The critical role of personality for creating an artificial mind

Material Information

The critical role of personality for creating an artificial mind a model for self-evolution based on dynamic experience
Portion of title:
Model for self-evolution based on dynamic experience
Prebble, Carrie Todd
Publication Date:
Physical Description:
ix, 86 leaves : illustrations ; 29 cm


Subjects / Keywords:
Artificial intelligence ( lcsh )
Cognitive learning theory ( lcsh )
Cognitive science ( lcsh )
Artificial intelligence ( fast )
Cognitive learning theory ( fast )
Cognitive science ( fast )
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )


Includes bibliographical references (leaves 71-86).
General Note:
Submitted in partial fulfillment of the requirements for the degree, Master of Science, Computer Science.
General Note:
Department of Computer Science and Engineering
Statement of Responsibility:
by Carrie Todd Prebble.

Record Information

Source Institution:
|University of Colorado Denver
Holding Location:
|Auraria Library
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
37306931 ( OCLC )
LD1190.E52 1996m .P74 ( lcc )

Full Text
Carrie Todd Prebble
B.A., University of Colorado at Boulder, 1984
A thesis submitted to the
University of Colorado at Denver
in partial fulfillment
of the requirements for the degree of
Master of Science
Computer Science

1996 by Carrie Todd Prebble
All rights reserved.

This thesis for the Master of Science
degree by
Carrie Todd Prebble
has been approved
Jody Paul

Prebble, Carrie Todd (M.S., Computer Science)
The Critical Role of Personality for Creating an Artificial Mind:
A Model for Self-Evolution Based on Dynamic Experience
Thesis directed by Assistant Professor Jody Paul
Personality is a dynamic process that evolves with
experience. Personality accounts for the properties of intuition,
insight, drive, commitment, choice, etc. These are integral
properties for a truly intelligent mind natural and artificial.
Personality is essential for an artificial mind
The model presented in this thesis for building an
artificial mind with personality is based on the theory of
Dynamic Memory. Dynamic Memory theory is a computer
memory theory that models human memory. It interprets new
experiences based on the experiences already in memory. The
most similar experience is found and is linked to the new
experience. Finding the most similar experience is
understanding. The structure of the memory changes with
each new experience a new experience is understood and the
state of the memory changes.
An experience is represented as a collection of specific
information surrounding that experience. This information
might include details about the situation, the participants, the
objects involved, the intent, the results, etc. This collection of
information is called a case. Each specific piece of information
is called an attribute.

I claim that finding a similar experience in memory
understanding also depends on what parts of an experience
are more important to remember at the time: the current
context. Context can be incorporated into the Dynamic Memory
model by differential weighing of various attributes of a case.
For instance, in a scavenger hunt, shape and color may become
more important than size and texture, or vice versa.
So, this memory structure changes with each new input.
Links are created between similar cases. Similarity is
determined by what is already there and the context in which
this new case is perceived. A tangled web of links forms. The
formation of the tangled web is dependent on the order that
the cases came in to memory the order of experiences, and
the context. This tangled web of connections is unique from all
other tangled webs formed by any different input sequence.
Memory's tangled structure interprets and understands new
experiences in ways whose behavior suggests intuition, insight,
choice, etc. That is, the unique structure of memory embodies
This abstract accurately represents the content of the
candidates thesis. I recommend its public^tio

Figures .............................................. ix
1. INTRODUCTION .................................... 1
Understanding ................................... 1
Chinese Room ............................... 2
Degrees of Understanding ................... 3
Memory .......................................... 4
Personality ..................................... 5
Why Have a Personality in a Computer? ...... 6
Personality Versus Computation ............. 8
Summary ....................................... 9
The Following Chapters ......................... 10
Personality, Understanding and Memory in the
Human Mind ..................................... 12
Six Theories of Personality ............... 12
Which Brings Us to Memory ................. 17
Learning ............................ 18
Retrieval and Forgetting ............ 21

Personality and Memory Studied
Together .................................. 23
Personality, Memory and Understanding in
Computers ....................................... 25
Personality Studied in Computers .......... 25
Computer Memory and Understanding ......... 26
3. APPROACH AND METHOD ............................. 30
Case-Based Reasoning ............................ 30
Definition of a Case ...................... 31
The Range of Case-Based Problem
Solving ................................... 32
Dynamic Memory .................................. 32
Example of Inference from Limited
Information ....................,......... 33
Reminding and Understanding ............... 34
Memory Structure .......................... 35
Case-Based Learning ....................... 35
Context ......................................... 36
Permanance in Memory and Personality .. 38
DMP Memory ...................................... 39
Cases Represent Objects ................... 39
Classifying Objects ....................... 39
Cases ........................................... 42
Case Example .............................. 43

Generalizations ................................. 44
Examples of Generalizations ............... 44
Generalization Example 1............. 44
Generalization Example 2............. 44
Runs of DMP ..................................... 45
Transcript of Run #1 ...................... 48
Transcript of Run #2 ...................... 49
Transcript of Run #3 ...................... 52
Transcript of Run #4 ...................... 55
Summary of Runs ........................... 56
5. CONCLUSIONS .................................... 57
A. Scheme Source Code ................................ 60
ANNOTATED BIBLIOGRAPHY .................................... 71

4.1. State of memory before input, Run #1...... 46
4.2. State of memory after input, Run #1....... 46
4.3. State of memory before input, Run #2...... 47
4.4. State of memory after input, Run #2....... 47
4.5. State of memory before input, Run #3..... 51
4.6. State of memory after input, Run #3....... 51
4.7. State of memory before input, Run #4...... 53
4.8. State of memory after input, Run #4....... 54

My thesis is about computer-based understanding, memory and personality.
To propose an explanation for the mechanisms responsible for personality, my
work builds on the theory of dynamic memory, that understanding and memory are
the same thing. I will show why personality is also a necessary component for the
creation of an artificial mind.
An example of understanding is when we come to grasp a new concept like
what a book is or what a chair is. We understand what they look like and that they
dont all look alike. We understand what they feel like and how heavy they are and
what their function is. And a chair can be many things: a rocker, a sofa, an
adirondack, a stump, a countertop when Mom is cooking. We have an
understanding of the concept chair. Similarly, we understand concepts about
traveling and studying and about conflicts and relationships and desire.

