reach. that allow us to work with it?

Not knowing what else to do, we looked to sources of what The standard forms in which thought is represented
we hoped were authority: priests, oracles, prophets, sacred outside the mind are (a) drawings, diagrams, sound re-
writings, divinities, etc., to tell us what to think. We looked cordings, and images and (b) symbols such as words,
to these sources to tell us what thoughts about the world to numbers, and other symbolic notations. These categor-
install in our minds.6 ies correspond to what will later become the distinc-
tion between analog and digital computers.
Worse, we often fought with each other about whose
sources of knowledge were right. In a recent op-ed piece • Can we devise tools that provide mechanisms for
(Albacete, 2006), Lorenzo Albacete, a Roman Catholic manipulating and otherwise working with such ex-
priest, laid out the position of those who fear the use of re- ternalized thought?
ligion as a source of knowledge. From here on we will refer to thought tools, tools that al-
For [nonbelievers], what makes Christianity potentially dan- low us to work with externalized thoughts, as computers.
gerous [is not its other-worldliness but] its insistence that faith Today’s word processing programs present a challenge
is … the source of knowledge about this world and that, there-
fore, its teaching should apply to all, believers and nonbeliev-
with respect to the sort of thought tool it is. Consider the
ers alike. … difference between a word processor and a calculator.
The Roman world celebrated religious pluralism and was will- • Calculators manipulate number representations,
ing to welcome Christianity as an ethical or “spiritual” option, which represent our ideas about counting. The oper-
but not as a source of truth about this world—that was con- ations calculators perform, e.g., the arithmetic oper-
sidered to be the realm of the philosophers. ations, etc., on number representations correspond
At that time, Christianity would not accept a place with the re- to operations that we think of as being performed on
ligions of the empire. It saw itself as a philosophy, as a path to count values. Calculators are easily understood as
knowledge about reality, and not primarily as a source of spir- thought tools to help us think about numbers.
itual or ethical inspiration. …
• Word processors manipulate words, which corres-
Indeed, throughout history Christians have used this claim to pond to ideas. The operations word processors per-
justify their intolerance of other views, even turning to viol- form on words do not correspond to operations that
ence in order to affirm and defend their idea of what is true.
The events of Sept. 11, 2001, reminded us that this unhappy
we perform on the ideas those words represent. The
tendency was not limited to the Christian faith, but seems in- operations that word processors perform do corres-
herent in religious belief. If a god offers absolute truth, then pond to operations that we perform on meaning rep-
those who disagree with that god's teachings are enemies of the resentations. Word processors are thought tools with
truth, and thus harmful to society. respect to our ideas about organizing words to con-
vey meaning but not (yet) with respect to the ideas
As Albacete notes, by the time of the Roman Empire, the
those words represent.
use of religion as a source of ideas about how nature works
had been discarded by enlightened thinkers. Greek and Ro-
man philosophers believed that they themselves could be a 3 The history of thought tools
source of knowledge about the world. The history of early computing may be traced along three
paths. Each path traces devices that help us think about a
The step from looking for external sources of knowledge to particular (and fundamental) subject area: time, counting
supposing that perhaps we can figure it out for ourselves is (arithmetic), and space (geometry).
what we are referring to as the internalization of thought—
attributing to oneself the power to produce thoughts of Time computers
value and rejecting the notion that thoughts must originate We used natural processes to help us express our ideas
externally to be valid. about time—the daily, monthly, and yearly cycles of the
earth, moon, and sun about each other. Not to beat this
Externalized thought and tools to work with it point into the ground, but day, month, and year are ideas.
If human beings can be sources of valid thought, then tools As ideas, they exist only in the mind—no matter how ac-
of thought, means and mechanisms outside the mind that curate or true they are as descriptions of nature.
help us formulate and work with our thoughts, also have
The first (analog) time computers were the actual processes
6
One wonders what priests, oracles, prophets, and other that corresponded to our thoughts. The rising and setting
human authorities believed about how the ideas they transmitted arrived in of the sun were the physical events that we used to keep
their own minds. Perhaps they believed that the ideas had been implanted track of the mental events: the start and end of a day.
in their minds as a result of their special status or as a result of some spe-
cial words or rituals that they performed. Presumably they didn’t believe
that they themselves made them up. Most likely they didn’t ask themselves
this question.