There are degrees of understanding (Franklin 1995) which include
intermediate states that correspond to saying: I sort-of understand what you
mean. An artificial mind, if it is to be anything like ours, must understand to
some degree.
Some researchers dont think computers will ever understand anything
(Dreyfus 1992, Searle in Boden 1990, Penrose 1989). Dreyfus (1992) feels that
we can never mechanize human intelligence, which includes our intuition, our
hunches and emotional responses that influence a judgement. More on this later.
Chinese Room
Searle (Boden 1990) uses a thought experiment called the Chinese Room to
illustrate understanding. Say Searle is in a closed room. A story and a question
about the story, both written in Chinese, are slipped in under the door to him.
Searle doesnt speak or read Chinese (its Greek to him). Hes been given
comprehensive instructions, written in English, that provide an algorithmic way of
answering the question. Searle follows the directions and produces squiggles on
the page that form his answer to the Chinese question. Searle has answered the
question that was written in Chinese in Chinese. He slips it under the door.
Chinese speakers outside the room could be led to believe that the guy inside the
room understands Chinese because of his answer to the question. Searle doesnt
think hes understood anything about the story or the question. Searle asserts that
the guy inside the box can spit out an answer but understands nothing. Franklin

(1995) responds by saying that Searle by himself doesnt understand Chinese, but
that if you include the instructions and Searle together as a system, then the system
does have a degree of understanding. If Searle were to memorize the instructions
then in some fashion he would understand Chinese even if it were only how to
scrawl out characters. Franklin says this and Searle would disagree about any
degree of understanding.
Degrees of Understanding
Degrees of understanding can be accounted for by the connections to
different kinds of knowledge. I dont understand numerical analysis as well as my
teacher. He has a much deeper understanding of how numerical analysis is used as
a tool. I probably have a little better understanding of it than someone whos never
had the course. How do we convince ourselves that someone understands what
we are saying? We usually ask questions (in the form of exams in my class for
instance). We judge the degree of understanding based on the answers the
understander gives. Amie, a computer program that answers quadratic equation
questions, illustrates degrees of understanding (Franklin 1995). Amie, when given
the coefficients a, b and c, solves the equation ax2 + bx + c = 0. Amie has a degree
of understanding of numbers and algebra and quadratic equations, certainly more
than a program that doesnt give the right answer.
Its the connections to other knowledge that give us understanding. [T]he
more connections, the more understanding.(Franklin 1995 pl08) Amie has

some connections between how a variable a might equal five and that five is this
many and how a variable fits into an equation and so forth. Amie has nonzero
understanding. A system with more understanding can be reduced to Amies level
connection by connection. (Franklin 1995 pl08) So for an artificial mind to have
lots of understanding it will have to have lots of connections among its knowledge
of concepts. The more connections there are from a concept to different kinds of
knowledge, the better the understanding of that concept.
The memory part of my research is, naturally, connected to the
understanding part. The memory is where all the connections are. Now I dont
mean memory as in RAM and ROM and hard drives, I mean memory like humans
have. I have memory for the concept of chair because Ive encountered many
chairs in many situations. Understanding means being reminded of the closest
previously experienced phenomenon.(Schank 1982 p24) Some of my closest
previously experienced phenomenon to chair is that: Ive used a log as a chair
when I go camping, and Ive sat on the counter while Mom made cookies. The
concept I have for chair revolves around sitting; the putting of my butt on
something that is not the floor. [W]hen we look up a word in our mental
dictionaries, we find an entire episode from our experience located with the
definition, almost as if it were a part of that definition.(Schank 1982 p25) Chair
and sitting and lots of instances of sitting on chairs are in my memory connected to

each other and other things like: Did I get a sliver when I sat on the log?, or Was
there a puddle on the deck chair after the rain?, or How hard is it to get up from a
chair when Ive been skiing all day? I have lots of connections in my memory to
the concept of chair. Similarly, for this computer to have some understanding of
chair, it needs to have a concept of chair in its memory and connections with other
Where does the personality part come in? you ask. Personality, I claim,
is embodied in the network of connections in memory. Every human has a
different personality because each has a different network of connections that
comprise understanding. (Schank 1982 p226) Memories should only be identical
for people with identical experiences, and it is not even clear if they should be
identical then.(Schank 1982 p224)
My network continues to be built over the course of my life. The building
process is what we generally call learning. Connections are made depending on
the current learning situation: the context (more on this later). And what Ive
already learned, what is already in my memory, determines how I interpret or
understand new experiences. How we understand is affected by what is in our
memory.(Schank 1982 p4) My. sister and I have had a similar growing up
situation. Yet our personalities, our networks of connections, are very different. I
learned a connection between school and fun whereas my sister made a connection

between fun and sun. Now I have a connection between fun and sun too, but even
without looking inside our heads anyone who knows us can tell that we have
different connections. One such difference is because we have a different collection
of concepts connected to the fun concept. Thats part of our individual
Whv Have a Personality in a Computer?
A personality, an individualness, is necessary for a mind natural or
artificial. Emotions, motivation, beliefs, coping, needs, desires are the things of
personality. Truly intelligent machines require emotions. Our needs, desires, and
emotions provide us directly with a sense of the appropriateness of our
behavior.(Dreyfus 1992 pxlv) If we are to build truly intelligent systems, we must
allow for multiple sources of motivation. Further, we must expect gaps and errors
to occur in the systems knowledge (beliefs about the world). (Franklin 1995 p387)
Descartes was a pioneer of the idea that soul stuff helps govern the physical
world.(Wright 1995) Personality helps govern the world. Personality is the
more than the sum of its parts part. Personality is the complexity gone chaotic
piece that needs to happen to a collection of circuits and logic gates to invoke
intuition and hunches and common sense. It is the stuff of human intelligence that
Dreyfus (1992) claims is non-computable.
Personality is the mechanism of focus. Personality is the channel through
which all the power of the mind the intellect, the creativity, the desire is

directed. Personality is how we choose to do something or not do anything.
Personality is like the steering wheel and gas pedal of a race car. By itself the car
has potential. Only when the pedal is pressed will the car do anything. Without
personality a mind has potential but unless that potential is channeled and directed
the mind doesnt do anything. Human nature includes "the planning, organizing,
building, imagining, dreaming, creating activity without which we cannot
understand either individuals or the civilization they have constructed.(Zucker et al
1992 p20) We cant understand individuals without the personality. Einstein, say,
had an interest in physics. He was driven to explain physical phenomenon to
himself and others. He set up his life in such a manner as to promote thinking
about and exploring his chosen field. This interest, this drive, this focus, and his
theories are due to the personality of Einstein. "[I]ndividuals will choose to spend
time in situations that offer them the chance to express their personalities."(Gale
1992 pl23)
Personality is not a goal, it is the creator of goals. Personality determines
what each of us does with whatever we have intellect, creativity, drive,
persistence, intuition, etc. Personality is what does the choosing. Our personality
helps us decide what is important.
One personality is responsible for the works of Mozart. One personality is
behind the works of Hemingway. One personality invented calculus. Someone
could argue that it wasnt really just one mind, after all, there were teachers and co-
workers who helped bring about these works. But, the synthesis of an idea
converges in one mind, even if others are there to help flesh it out.

Personality Versus Computation
We have to face the important way desires, emotions, and a persons
interpretations of what it means to be a human being open up endless possibilities
for human life.(Dreyfus 1992 p45) Dreyfus answers his own doubts about
mechanizing human intelligence. Dreams, desires and emotions the stuff of
human intelligence that eludes algorithmic definition, personality in other words
opens up the endless possibilities. There is also imagination and intuition and
doubt that contribute to mind. Whats the algorithm for computing intuition and
insight? We don't know these algorithms. We only know that we have them and
that they contribute to our intelligence.
Making choices comes from personality. Dreyfus (1992) criticizes
Winograd: he gives no formal account of how a computer program could
determine the current context. He criticizes Schank: Schanks proposal leaves
completely unanswered the problem of how do we choose the right script.
Dreyfus is pointing out the lack of an algorithm that could compute the proper
context or script. We humans choose context and scripts every day. We can
usually figure out which context or script is appropriate. We have interests and
goals and we ask ourselves what does this mean for me? and make our choices.
Sometimes we are wrong, the script we have chosen is inappropriate for this
situation. But we have chosen for ourselves. Could we choose these things for
other people? Could we choose them for another mind? It is this sort of algorithm
that Dreyfus wants: a choose-for-another-mind algorithm that we can plug in and it

would choose a context and have insight into the composition of woven cloth
because it has an interest in cloth, when in fact, that mind needs to do its own
choosing. It needs to create its own algorithm. It needs to evolve itself.
"Mathematically, we can describe two things interacting, like two planets in
space. Three things interacting three planets in space well, that becomes a
problem. Four or five things interacting, we can't really do it." (Crichton 1995
p310) Life is full of these complex interactions that we can't compute. Desires,
motivation, emotions personality can't really be computed mathematically.
We should look at it as something that emerges. It is an algorithm that emerges as
opposed to an algorithm we can program in.
Some people think the answer is that living forms organize themselves.
Life creates its own order, the way crystallization creates order. Some people think
life crystallizes into being, and thats how the complexity is managed.(Crichton
1995 p310) The answer could be in creating an artificial mind with enough
complexity such that there will emerge a more than the sum of its parts behavior.
Maybe theres a characteristic order to living things that is generated by their
interacting elements. Maybe life just happens.(Crichton 1995 p310) Maybe
intelligence just happens. Maybe intelligence creates its own algorithms.
A dream of artificial intelligence is to create an artificial mind (Bock 1993,
Minsky 1982, Franklin 1995). A mind that will think, dream and reason similarly

enough to us so that we could chat. A set of values, interpretations, interests, goals
and needs a personality would make that mind interesting. One might still
hope that networks different from our brains will make exciting new generalizations
and add to our intelligence.(Dreyfus 1992 pxxxvii)
As individuals, we humans interpret the world in a way that makes sense to
us (Franklin 1995). We each have our own model of the world inside our minds.
Our mind senses the world and creates information for its own use. Franklin
suggests that the task of mind is to decide what to do next. Our individual minds
decide what to do next based on our model of the world. Our power, our
creativity, our accomplishments may all come out of what I have decided to do
next. Individual minds have discovered mathematics, invented calculus, reasoned
relativity, built civilizations. Until computers can simulate human curiousity and
insight, robots will remain tools designed to discover what we already expect to
find.(de Grasse-Tyson 1995 p22) A mind that is capable of creativity, discovery
and invention, will need to be an individual with its own interests, its own driving
needs and goals, its own way of interpreting and modeling the world, that is its
own personality.
The Following Chapters
My study of personality in computers has led me to study memory
connections. I begin by using Schanks theory of dynamic memory to build a case-
based memory. My program, which I call Dynamic Memory with Personality or