Abbott – 10/17/2008 If a tree casts a shadow … 3/13

Similarly for the moon. Yearly events such as river flood- Patterns and shadows themselves are not capable of mov-
ings and the comings and goings of the seasons helped us ing either under their own power or as a result of some
keep track of the mental event: the yearly solar cycle. force being applied to them. Neither shadows nor patterns
can propel themselves. Nor can one push or pull them.
These natural processes might be considered early primit-
ive analog computations. The physical analogs that we The mechanisms that produce both shadows and Game of
used to help us think about and track mental events were Life patterns are fixed—and thoughtless. The Game of Life
actual physical events that corresponded to the mental rules are simply rules for how and when cells turn on and
events. off. The relative motion of the sun, a tree (or other shadow
casting object), and a background surface is equally fixed
It didn’t take us long to invent more sophisticated analog
and thoughtless. Yet in both cases, we can use the gener-
computers. The sundial, for example, is an analog comput-
ated patterns to represent our thoughts.
ing device. The position of the sun’s shadow is an analog
for the mental event time-of-day which corresponds to the Using Game of Life patterns we can generate very complex
physical event relationships between the relative positions ideas: patterns that fill unbounded portions of the grid, pat-
of the sun and the earth. terns that (can be used to) count, patterns that can be used
to generate prime numbers, and patterns that can be inter-
It is worth noting that with the sundial we started to ar-
preted as Turing Machine computations. Similarly, we can
range physical materials to help us track our thoughts. In
interpret sun/object/ground shadow patterns to help us
building sundials we set up shadow casters, which in con-
think about ideas such as the time of day or day of the year.
junction with the sun and markings on the surface on
which the shadow is cast, help us track (our ideas about) We have been quite creative about this. We apparently used
the passing of the day. Presumably this was not a very sig- shadow patterns to help us think about the time of day and
nificant step from simply using existing shadow-casting day of the year before we understood the mechanisms that
objects, e.g., trees, for the same purpose. (If a tree casts a caused those shadows patterns to be connected to daily or
shadow, is it telling the time?) yearly cycles. We had the idea of a day and a year, and we
used shadows to help us think about them before we knew
Using epiphenomenal shadows to tell the time what produced them. In both cases, we used patterns gen-
The use of shadows as thought tools deserves special atten-
erated by fixed rules to help us think.
tion. In (Abbott, 2006) we discuss naturally occurring en-
tities—entities that persist independently of human obser- Number computers
vation. These include atoms, molecules, animals, organiza- Apparently we started to count quite early. Bones with
tions, hurricanes, galaxies, and most of the things we intu- notches carved into them appeared in western Europe
itively think of as entities. Shadows are not in this cat- 20,000 to 30,000 years ago. There is apparently evidence
egory. of the use of a tally system—groups of five notches separ-
ated from each other. (Maxfield, 2006).
A shadow, after all, is that portion of a background or sur-
face that is not illuminated by a light source because an ob- With tally systems not only did we mark physical materials
ject is blocking the light. The light source may be an en- to help us keep track of numbers (which are also mental
tity; the object blocking the light may be an entity; the events), we also invented ways to make counting easier by
background or surface on which the shadow is cast may be the specific way in which we arranged these markers, i.e.,
an entity. The shadow itself is not an entity. At best, a in groups. Although notches are digital, this is analog com-
shadow—and more importantly, a moving shadow and the putation at both a direct level (individual notches are entity
leading edge of a moving shadow—are epiphenomena of analogs) and a basic meta-level (notch groups are collec-
the possibly changing relationships among the light source, tion analogs). (See discussion of analog computation be-
the background, and the object. low.)
Although the mechanisms are completely different, a With these primitive analog computers we separated the
(moving) shadow is very much like a (moving) pattern computational process from its dependency on natural pro-
(such as a glider pattern) in the Game of Life. In both cesses. Sundials and astronomical masonry depend on the
cases, the apparent object (the shadow or the Game of Life sun and the stars. Counting depends on nothing other than
pattern) consists of illuminated/not illuminated or on/off human activity.
elements on a surface. Over time, the on/off elements may At approximately 5,000BC, the abax, followed soon by the
appear to move across the surface. In fact, the on/off ele- abacus, was invented. The abax was originally a dusty sur-
ments don’t move; it is only the pattern of on/off elements face that was used—by drawing lines in the dust—to rep-
that appear to move. resent numbers using positional notation. Pebbles or other
But patterns don’t move either. With both Game of Life markers could also be used as counters in the positions.
patterns and shadows, portions of the surface are on and More sophisticated versions had grooves.
portions of the surface are off at any given time. That the Once we invented computational devices that were essen-
same or similar on/off configurations appear first at one tially independent of non-human physical processes it was
location and then at another is strictly a consequence of the a short step to the use of written notation instead of pebbles
mechanisms that generate those shadows and patterns. and grooves. At approximately 3,000 BC cuneiform writ-

Abbott – 10/17/2008 If a tree casts a shadow … 4/13

ing on clay tablets in positional notation was known in The idea for the machine came to Lull in a mystical vision that
Babylonia. appeared to him after a period of fasting and contemplation. It
was not unusual in that day before the dawn of scientific obser-
Space computers vation for scientific advances to be attributed to divine inspira-
Besides time and numbers, the Pythagoreans in Greece and tion. He thought of his wheels as divine, and his goal was to
Euclid in Egypt developed ways to think about space. We use them to prove the truth of the Bible.8
know that early geometers thought about construction is- In reality, Lull may owe what he considered divine inspiration
sues. The straight edge and compass were their thought to the influence of mere mortals. While studying Arabic and
tools. They used them to externalize, to create representa- working as a missionary to convert the Moors to Christianity,
tions of, and to manipulate the ideas of straight lines and he came across a device that Arab astrologers used called the
circles. These were the (human-powered) computers of the zairja. It used the 28 letters of the Arabic alphabet to represent
early geometers. the 28 categories of thought in Arab philosophy. By combining
numerical values associated with the letters and categories,
Is it reasonable to call abaci and geometers’ new avenues of thought and enlightenment were opened.
tools computers? Lull’s fame as a mystic spread throughout Europe during the
Even though abaci and geometers’ tools are completely in- late Middle Ages and Renaissance. His followers were called
dependent of non-human physical processes, i.e., they are Lullists. Centuries later, the German mathematician Leibniz
entirely dependent on human activity to make them oper- acknowledged the importance of Lull’s logic machine.
ate, we feel justified in calling them computers because
In “Gulliver’s Travels,” Swift satirizes the machine without
they are used according to mechanical rules. Even though
naming Lull. In the story, a professor shows Gulliver a huge
the source of power for an abacus is the user, the abacus contraption that generates random sequences of words.
user follows strict rules—rules which could be automated. Whenever any three or four adjacent words made sense togeth-
The fact that an abacus is human-powered doesn’t detract er, they were written down. The professor told Gulliver the ma-
from its status as a computer. At the dawn of the 20 th cen- chine would let the most ignorant person effortlessly write
tury, David Hilbert famously dreamed of automating math- books in philosophy, poetry, law, mathematics, and theology.
ematics. His dream was that a mechanical procedure would This may be the first use of non-determinism in comput-
allow one to determine whether or not a statement was ing.
provable from a set of axioms. Hilbert dreamed his dream
Soon thereafter William of Ockham discovered the founda-
before there were any devices that could carry out his ima-
tions of what were to become De Morgan’s laws of logic.
gined procedure—were it even possible to do so. 7 Yet Hil-
More specifically, from (Sowa, 2002):
bert’s dream was of a mechanical computational procedure
in the same sense that the rules followed by an abacus user (Ockham, 1323) showed how to determine the truth value of
is also a mechanical computational procedure. Today’s compound propositions in terms of the truth or falsity of their
computers are not different in kind from these pre-com- components and to determine the validity of rules of inference
puter computers. They just represent our success in har- (regulae generales consequentiarum) in terms of the truth of
their antecedents and consequents.
nessing processes of nature to carry out the mechanical
procedures in terms of which we have long thought. Thought tools vs. scientific instruments
Clocks, abaci, straight-edges, hierarchies, non-determin-
Thought tools for symbol manipulation ism, laws of logic, and other thought tools differ in kind
Beyond time, numbers, and space, we have for a long time
from microscopes, telescopes, and other scientific instru-
built thought tools to represent symbolic thoughts and rela-
ments of observation. The former are intended to allow us
tionships. (Sowa, 2006) has this to say about the Tree of
to externalize and manipulate our thoughts. The latter al-
Porphyry.
low us to investigate nature—to see what’s out there and
The oldest known semantic network was drawn in the 3rd cen- perhaps to see things that will require new ideas to under-
tury AD by the Greek philosopher Porphyry in his commentary stand them.
on Aristotle's categories. Porphyry used it to illustrate Aris-
totle's method of defining categories by specifying the genus or Of course thought tools also help us create new concepts.
general type and the differentiae that distinguish different sub- But the new concepts that result from the use of thought
types of the same supertype. tools tend to build on existing concepts. The new concepts
Another attempt to externalize symbolic thought has been that arise from the use of instruments of observation tend
credited to Ramon Lull in the late 13th century. (Smart to be new primitives. Thought tools are constructive; in-
Computing, 2006) describes it as follows. struments of scientific observation are reductive.9
Ramon Lull’s logic machine consisted of a stack of concentric
disks mounted on an axis where they could rotate independ-
ently. The disks, made of card stock, wood, or metal, were pro-
gressively larger from top to bottom. As many as 16 words or
8
Compare this with our notion of how oracles thought
about the manner in which ideas were installed in their minds. Lull took
symbols were visible on each disk. By rotating the disks, ran- the next step and imagined that ideas may just as well be installed in a
dom statements were generated from the alignment of words. mechanical device through similar divine intervention. All we would need
Lull’s most ambitious device held 14 disks. do is read them out.
9
Reductionism has recently received a lot of bad press.
7
We know now, of course, that no such procedure is pos- As explicated here, the reductionist impulse often leads to observations
sible. that require the development of important new ideas.

Abbott – 10/17/2008 If a tree casts a shadow … 5/13

Thought tools and conceptual models Nothing interesting came out of this student’s work—
It is commonplace (and correct) to say that to understand a it was done by mistake. Yet now that the idea is out
computer program (from the perspective of a user of that there—and it is out there because of a mistake in ma-
program) one must understand the conceptual model that nipulating externalized thoughts—it may be worth in-
the program implements. This may sound inconsistent with vestigating.
our notion of externalizing thought, i.e., that the thought 3. Externalized thoughts are generally not the same as
comes first and is then externalized as a program. the original even when no error or misunderstand-
1. Most users of most computer programs are not the ing occurs. This is simply a fact of life. To external-
people who developed those programs. It is the de- ize a thought, the thought must be expressed in
velopers’ thoughts which were externalized. 10 Users some symbolic or physical medium. Every such me-
must learn to understand the developers’ thoughts in dium is limited—often in ways that prevent it from
pretty much the same way that someone who reads a representing the original thought in its precise ori-
book is being asked to understand the writer’s ginal form. To take two obvious examples: (a) a
thoughts. computer’s memory is not unbounded even thought
one often thinks in terms of unbounded memory and
2. More importantly, a fundamental reason for extern-
(b) the way we choose to represent numbers may
alizing thought is to allow us to work with it. It is
limit their range and precision even though we tend
almost inevitable that when one works with one’s
not to think of numbers as being limited in either of
own externalized thoughts, be they expressed as
these ways.
computer programs or as words, one develops those
thoughts beyond the form they took when first ex- The state of the art of thought externalization
ternalized. Thus developers of computer programs Every computer application is a thought tool. The thoughts
often find that the programs they write implement that are being manipulated are the thoughts that are repres-
models that extend and modify their original ideas ented by the conceptual model implemented by the applica-
in important ways. So even software developers of- tion. More importantly every programming language is a
ten find themselves working hard to understand the thought tool. It allows us to externalize our thoughts about
conceptual models of the programs that they write. behaviors as computer programs. One writes computer ap-
plications in programming languages. Thus a program-
A wonderful example of this occurred in a recent class.
ming language is a thought tool for building thought tools,
Students were asked to write a program that evolved
i.e., a thought tool for externalizing thought.
iterated prisoner dilemma (PD) strategies. Each popu-
lation element was a strategy that determined whether It is important to realize that a programming language is
to cooperate or defect as a function of the player’s pre- itself a computer application. As a computer application, it
vious move and the opponent’s previous move. All implements a conceptual model; it allows its users to ex-
population elements played each other; the more suc- press their thoughts in certain limited ways, namely in
cessful ones were more likely to be selected as parents terms of the constructs defined by the programming lan-
for the next generation of strategies. This was a stand- guage. But all modern programming languages are also
ard introductory genetic algorithm assignment. conceptually extensible. Using a programming language
one can define a collection of concepts and then use those
One student misunderstood the model. Instead of
concepts to build other concepts. Because of that extensib-
building a population of strategies, he built a popula-
ility computers have become extraordinarily powerful
tion whose elements were (as it turned out) meta-
thought tools.
strategies, e.g., play Tit-for-Tat for three games then
play Pavlov for two games; etc. It’s important to em- We are still learning to use the power of computers to ex-
phasize that the student did this as an error; he simply ternalize thought. In one way or another, much of soft-
didn’t understand the assignment. ware-related research is about developing more powerful,
more specialized, faster, easier to use, or more abstract
This student was able to make this error because we
thought tools. We now have thought tools for images,
had discussed (named and externalized) (a) the various
thought tools for sound, thought tools for business, thought
known PD strategies and (b) iterated PD in terms of a
tools for science, etc. There is virtually no realm of thought
notation in which one could represent a player’s activ-
for which some thought tool has not been developed.
ity as a sequence of Cooperate/Defect moves. In his
confusion this student chose as elements to evolve, se- We also have increasingly powerful languages in which to
quences of named strategies externalize and work with our thoughts. The more we learn
the higher we ascend the mountain of abstraction and the
broader the vistas we see. Recent work in externalize
thought includes work in declarative programming (e.g.,
10
In large formally defined systems, the origin of the sys-
tem’s conceptual model is typically more diffuse. The eventual users (or logic programming, functional programming, constraint-
their proxies or representatives) may write requirements documents, which based programming, rules-based systems such as expert
may include use cases. A requirements document—or the use cases it may systems, etc.), meta and markup languages such as XML
include—often does (but need not) include elements of a conceptual mod-
el. Users and their proxies may also write concept of operations (CON-
(W3C, 2006.X) and its extensions and derivatives, the Uni-
OPS) documents, which also may or may not include elements of a con- fied Modeling Language (UML) (OMG, 2005) for repres-
ceptual model.