DMP, develops a personality based on its experiences which it uses to interpret new
experiences. Cases, input into memory that are similar to each other are
generalized. Which cases get generalized is dependent upon the current memory
structure which in turn depends on the order they come into memory. The
collection of connections in the memory is personality. Each different input
sequence that results in a different memory structure is therefore responsible for
generating a different personality.
In chapter two I discuss the current studies of personality and memory in
humans and computers. In chapter three I discuss the theory of dynamic memory
and the methods I used to build DMP. In chapter four I illustrate my findings and
chapter five contains my conclusions and a vision of the future. The bibliography
is annotated.

Personality has led to the discoveries of calculus and the Odes of
Shakespeare. Just what is this personality thing? I went and looked it up in the
library. Theres been lots of studies done on human personality. Ill discuss some
of what I found in this chapter. Ill start with theories of personality, then Ill
discuss what I found about human memory and understanding. From there, Ill
talk about the literature on personality, memory and understanding in the computer.
Personality. Understanding and Memory in the Human Mind
Six Theories of Personality
To psychologists, personality is ones relatively distinctive and consistent
pattern of thinking, feeling, and acting.(Myer 1992 p440)
The psychoanalytic perspective is the work of Freud. The larger part of our
mind, the unconscious, is hidden from us; thoughts, wishes, feelings, and

memories of which we are largely unaware. For Freud, human personality its
emotions, strivings, and ideas arises from a conflict between our aggressive,
pleasure-seeking biological impulses and social restraints against them. In his
view, personality results from our efforts to resolve this basic conflict to express
these impulses in ways that bring satisfaction without also bringing guilt or
punishment.(Myer 1992 p413)
The trait perspective explains personality in terms of the dynamics that
underlie behavior. Gordon Allport defined personality in terms of identifiable
behavior patterns. He was less concerned with explaining behavior and more
concerned with describing individual traits. Trait or factor analysis identifies
clusters of behaviors that tend to occur together, and that this cluster reflects a basic
trait or factor. Hans and Sybil Eysenck believe that individual variations can be
reduced to two genetically influenced dimensions: extraversion/introversion and
emotional stability/instability.(Myer 1992) The Big Five personality factors
describes an expanded set of these trait factors. These five factors include
agreeableness, conscientiousness and openness to experience in addition to
Eysencks two. [F]or now, it seems our best approximation of the basic trait
dimensions.(Myer 1992 p424)
An objection to the trait constmct is that the behavioral consistency of
individuals across different situations is not accommodated by the trait and factor
theorists.(Eysenck 1977) Behavior can plausibly be viewed as a joint function of
situational determinants and individual characteristics, and trait theorists have
frequently been criticized for their relative lack of interest in situational
factors.(Eysenck 1977 pl93) In comparing groups with a different trait, it is

difficult to decide whether observed differences in behavior between the groups is
attributable to differences in that trait or to some other difference between the
groups. These objections center around the idea that the study of personality is
more than a sum of parts. It is a study of a whole person.
The humanistic perspective focuses on the strivings of healthy people for
self-determination and self-realization. The humanistic approach studies personal
experiences of sorrow and joy, alienation and intimacy, frustration and fulfillment.
Abraham Maslow proposed that we are motivated by a hierarchy of needs.(Myer
1992 p428) We seek food, shelter then safety, then love, and ultimately, we seek
self-actualization, the process of fulfilling our potential. For Maslow and other
humanists, a central feature of personality is ones self-concept. Maslow studied
self-actualization in the rich and productive lives of people among whom were
Abraham Lincoln, Thomas Jefferson, and Eleanor Roosevelt. He found common
characteristics. Among these characteristics were a self-awareness, an acceptance
of self, an openness and a spontaneity. These self-actualized people were loving
and caring and not paralyzed by others opinions.
The social-cognitive perspective derives from psychological principles of
learning, cognition and social behavior. Social-cognitive proponents emphasize the
importance of external events. They assume that our environment shapes our self-
understanding. How we think and feel about the situations we find ourselves in
affects our behavior. (Myer 1992 p435) Social-cognitive proponents ask
questions about how our memories and expectations influence our behavior
patterns. This interactive, reciprocal perspective has inspired researchers to study
how the environment shapes personal factors, such as self-control and self-concept,

and how these in turn influence behavior.(Myer 1992 p435) This differs from the
humanistic viewpoint in that theorists do not believe that ones self is innate,
waiting to be discovered, but rather they believe that our environment shapes our
self-understanding. Behavior, internal personal factors, and environmental
influences all operate as interlocking determinants of each other.(Myer 1992 p435)
Another perspective on personality was explored by Henry Murray in his
book Explorations in Personality (1938) which he calls personology. Living
beings must be studied as living wholes. Personality is a dynamic process, a
constantly changing configuration of thoughts, feelings and actions.(Zucker, et al.
1992 p8) It must be studied over periods of time. Personality is seen as
something happening rather than something fixed. (Zucker, et al. 1992 p8) Needs
are the precipitating dynamic that drives the process. Personality is a complex
organization and cannot be properly studied by shortcuts. (Zucker, et al. 1992
p20) Murray teaches us to take people as wholes, as they are and as they think they
are. People have needs, they are engaged in strivings, they are trying to produce
endstates and reach goals; it is a dynamic process. Any single episode of human
behavior must be understood in its developmental context.(Zucker, et al. 1992
p324) It is from a subjective standpoint [, how does a person define himself or
herself as a person,] that the person can be said to be a historya subjectively
composed and construed life story that integrates ones past, present and
future.(McAdams in Baron 1982 p325)
Cognitive-perceptual theory is a theory of personality that places
autobiographical memory at its center.(Singer 1993 pi69) A.R. Bruhn thinks that
a persons memories are fantasies about the past that reflect present concerns