Abbott – 10/17/2008 If a tree casts a shadow … 6/13

enting a wide range of modeling concepts, and the Semant- There are numerous competing definitions of computation.
ic Web (W3C, 2006.S) and the OWL Web Ontology Lan- Along with the initial definition provided here, the following
guage (W3C, 2004) for externalizing how we look at the three definitions are often encountered:
world. With OWL we are continuing to work in a tradition • Rule governed state transitions
that dates back to Porphyry—and before.
• Discrete rule governed state transitions
What is the field of thought externalization?
It is not clear what name to apply to the discipline of • Rule governed state transitions between interpretable
thought externalization and manipulation. It would seem to states
be a branch of philosophy. But it isn’t metaphysics, epi- The difficulties with these definitions can be summarized as
stemology, or ethics. Nor is it either logic (which is more follows:
about how to draw conclusions than about what thinking is The first admits all physical systems into the class of compu-
about) or meta-philosophy (which is concerned with the tational systems, making the definition somewhat vacuous
scope, aims, methods, and goals of philosophy).
The second excludes all forms of analog computation, per-
It might be considered a branch of philosophy of mind, haps including the sorts of processing taking place in the
which according to (Eliasmith, 2004.P) is the brain.
branch of philosophy that is concerned with the nature of men- The third necessitates accepting all computational systems
tal phenomena in general and the role of consciousness, sensa- as representational systems. In other words, there is no com-
tion, perception, concepts, action, reasoning, intention, belief, putation without representation on this definition.
memory, etc. in particular. Standard problems include those of
free will, personal identity, mind-body problem, other minds,
Although Eliasmith wants to allow for “the sorts of pro-
computationalism, etc. cessing taking place in the brain,” his definition appears
to leave out asynchronous multi-threaded and distributed
But the discipline of thought externalization and manipula- computing since it defines a computation as a (single)
tion, takes as a starting point assertions (such as an idea series of state transitions. Perhaps “a series” may be inter-
exists when and only when someone is thinking it) which preted to include vector-style parallel computations, but it
philosophy of mind would analyze in depth. Nor is thought would seem to exclude non-synchronized multi-threaded or
externalization concerned directly with intentionality in distributed state transitions.
the sense of Dennett or the symbol grounding problem. It
takes for granted our ability to relate our thoughts to real- Even more importantly, we suggest that the notion of alter-
ity. Look up Dennett and the intentional stance. Is this the able rules is not well defined. Eliasmith requires that the
next step? Are we presuming the intentional stance? What rules governing the state transitions must be alterable in
was Dennett’s original insight? order to distinguish a computation from a naturally occur-
ring process—which presumably follows rules that can’t be
Another related field is cognitive science (Thagard, 2004). altered. But all computing that takes place in the physical
Cognitive science is the interdisciplinary study of mind and world is based on physical processes. If we set aside the is-
intelligence, embracing philosophy, psychology, artificial sue of the probabilistic nature of quantum physics, and if
intelligence, neuroscience, linguistics, and anthropology. we suppose that physical processes operate according to
As thus defined cognitive science is broader than thought unalterable rules, it’s not clear what it means to say that it
externalization and manipulation. The primary focus of must be possible to alter a set of rules.
cognitive science, at least according to Thagard, is how the This is not a matter of being difficult. Certainly we all
mind works. Thought externalization and manipulation is know what it means to say that one program is different
not so much concerned with how the mind works as with from another—that a program, i.e., the “rules” which gov-
providing tools that help minds at work. ern a computation, may be altered. But the question we
Of course just as research in kinematics and sports medi- wish to raise is how can one distinguish the altering of a
cine help produce better athletic performance, if we knew program from the altering of any other contingent element
more about how the mind works, we would probably be in an environment?11
able to build tools that would help it work better. It is the program that is loaded into a computer’s memory
that distinguishes the situation in which one program is be-
4 Defining computation ing executed from that in which some other program is ex-
In this section we return to the question of how to define ecuting. But a computer's memory is the environment
computation. We approach the question “from scratch,” within which the computer’s cpu (or some virtual ma-
without assuming that computation is externalized chine) finds itself, and a program is the state of that envir-
thought. onment. The cpu (or the virtual machine) is (let’s presume)
fixed in the same way that the laws of nature are fixed. But
It is surprisingly difficult to find a well considered defini- depending on the environment within which it finds itself
tion of computation. The definition offered in (Eliasmith, —depending on the program that it finds in its environ-
2006.C) appears to be the most carefully thought out. Here
is his definition and commentary. 11
Another question about alterability is that it’s not clear
what to do about hard-wired programs. How fixed must a program be be-
Computation: A series of rule governed state transitions fore one is no longer willing to say that it is alterable? We don’t address
whose rules can be altered. this issue.