(Singer 1993 pi69). Individuals remember what is useful to them and what helps
them adapt to their situations. Memory is a dynamic process where [i]f the
circumstances of an individual change, the content and selection of memories will
subtly change to conform to the demands of these new experiences.(Singer 1993
pi69) Autobiographical memories coalesce around the attitudes or frames of
reference of a personality. Recollections are shaped to fit the current attitude.
As we can see, there are lots of ways at looking at personality. The trend of
study seems to be headed in the direction of a wholeness, a seeing of the person as
the persons entire life. "Personality may be regarded as a compound or Gestalt
possessing properties not present in its separate constituents."(Gale 1992 pl21)
This is a relatively old concept (Murray 1938, William James 1890). The western
idea of logic and the finding of a system of objective principles to explain
everything (Dreyfus 1986 p2) subsumed the wholeness idea of personality. We
westerners try to codify things, find the rules, break it down into objective, non-
subjective behaviors. Now we are learning to see personality as more than
objective rules of behavior that can be separated from the rest of a persons life.
We are learning to study the whole person. "By viewing personality as emergent,
the constructivist approach emphasizes the study of the whole, not the isolated
parts."(Gale 1992 pl21)

Which Brings Us to Memory
Human memory has been studied for many years. It keeps getting studied
because we cant seem to pin it down. [T]he enormous recent proliferation of
articles and books on human memory does not necessarily imply a commensurate
increase in knowledge. (Eysenck 1977 pi) Human memory studied in
conjunction with an awareness of the importance of individual differences is a
relatively unstudied subject (Eysenck 1977). Eysenck doubts whether memory can
be decoupled from other cognitive processes. Locke argued that personal identity
is not a matter of sameness of immaterial soul, or of bodily continuity, but of
memory.(McLeish 1993 p551) To start I will discuss the terms and concepts
surrounding the study of decoupled memory.
According to the literature, we have iconic, echoic, active, working,
acoustic, articulatory, primary, secondary, episodic, semantic, short-term,
intermediate-term, and long-term memories, and these memories contain
tags, traces, images, attributes, markers, concepts, cognitive maps, natural-
language mediators, kernel sentences, relational rules, nodes, associations,
propositions, higher-order memory units, and features. (Eysenck 1977 p4)
Human memory has been poked and prodded and coerced. Weve given
names to all sorts of pieces of memory. Im less interested in defining these
memory terms than in describing the general processes and structures that form the
network of connections that is understanding and personality.
The word, memory, is used for different things. It is representational;
memory refers to an episode that we remember such as a particular birthday party.
and all that was involved at that party. Sights, scents, tastes, sounds, what games

were played, what the weather was like, whose mother hosted it, and so forth leads
to an internal representation of that party. A different party attendee may remember
all those things in a slightly different way. The simple fact is that neither animals
nor humans are machines that take in information exactly as presented and give it
back in the same form.(Spear 1994 pi 1) We all add our own individual
interpretation to represent the memory. Memory also refers to the entire process of
learning, storage and retrieval. Memory is representation and memory is
process.(Spear 1994)
Learning. The first part of the process of memory is the acquisition of
information or learning. Spear (1994) makes a distinction between memory and
learning. He says it is evident that they are not the same thing because all things
that we have learned simply are not equally accessible from memory.(p2)
Schank (1982), on the other hand, more closely associates memory with the
processing of new input, which is learning. The structures for storage of old
experiences are the same structures we use to process new experiences. We are
reminded of old experiences as we process new ones. The new experiences are
passed down through the old ones, so we are reminded of those previous
experiences. The structures adapt as new information is processed; different
indices, more generalizations, new structures. Thus, reminding not only tells us
about memory organization, it also signals memory that it will have to adapt to the
current episode. Reminding is the basis for leaming.(Schank 1982 p36)
Understanding is reminding (or re-cognizing, to take that word literally) and
reminding is finding the correct memory structure to process an input.(Schank

1982 p79) Schanks conclusion of learning, that understanding and reminding are
the same thing, combines these processes of memory into a description of learning.
Understanding is finding the right place in memory to put a new experience.
Reminding is running across other memories while looking for the new place.
Learning is having been reminded of other memories and having put the new
experience in the right place.
Learning is one part of the memory process. The acquisition of memory is
learning. Norman identifies three kinds of learning: accretion, tuning and
restructuring (Norman 1993).
Accretion is the accumulation of facts; vocabulary, learning about current
events and trade information in your field, learning the history of a place before you
go there. Learning by accretion is adding to the stockpile of knowledge. With the
proper conceptual framework, accretion is easy, painless, efficient. Without a good
conceptual framework, accretion is slow and arduous and difficult.
Tuning is the hours and hours of practice between being a novice at
something and being an expert. Tuning is the learning by practice. Note the
discrepancy between the apparent ease and automatic nature of the experts
experiential behavior and the laborious effort over a period of years required to
reach that stage. Moreover, expert behavior must constantly be retuned.(p29)
The third kind of learning that Norman identifies is that of restructuring.
Restructuring is the forming of the right conceptual framework. Restructuring is
the hard part of learning, where new conceptual skills are acquired.(p30) A proper
conceptual structure gives the learner the proper tools to reflect, to explore, to