Abbott – 10/17/2008 If a tree casts a shadow … 7/13

ment—the cpu operates differently, i.e., it performs a dif- environmentally driven. Computing involves configuring
ferent computation. environmental contingencies, i.e., setting up an environ-
ment within which a process (or multiple processes) will
But the same sort of analysis may be applied to virtually
occur. One may refer to this perspective on computing as
any natural process. When we put objects on a balance
non-algorithmic computing because one’s focus is more on
scale, for example, the behavior of the scale will depend on
how an environment will shape a process than on a specific
the particular objects, i.e., on the environmental contingen-
sequence of steps that the shaped process will take. No ex-
cies. In both the case of the program that we load into a
plicit algorithm is involved. Most of what we call uncon-
computer and the object that we put in the pans of a bal-
ventional computation appears to be non-algorithmic in
ance scale we (the user) determine the environment within
this sense.
which some fixed process (that defines the rules) proceeds.
Given this perspective, it may seem ironic that what we
This brings us back to our original position. A process in
think of as conventional computation (as carried out by von
nature may be considered a computation only when we use
Neumann computers) is a very constrained form of uncon-
it as a way to work with externalized thought.
ventional computation. We are attracted to this model be-
A physical or otherwise established process—be it the op- cause its single threaded linearity makes it easy to manage.
eration of a balance scale, of a cpu, of the Game of Life, or But clearly, nature is not linear. Any electrical/computer
of the sun in motion with respect to trees and the ground— engineer will confirm how much work it takes to shape
is just what it is, a fixed process. 12 But for almost all pro- what really goes on in nature into something as easy to
cesses,13 whether we create them or they arise naturally, manage as a von Neumann computer. Perhaps even more
how the process proceeds depends on environmental con- ironically, we then turn around and use conventional
tingencies, i.e., its history and its initial conditions. When single-threaded computations to simulate nonlinear uncon-
we control (or interpret) the contingencies in such a way ventional computation.
that we can then use the resulting process to work with our
One might say that one of the goals of this conference is to
own thoughts, then the process may be considered a com-
find ways to eliminate the von Neumann middle man.
putation. This is the case whether we control the contin-
gencies by loading a program into a computer, by placing Analog vs. digital computers
objects in the pans of a balance scale, by establishing ini- As (Bromley, 1990) points out, an analog computer is
tial conditions on a Game of Life grid, or by giving mean- known as such because it embodies an analog of nature—
ing to shadows cast by trees. really an analog of our ideas about nature. These analogs
Consequently we agree with Eliasmith that it must be pos- are usually simplified models that can be seen as analogous
sible to alter the rules, but we would express that condition to some aspect of (our ideas about) nature. A scale draw-
in other terms. For a physical or otherwise established pro- ing is a simple but illustrative example. By making obser-
cess to be considered a computation there must be vations about a model, one can project those observations
something contingent, i.e., alterable, about it. It must be back onto our ideas about nature and then onto nature it-
possible to separate a computational process into its fixed self. Our first computers—and in some sense all our com-
or established part and its contingent or alterable part. puters every since—have been (and continue to be) analog
computers in this sense.
The established process may be some concrete instances of
the playing out of the laws of nature—in which case the (Bromley, 1990) also points out that it was not long before
contingent environment is the context within which that we started to use our analog thought tools to represent not
playing out occurs. Or the established process may be the only our intuitive models of nature but to represent equa-
operation of a cpu—in which case the contingent environ- tions that characterize those models. This is the domain of
ment is the computer’s memory which contains the pro- the analog computer from the mid-19th century onward. We
gram that is being executed. Or the established process built thought tools that were capable of performing both
may be the operation of a program that a cpu is executing simple arithmetic as well as more complex mathematical
—in which case the contingent environment is the input to operations.
that program. No matter how it’s sliced up, a computation The distinction that is usually drawn between analog and
occurs when we alter (or interpret in the case of shadows) digital computers is that analog computers represent values
the contingencies in the environment of an established pro- as physical quantities, e.g., voltages angles or even lengths
cess as a way to work with our thoughts. when we think of a scale drawing as an analog thought
Non-algorithmic computing tool, whereas digital computers represent values as symbol-
ic encodings. The fundamental trick of a digital computer
A corollary of the preceding discussion is that all computa-
is that it restores its physical values to standardized levels
tion performed by established processes must be considered
—a bit is either on or off—thereby precluding the use of its
physical values to represent problem-specific values. Al-
12
Of course many processes—such as the operation or a though the analog/digital distinction is clearly important
cpu and the operation of a balance scale—are what they are because we
built them to be that way. And we built them to be that way because we when one considers implementation, it is much less im-
anticipated using contingencies that we could control in their environment portant when one asks about what thoughts can be extern-
to help us think about something. alized.
13
Some quantum processes may occur on their own and
without regard to their environments.

Abbott – 10/17/2008 If a tree casts a shadow … 8/13

Whether one is using an analog or a digital computer, the Neumann computers, that we built based on the Turing
objective is to create an externally manipulable model— Machine.
where a model is a representation of an idea we have about But many of our fundamental computer science models are
how something (or how an aspect of something) is organ- non-deterministic. Consider non-deterministic automata of
ized and how it works. various sorts, regular expressions, context free grammars,
An advantage that analog computers have over digital general rewrite systems, etc. Computation as we have de-
computers is that they operate in parallel. Once one has veloped it in theory is certainly not a linear, single
configured an analog computer to represent a set of interre- threaded concept.
lated equations, all one has to do is to apply voltages that Yet we have been unable to build thought tools to help us
represent the values of the independent variable and then externalize the kinds of non-determinism just listed. In
read off the values of the dependent variables. some cases (NFAs, regular expressions), non-determinism
But as far as thought tools are concerned, as long as a tool yield no additional computational power. In other cases
can represent and manipulate a model fast enough to serve (NP), non-determinism results in a combinatorial explo-
our needs, it shouldn’t matter whether the tool uses direct sion and a demand for computational resources that simply
(analog) or indirect (symbolic) mechanisms for represent- cannot be met by any physical thought tools.15
ing values. About four decades ago an intermediate form of computa-
Although analog computer have some advantages, as we tional framework began to emerge. It has been spreading
all know, digital computers have others—they are able to slowly and is becoming the primary computational
represent symbolic as well as numeric values, for ex- paradigm.
ample.14 Furthermore, digital computers are able to store It is a framework that corresponds very closely to how we
and represent far more data than analog computers. think about the world.
One might argue that since analog computers use actual
physical quantities for values, there is no limit to the preci- • The programming language Simula may reasonably
sion with which analog computers can calculate. That may lay claim to being the first object-oriented multi-
be the case, but it is also not relevant. For a computation to threaded programming language. Since an agent
be useful one must be able to extract the results. The may be understood most simply as an object (in the
primary value of thought tools is to enable us to work with object-oriented programming sense) that is con-
our ideas. But to be useful for that purpose, we must be trolled by its own thread, 16 Simula may be con-
able to bring the fruit of the work back into our minds. sidered the first agent-based language.
Thus a computation (analog or digital) whose results are • In the mid-1970s (Chen, 1976) developed the En-
beyond our ability to observe has no value to us and should tity-Relationship model for use in describing those
not be called a computation. aspects of the world about which information is to
On the other hand, if one is using an analog computer to stored in a database. The ER model adds to the no-
control some other device, there is no need to extract res- tion of entities (agents by another name) an ability
ults and bring them back into our minds. This is the sim- to express static relationships among them.
plicity and elegance of using an object to represent its own • At the beginning of the 1980s, the spreadsheet
dimensions. When preparing to cut a hole into which an brought agents and relationships to the masses. A
object must fit, trace the object instead of measuring it and spreadsheet is, in effect, an agent based system—
then drawing an outline of a hole with the measured di- really a limited form of cellular automaton—in
mensions. More interestingly, with a balance scale one can which each cell is a simple agent. Each cell con-
determine which of two objects is heavier without ever ex- strains itself to have a value that depends on other
pressing the weight of either as a number. cells.
When a balance scale compares the weights of two objects Agent-based programming, the ER model, and spread-
and returns an “output” (selected from left-is-heavier, sheets illustrate the development of thought tools that sup-
equal-weights, and right-is-heavier), is it performing a port the agent-based computing paradigm.
computation? It is if we are using it for this purpose. It is-
n’t if we are using it as a designer setting for flower pots. Agent-based computing is attractive from a resource per-
spective because it requires what we might call manageable
parallelism—which allows for the operation of asynchron-
5 Agent-based computing ous computing threads but doesn’t result in a combinatori-
Traditionally when one thinks of computing and comput-
ally unrealizable demand for computing resources.
ability, one thinks of Turing Machines and what they can
compute. The Turing Machine model is explicitly single In the most general case of agent-based computing, each
threaded. So are the thought tools, the single processor von entity has the power of a Turing Machine. The agents are
15
If we get it to work on a useful scale quantum comput-
ing may be the first such thought tool.
14
A decade ago (Brocket, 1997) apparently showed that
any finite automaton can be emulated by a set of continuous mathematical 16
In our view, an agent is an object (a) whose actions are
equations. It is thus possible to use analog computers for symbolic compu- a consequence of its own internal thread(s) and (b) whose methods may
tations. not be called freely from the outside.