compare and integrate new knowledge. A proper conceptual framework makes
accretion and tuning easier.
Wyer and Srull identify three levels of encoding and organization, which is
learning. The first level is the Comprehender which interprets input information in
terms of low level noun, attribute, and action concepts(Wyer 1989 pi 15). The
second level occurs when a specific processing objective (e.g. to form an
impression of someone, to explain the occurrence of an event, etc.) requires an
interpretation of pieces of information in terms of more abstract goal-relevant
concepts. The third level is when individual pieces of information are organized
into a configuration of interrelated features that in combination function as a single
unit of knowledge. Initial processing objectives affects where the information is
stored. Objectives at the time of recall determine where information is sought.
Wyer and Srull use the example of a subject describing another person in
terms of personality traits to illustrate that how a person interprets another persons
behavior depends on the interpreting person.
The initial interpretation of the persons behaviors in terms of trait concepts
presumably influences the nature of this representation and its implications.
Consequently, the particular concepts that are used to interpret the behaviors
at the time they are first considered may have a substantial impact on the
judgements that are ultimately made. (Wyer 1989 pi 17)
A recently activated trait concept is more likely to be used to interpret behavioral
information than a concept further from consciousness. This leaves us with the
possibility of interpreting a persons behavior via a concept used recently for some
other, unrelated purpose, that just happened to be the most recently used concept.
The concepts that happen to be activated at the time information is received may

not only determine how information is encoded (i.e., the interpretation given to it)
but also may influence which information is encoded and stored in memory.(Wyer
1989 pl40) Differences in the accessibility of concepts may reflect differences in
the frequency with which the concepts are typically used by people of different
social or vocational roles, have different recreational interests or for other reasons
make use of these concepts (Wyer 1989).
Moods can also take the place of concepts if mood is more readily accessed
than a goal or objective. [Ejmotionally charged material can automatically capture
the attentional resources one has available. (Wyer 1989 pl42)
Retrieval and Forgetting. Another process of memory that has been studied
separately is the process of retrieval and the very common break-down of this
process, forgetting. There are many processes that may affect retention:
interference from previous or subsequent learning; hypothetical decay of the
physiological underpinnings of memory; a consolidation or elaboration process
occurring subsequent to learning to facilitate storage of the memory; changes in
stimuli (whether subtle or obvious and whether external or internal to the organism)
from those that had been present during original learning; and special cues just prior
to testing that may facilitate retrieval.(Spear 1994 pi5)
With the subject of retrieval from memory comes the lack of it or forgetting.
Forgetting is something humans expect to do; its ubiquitous nature is one of its
striking features.(Spear 1994 p36) Spear outlines a framework of assumptions
and ideas around forgetting. Whatever is learned is permanent, so long as there is
no neurophysiological damage, is the major assumption. One of the ideas is that a

memory is most readily remembered in similar circumstances to those when first
learned. Another is that an episodic memory is multi-dimensional it has a group
of attributes. And a certain number, kind or percentage of those attributes needs to
be aroused for that memory to be retrieved. These are the ideas and assumptions of
this conceptual framework. So then the issues of forgetting revolve around these
assumptions about human memory.
Remembering is largely influenced by context. That is, the more nearly
similar the contextual circumstances (for example the context of the physical
environment: color, smell, shape of the room; and the internal circumstances of the
rememberer; mood, hormonal conditions, state of hunger or fatigue, etc.) at the
time of remembering to the time of acquisition, the better the retrieval. As time goes
by, ones circumstances change; the external environment changes, ones internal
conditions change: one gets older, hormones change, habits change, what we attend
to may be changed by new learning in the interim.
Compared with contextual change, interference probably is more
pervasive, influential and widely accepted as a source of forgetting.(Spear 1994
p85) Interference is the effect produced by conflicting memories. Spear uses the
interference example of associating the English word white with the French word
blanc and the German word weiss. There is the interference in remembering the
learned-first association caused by learning a second association, called retroactive
interference. The interference of learning and recalling a second association by the
first association is called proactive interference. Spear feels that proactive
interference is the more important determinant of most forgetting. However, if the

respective contents of two memories are not at all similar, there is no reason to
expect that learning one will influence retention of the other.(Spear 1994 p88)
Spear claims that there is no evidence for the decay theory, that acquired
memories disintegrate in the sense of being lost forever. He feels that a more
reasonable assumption may be that memories do not decay and that they are at least
potentially accessible.
There is the theory that a certain amount of neurophysiological activity is
necessary after the learning of a memory in order to solidify that memory into a
relatively permanent form. If a second set of learning happens right after a first set,
then the second learning interferes with the consolidation of the first set, so
forgetting of the first set of learning happens.
Weve all forgotten things. Weve all had the experience of remembering a
long forgotten memory while reminiscing with an old friend. Havent we all said:
Id almost completely forgotten about that. There are lots of reasons for not
being able to access a memory. Some of those reasons are truly trouble with
retrieval. A few of the reasons are really trouble at the time of learning and so they
never are satisfactorily put in memory in the first place.
Personality and Memory Studied Together
Personality and Memory havent often been studied together. Eysenck
considered his 1977 book one of the first to present the ideas of both together.
Spear briefly acknowledges that individuals encode and represent memories in