Abbott – 10/17/2008 If a tree casts a shadow … 9/13

all situated in a common, although possibly complexly nomic system the rules and mechanisms are defined by
structured environment. Since the agents operate asyn- the legal system and its associated enforcement mech-
chronously and in parallel, each agent’s environment may anisms.
change as it operates. The agents are situated in the envir- (b) a collection of standards, which enables the parti-
onment in such a way that they are able to interact with cipants to interact in terms of a common vocabulary.
each other either directly or indirectly. Direct interaction is In the economic system these include the monetary
the sort of communication that takes place among elements system, standards for weights and measures, etc.
in a network. Indirect interaction has been called stigmer-
gic. It occurs when agents leave markers for each other in (c) an infrastructure of services. These are widely used
the environment. Computational paradigms that make use services that are not considered fundamental enough
of pheromone-like markers (e.g., as ant-colony optimiza- to be built into the basic framework. In the economic
tion) illustrate this form of computing. system these include the transportation and commu-
nication infrastructures.
Ubiquitous computing
Agent-based computing envisages agents situated in an en- Although non-trivial to implement and essential to the suc-
vironment. The agents are in the conceptual foreground; cess of the system of systems, these elements may be com-
the environment serves as background. If you turn that pic- paratively small compared to the system of systems as a
ture inside out and think of the environment as primary whole, i.e., the entire economic system.
and the agents as secondary one gets ubiquitous comput- Notice that these features describe not only systems of sys-
ing. tems but agent-based systems in general. In other words, a
Ubiquitous computing is distinguished from traditional system of systems is really just another agent-based system.
computing in that one imagines computational activity to In this case the agents are the participating systems and the
be disbursed into the environment instead of being local- system (of these systems) is the environment or framework
ized to a few relatively isolated computational sites. With within which they interact.
ubiquitous computing one imagines that processing threads Far-from-equilibrium (open) computing
may be anywhere—and that there be a great many of them. Wegner’s interactive computing agents
But still, each processing thread is an independent comput-
ing activity—just as in the agent-based model—and there (Wegner, 2005) (and elsewhere) perform their computa-
is no combinatorial explosion. tions over an indefinite period—continually accepting in-
put and producing output without ever completing what
Systems of systems might be understood as a traditional algorithmic computa-
The term system of systems has been in use in the systems tion. Although Wegner’s focus is somewhat different, his
engineering community for at least the past decade, see interactive computing model appears to be more or less the
(Maier, 1996). Unfortunately, the most widely accepted same as the standard agent-based model. Wegner’s inter-
sense of the term is that a system of systems is nothing active agents are agents.
more than a very large system. Jamshidi, program chair of
the IEEE Systems of Systems Engineering conference, But Wegner is after far more than simply to support agent-
quotes as his “favorite” definition of system of systems based. Wegner and Goldin “claim that interactive finite
from (Kotov, 1997), which reads, computing agents are more expressive than Turing ma-
Systems of systems are large scale concurrent and distributed chines.” (Wegner, 1999) Since agent-based frameworks
systems that are comprised of complex systems. are implemented using standard computers, and since
standard computers are considered no more powerful than
In fact, a system of systems is not just a large scale concur- Turing machines, it appears on the face of it that no formal
rent and distributed system that has complex systems as power is gained from an agent-based (or interactive)
subsystems. A system of systems is not a system in the tra- framework. On what grounds do Wegner and Goldin claim
ditional hierarchical systems engineering sense at all. In to have more than traditional computing?
our opinion, a system of systems is a framework or envir-
onment within which independently operating systems op- In focusing on interaction, Wegner and Goldin do for Tur-
erate and communicate. (Abbott, 2006) ing machine computations what one does for the Prisoner’s
Dilemma by converting it from a one-shot event into an it-
A familiar example of a true system of systems is the eco- erative process. They convert it from a closed into an open
nomic system. An economic system is not a system for system. By definition, interaction involves an external en-
which one defines requirements, which are then allocated vironment. As Wegner points out, Turing invented this
to subsystems, which are then designed and built. The eco- model of computing as well as the traditional computation-
nomic system is an open-ended framework that can accom- al model. Turing called it an oracle machine—although
modate the comings and goings of participating members, Turing’s oracles weren’t thought of either as users or as the
which are not designed into the framework from the start environment.17 Nonetheless an interactive system is an or-
as is the case with traditional engineered systems. The
framework of a system of systems requires the following.
(a) rules along with mechanisms for implementing the
17
Turing distinguished among oracle machines on the
basis of the number of times they interacted with the oracle. Wegner’s in-
rules, which define how the framework works and teraction machines interact with an external environment continually and
what options are open to the participants. In the eco- without limit. For our purposes Wegner’s model makes more sense.