accord with their own past experiences as well as contemporary environmental and
physiological influences. (Spear 1994 p305) In many of the books on memory
that Ive researched, the authors acknowledge that individual differences influence
memory, but few of these authors go into any more detail, as if its not a major
contributor (Spear 1994, Myer 1992, Wrench 1969). Some authors do, however,
incorporate the individual differences into memory structures and processes (Wyer
and Srull 1989, Eysenck 1977, Schank 1982). [T]he focus is on things that can
be studied, measured, and reproduced in carefully controlled laboratory
experiments. This leaves out much of social interaction, humor, emotion,
motivation, and creativity.(Norman 1993 pi 16) It is hard to study those parts of
people for which we dont have a ruler. Unfortunately, because of the complexity
of its scoring methods, ego psychologys approach to early memories has yet to
have a major impact on mainstream personality and clinical psychology.(Singer
1993 pl67) When we dont have rulers for measurement it becomes subjective
measurement which is less than satisfactory to western researchers.
One of the more immediately obvious facts about human learning and
memory is that there are enormous individual differences.(Eysenck 1977 pl59)
Eysenck presents theories that indicate that there are several consistent, replicable
differences in learning and memory between those high and low in arousal,
between introverts and extroverts, between those low and high in anxiety and
between young and old. In many cases, no theory ignoring individual differences
could possibly be accurate.(Eysenck 1977 pi60)
Wyer and Srull realize that a persons vocation, interest and mood influence
what and how information is encoded into memory. Schank says that our

experiences alter an individuals view of the world to such an extent that we can
expect major differences(Schank 1982 pl27) in their memory structures and
expectations. Singer and Salovey write about self-memories as descriptions of
personality. They emphasize that what one remembers and talks about is more
important than what one has forgotten.
The influence of personality in memory is just starting to catch the attention
of mainstream researchers. There are those who study and have studied this subject
(Murray 1938, Eysenck pere et fils 1977) and their work seems to be spreading.
Personality. Memory and Understanding in Computers
Personality Studied in Computers
Sloman (Franklin 1995) offers a step towards a computational theory of
emotions, attitudes, moods, character traits, and other aspects of mind, so far not
studied in Artificial Intelligence.(p386) Attitudes are dispositions to behave in
certain ways. Emotions are episodes with dispositional elements, they are states
produced by motives and beliefs. Sloman views a motive as a disposition to
produce certain effects. Motives have quantitative dimensions. Insistence of a
motive measures its power to interrupt. Importance measures its likelihood of
adoption, of success in producing action. Urgency of a motive measures the time
left in which to act. Intensity measures how vigorously it is pursued. Distress .

quantifies failure to achieve and pleasure quantifies fulfillment. Sloman expects
intelligent machines to require emotions.(Franklin 1995 p387)
Sloman has provided us with a rich and varied account of goals, motives,
emotions, and so on as high-level, perhaps virtual, mechanisms of mind. I
suspect these concepts from human folk psychology will prove useful,
and perhaps indispensable, in the design of sophisticated, adaptive,
autonomous agents. (Franklin 1995 p391)
Several other theorists insist that a uniqueness is necessary for the qualities
we wish in an artificial mind (Franklin 1995, de Grasse-Tyson 1995, Penrose
1989). There is some study into creating the personality without the mind (Nass
Computer Memory and Understanding
Memory in computers has been something researchers could create,
measure and study different aspects in measured, controlled ways.
Franklins new mind paradigm, one supported mind as a control device, its
function being to produce the next action(Franklin 1995 p329), proposes that we
view memory as an active process of reconstruction as opposed to taking a
photograph from a folder. Memory plays a supporting role to the mind as a control
Franklin likes the theory of a sparse distributed random access memory by
Kanerva. Kanerva lists several similarities between his sparse distributed memory
and human memory. One of which is the way we can immediately know if we

know something or not. Kanervas memory can make such decisions based on the
speed of initial convergence of a search. If its slow, I dont know.(Franklin
1995 p337) Another similarity is the on the tip of my tongue phenomenon. In
sparse distributed memory, this corresponds to the cue having content just at the
threshold of being similar enough for reconstruction.(Franklin 1995 p337)
Rehearsal is similar in sparse distributed and human memories. An item is written
lots of times to many different locations so to be retrievable with fewer cues. And
forgetting happens over time as a result of other writes to memory.
Kanerva says that understanding can be measured by the ability to predict.
Kanervas system can learn using prediction and feedback. It has an internal world
model to assist in this prediction mechanism. Kanerva proposes to add a preference
function into his system for value based decisions; this pattern is good, that one
bad, a la Ackley-Littman(Franklin 1995 p341). Wed want this system to choose
desirable actions and avoid undesirable ones. Wed want this system to have
values; a sense of good and bad (a subjective judgment).
A small model of Kanervas system was built. The bit vector used was a
factor of ten smaller than proposed by Kanerva. It didnt work well, almost any
pattern was accessible from any other. More complexity is needed to adequately
test a sparse distributed controller. Kanerva has provided a mechanism potentially
capable of reconstructing, or perhaps creating images, actions, plans, narratives,
plays, paintings, concerti, poems, mathematics. The line between memory and
creativity has become blurred.(Franklin 1995 p344)
The paradigm I explore in this thesis is Schanks (1982) theory of dynamic
memory; a memory that can alter the memory structures that catalog what is known.