Abbott – 10/17/2008 If a tree casts a shadow … 10/13

acle machine, which is more powerful than a standard Tur- build (mental) models of their surroundings and then
ing machine.18 (b) to run an evolutionary computation on those models
as a way to make decisions about how to act. We are
The distinction between a Turing machine and an oracle
nowhere near being able to build agents with that level
machine parallels that between a closed and an open sys-
of sophistication. But it is not out of the question.
tem in physics. Turing machines are closed. Given a set of
rules and some fixed input, a Turing machine does its Another way to think about this is that science is a fixed-
thing in isolation. It may make use of unlimited memory for-the-moment framework for thinking about nature. But
resources, but it has no access to any information other it has never been a completely successful framework. It
than the information it had at the start of the computation. tests itself against nature by asking questions of nature as
an oracle. In answering, nature provides information that
Science as an open computing system was missing from the framework at any given time. With
In contrast to a traditional computing system, an oracle
the new information science is able to construct a new
machine is open. It interacts with an external environ-
framework that incorporates that new information.
ment.19 Besides oracular pronouncements, what might one
gain from having an open computing system? This is easi- What does it mean for a system to be ffe wrt information?
est to see by looking at how science progresses. Science It means that the system adds new processing capabilities?
does two things. Just being able to compute more functions doesn’t seem to
add much. After all, a Universal TM can compute any
1. Through observation and experimentation scientists
computable function simply by enumerating them and sim-
look at nature and observe new facts, e.g., that light’s
ulating them. Does it mean being able to do more in the
velocity is constant, that a photon may “interfere”
real world? Presumably since knowing more about the real
with itself, etc. In other words, we learn more and
world enables one to do more with what one now knows.
more about the threads that weave nature’s tapestry.
So having more information is valuable only wrt to a larger
Could a (very smart) agent-based system discover new environment.
facts? To discover new facts about nature requires ac-
A different subject. Thoughts are pure; reality is messy.
cess to nature. Facts cannot be discovered in isolation—
That’s why we like reality to be as much like thought as
no matter how smart one is. That’s why we make obser-
possible. By messy I mean having hair follicles and
vations and do experiments. Thus by an open system,
pimples, etc., not just complicated or complex—although
we mean a system that has nature (or some source of in-
that is a version of it. But we can make anything pure by
formation outside itself) as an oracle.
conceptualizing it. Once we conceptualize hair follicles
2. Scientists create new ways to think about nature that and include them as part of our concept of skin (or
incorporates both the new facts as well as what was whatever), they become pure. We want reality to match our
already known. We figure out how the threads of the thoughts—or vice versa. So perhaps it isn’t a different sub-
tapestry come together to paint a grand picture.20 ject after all.
Could a (very smart) agent-based system come up with Far-from-equilibrium with entities means that the entity
new ways to think about nature? We know of only one has access to energy that enables it to retain is reduced en-
mechanism to come up with new things: evolution. tropy state. What does it do in the information domain?
(Dennett, 1996) argues that the evolutionary process is How can information help an entity to persist—other than
the only source of creativity, be it evolutionary creativity by allowing it to get more energy? Is that what evolution
or human creativity. is? Increasingly information rich entities that are increas-
We know how to build both agent-based and evolution- ingly better at extracting energy from the environment for
ary systems. But beside Tierra (Ray, 1991) and work in- the sake of persisting? In that picture, what is it that is ffe?
spired by it, there has been little effort to merge the two It is the evolutionary process itself, i.e., the structure that is
into an integrated system. This is understandable. Oth- capable of extracting information from the environment,
er than a purely evolutionary model, a successful mer- which is DNA in biology and science in our version. Can
ger would require an agent-based system in which the one say that evolution is far from equilibrium? So in this
agents were computationally sophisticated enough (a) to picture, information is a second-level good. Is there a third
level good, something that will enable an entity or process
18
Neither Turing nor Wegner specifically refer to an or-
to extract information from the environment more effect-
acle machine as an open computing device. ively?
19
Most complex systems are open—to energy—which is Information transforms an entity into another form of en-
how they persist far from equilibrium. Complex system exist far from en- tity. Is there anything that can transform evolution into an-
ergy equilibrium and far from information equilibrium. A hurricane is an
example of a far-from-equilibrium energy system that is not far-from equi- other form of evolution? If it parallels evolution, the new
librium with respect to information. For a system to be far-from-equilibri- form of evolution would have to be able to extract informa-
um wrt information it must also be ffe wrt energy since it requires energy tion from the environment more effectively. How do we ex-
to provide power for its computational processes. tract information from the environment? By asking ques-
20
This second step represents scientific creativity. We are tions and then changing how we think. Is there anything
creative in non-scientific ways as well, most notably in the many ways we comparable with evolution? Is mental evolution the next
have found to externalize and manipulate our thoughts. For the purposes
of our example, we use scientific creativity as a proxy for creativity in gen-
eral.

Abbott – 10/17/2008 If a tree casts a shadow … 11/13

step after physical evolution? Is there a step after that? Cul- Chen, P, “The Entity Relationship Model - Toward a Uni-
ture/science? Then automated evolution? fied View of Data,” ACM Transactions Database Systems,
1 (1), 1976.
But all these are just better ways of extracting information.
What is the higher level resource that is enabling that? Ab- Dennett, D., Darwin’s Dangerous Idea, Simon & Schuster;
straction? Abstraction/conceptualization is the level above Reprint edition, (June 12, 1996).
information. Eliasmith, C, “Computation,” entry in Dictionary of the
Philosophy of Mind, May 11, 2004. Accessed March 6,
6 Conclusion 2006: http://philosophy.uwaterloo.ca/MindDict/computa-
The Church-Turing thesis is that the lambda calculus, Tur- tion.html.
ing machines, and Post production systems, all of which
Eliasmith, C, “Philosophy of mind,” entry in Dictionary of
characterize the same set of computable functions, charac-
the Philosophy of Mind, May 11, 2004. Accessed March
terize what we think of as computable.
25, 2006: http://philosophy.uwaterloo.ca/MindDict/philo-
The agent-based paradigm involves (a) autonomous but in- sophyofmind.html.
terrelated and interconnected agents, each of which has the
Google, Help Center, undated. Accessed March 12, 2006:
computing capacity of a Turing machine, situated in (b) a
http://mail.google.com/support/bin/answer.py?answer=294
structured, persistent, but evolving environment that (c) is
33&query=faq&topic=0&type=f.
not contained entirely within the system itself and that re-
veals itself to the agents only over time.21 IJUC Editorial Board, “Preface [to the first edition] from
the Editorial Board,” International Journal of Unconven-
The agent-based thesis is that the agent-based paradigm
tional Computing, (1, 1), pp 1-2, 2004. Accessed Feb 19:
represents a consensus view of how, at the start of the 21 st
http://www.oldcitypublishing.com/IJUC/IJUC1.1fulltext/E
century, we tend to think about much of the world.
ditorial.pdf.
If computation is the use of thought tools to help us work
Jamshidi, M. “System-of-Systems Engineering - a Defini-
with externalized thought, and if we tend to think in agent-
tion,” IEEE SMC (October 2005). Available as of Jan 2006
based terms, tools based on the agent-based paradigm en-
at: http://ieeesmc2005.unm.edu/SoSE_Defn.htm.
able us to work with these ideas. In developing such tools
we continue to explore the range of possible thought tools Kotov, V., "Systems of Systems as Communicating Struc-
we might build for ourselves.. tures," Hewlett Packard Computer Systems Laboratory Pa-
per HPL-97-124, (1997).
References Maier, M. “Architecting Principles for Systems of Sys-
Abbott, R., “Emergence Explained,” submitted for publica- tems,” Proc. of the Sixth Annual International Symposium,
tion. Accessed February 19, 2006: International Council on Systems Engineering, Boston,
http://cs.calstatela.edu/~wiki/images/9/90/Emergence_Ex- MA, (1996), pp. 567- 574. Accessed Jan 2006:
plained.pdf. http://www.infoed.com/Open/PAPERS/systems.htm.
Abbott, R. “Open at the Top; Open at the Bottom; and Maxfield and Montrose Interactive, A History of Com-
Continually (but Slowly) Evolving,” in Proceedings of the puters. Accessed February 20, 2006: http://www.maxmon.-
IEEE Conference on Systems of Systems Engineering, com/history.htm.
April, 2006. Ockham, William of (1323) Summa Logicae, Johannes
Albacete, L. “For the Love of God,” New York Times, Feb Higman, Paris, 1488; the edition owned by C. S. Peirce.
3, 2006. Available (for a fee) at: Also volume 1 of Opera Philosophica, ed. by P. Boehner,
http://select.nytimes.com/gst/abstract.html?res=F00E16FC G. Gál, & S. Brown, Franciscan Institute, St. Bonaventure,
3D5A0C708CDDAB0894DE404482. NY, 1974.
Brocket, R., “A Rational Flow for the Toda Lattice OMG, Unified Modeling Language, v 2.0: Superstructure,
Equations,” in Operators, Systems, and Linear Algebra August 2005. Accessed March 13, 2006:
(U. Helmke et al. eds.), B.G. Teubner, Stuggart, 1997. See http://www.omg.org/docs/formal/05-07-04.pdf.
web page for a brief discussion. Accessed February 19, Ray, T. S. “An approach to the synthesis of life,” in :
2006: http://hrl.harvard.edu/analog/. Langton, C., C. Taylor, J. D. Farmer, & S. Rasmussen
Bromley, A. “Analog Computing Devices,” in W. Asprey [eds], Artificial Life II, Santa Fe Institute Studies in the
(ed), Computing Before Computers, Iowa State University Sciences of Complexity, vol. XI, 371-408. Addison-Wesley,
Press, Ames, Iowa, 1990. Accessed Feb 27, 2006: 1991.
http://ed-thelen.org/comp-hist/CBC.html. Smart Computing, undated. Accessed February 20, 2006:
http://www.smartcomputing.com/editorial/dictionary/de-
tail.asp?guid=&searchtype=1&DicID=18707&RefType=E
ncyclopedia.
21
From the perspective of our shared real-life experience,
this third condition corresponds to knowledge gained about nature from
Sowa, J., “Existential Graphs,” 2002. Accessed February
observation and scientific experimentation. From the perspective of com- 20, 2006: http://www.jfsowa.com/peirce/ms514.htm.
putability, it corresponds to oracular pronouncements.

Abbott – 10/17/2008 If a tree casts a shadow … 12/13

Sowa, J. “Semantic Networks,” in S.C.. Shapiro (ed) Arti- W3C, Extensible Markup Language (XML), February 5,
ficial Intelligence-Encyclopedias, John Willy & Sons, Inc, 2006. Accessed March 25, 2006:
1992, pp 1493-1511. Accessed February 20, 2006: http://www.w3.org/XML/.
http://www.jfsowa.com/pubs/semnet.htm. W3C, Semantic Web, March 22, 2006. Accessed March
Thagard, Paul, "Cognitive Science", The Stanford Encyc- 25, 2006: http://www.w3.org/2001/sw/.
lopedia of Philosophy (Winter 2004 Edition), Edward N. Wegner, P. and D. Goldin, “Interaction, Computability,
Zalta (ed.), Accessed March 25, 2006: http://plato.stan- and Church’s Thesis,” British Computing Journal, 1999.
ford.edu/archives/win2004/entries/cognitive-science. Accessed March 20, 2006:
W3C, OWL Web Ontology Language Reference, February http://www.cs.brown.edu/people/pw/papers/bcj1.pdf.
10, 2004. Accessed March 13, 2006: Wegner, P. et. al., “The Role of Agent Interaction in Mod-
http://www.w3.org/TR/owl-ref/. els of Computing,” Electronic Notes in Theoretical Com-
puter Science, 141, 2005. Accessed March 19, 2006:
http://www.ecs.soton.ac.uk/~mml/papers/finco05.pdf.

Abbott – 10/17/2008 If a tree casts a shadow … 13/13