Lebowitz (1980), a student of Schank, wrote a dynamic memory that would
generalize stories that it read off the wire and apply those generalizations to
understand future stories. IPP, as it was called, developed different memory
structures dependent upon the order of the stories. Dyer (1983) developed BORIS,
a thematic pattern understander. Part of his concern was cross-contextual
remindings. Cross-contextual remindings indicate connections in memory that
touch different kinds of knowledge, which contributes to understanding. Dynamic
memories learn. New information comes in and changes the memory network just
by forming more connections.
Case-based reasoners compare and contrast new cases with previous cases
Kolodner (1993). Case-based reasoners make inferences from old cases and ask
questions about new cases. And learning happens as part of the process of
integrating a new case into memory.(pxiv) The more cases or previous learning
that a case-based reasoner has the better the inferences can be. Case-based
reasoners are a direct application from Schank* s dynamic memory. More on
dynamic theory and case-based reasoners in the next chapter.
These experiments of human-imitating computer memory points to the
complexity of the connection network as being important. More complexity is
necessary for Kanerva?s system to work as expected. IPP revealed differences in
the memory network after a few hundred stories had been input. This individuality
of memory networks was only briefly mentioned in Lebowitz (1980). It was
noticed at this less complex level, but not considered very important by Lebowitz.
More work needs to be done to achieve the necessary level of complexity.
for a memory system to reveal its individuality. Penrose (1994) thinks that a

chaotic dynamic evolution could manifest itself as an individual. Chaos needs
complexity to form. Understanding needs a tangled web(Minsky 1985) of
connections. The more complex a memory network, the harder it is to exactly
duplicate it. If it were chaotic, it would be nigh on impossible to exactly duplicate
it. Tiny differences in input could quickly become overwhelming differences in
output.(Gleick 1987 p8) Here is the individuality. The more complexity, the
more understanding. The more complexity the better the chance for chaos and

As I said before my hypothesis is that personality is constituted by
connections between concepts in memory. To investigate this hypothesis, I need a
way to represent concepts in memory and a method for connecting them. I use
cases to represent concepts in memory and I use Schanks (1982) theory of
dynamic memory to provide the method of connection. I extend the basic theory to
encompass context and personality.
Case-Based Reasoning
Cases provide sources of knowledge that can be used to support almost
every type of inference.(Kolodner 1993 p73) Case-based reasoning can reason
over a variety of tasks. Cases can help anticipate problems that might arise in
planning; constructing solutions, and repairing solutions. Cases can suggest
solutions and warn of potential failures in design.

Definition of a Case
Cases "represent specific knowledge tied to specific situations..."(Kolodner
1993 p9) A case is a collection of attributes and pieces of a concept. The concept
of a car might contain information about the make and model, the year and
maintenance history. A case-based reasoner that diagnoses car problems would
also want information about symptoms; car behavior that is anomalous or funny
noises it makes when it does this. A case will contain information surrounding a
concept. It might also contain information pertinent to the objective of the reasoner.
A case in a different case-based reasoner could involve the parts of an
activity, for instance, a playing-tennis case. This case might contain information
about the different kinds of court surfaces there are. It might also hold information
about rackets and shoes and strings. This case would include information about the
rules and scoring of the game. Backhands and forehands and service returns are
important to playing tennis. These parts of the concept of playing tennis are
complicated ideas in themselves. They may be contained in a different structure that
is referenced by the playing-tennis case. The playing-tennis case might also contain
other information about playing tennis that would help the reasoner solve tactical
and strategic problems. A case represents a concept by collecting the parts that go
with that concept.

The Range of Case-Based Problem Solving
The range of reasoning tasks cases support can be broken roughly into two
categories: problem-solving tasks, including design, planning, and diagnosis, and
interpretive tasks, including understanding, justification, and projection. (Kolodner
1993 p74) This is something of an artificial distinction since all tasks require some
degree of each. In problem solving, cases suggest ballpark solutions which the
solver then modifies and adapts to fit the current problem. In interpretive tasks,
cases are compared and contrasted with each other to get an understanding of the
new experience with respect to the experiences already known.
Dynamic Memory
An excellent description of dynamic memory is provided by Kolodner:
Remembering, understanding, experiencing, and learning cannot be
separated from each other. Our memories, which are dynamic, change as a
result of our experiences due to the new things we encounter, the questions
that arise in our minds as a result of these experiences, and the way we
answer these questions. We understand by trying to integrate new things
we encounter with what we already know. Thus, understanding causes us
to encounter old experiences as we process new ones. Those experiences
provide expectations that allow the understanding process to be simple and
top-down....[D]ynamic memory never behaves exactly the same way twice,
because it changes as a result of every one of its experiences. Reasoning
processes thus never encounter exactly the same knowledge on two
traversals through memory.(Kolodner 1993 pl01-102)
Dynamic memory is a computer-based theory that models human memory. Old
experiences provide expectations about new experiences. New experiences either

confirm or deny expectations. If a new experience denies old expectations, then the
memory has changed its expectations to including the expectation that sometimes
there are exceptions.
Example of Inference from Limited Information
An example, also provided by Kolodner (1993), may help illustrate what is
happening in understanding. Consider the following sequence of statements:
John went to a restaurant. He ordered lobster. It was good. We know, from
experience, what usually goes on in a restaurant. We understand that John is
probably a customer. That allows us to connect the 'he' in the second sentence to
John since customers order things to eat in restaurants. We also would be likely to
infer that John sat down at a table and that there was a menu for him to look at. We
could be wrong about this, but we need someplace to start to understand these
sentences. Our expectations of what usually happens in a restaurant provide this
place to start. The it in it was good probably refers to the lobster and doesnt
imply a moral sense of the word good, but a more physical sense of that word.
We also infer that the lobster was cooked and brought to John by a waiter or
waitress and that John ate at least some of the lobster and found it tasty. John
probably paid the bill, tipped the waiter. He was probably hungry when he came
into the restaurant and was no longer so when he left. Our knowledge about what
generally happens in restaurants allows us to explain the connections between the
different pieces of the story .(Kolodner 1993) What might we infer from this little

story if we had had previous experiences that everyone who eats lobster gets sick
and dies, or if our previous experiences with restaurants included paying for
ingredients and then cooking your own meal.
Our understanding of what goes on in restaurants is a sequence of events
we expect to happen. This information could be contained in a structure that
Schank (1982) calls a script or scene. Particular experiences of restaurants, if they
matched the script or scene, would become linked to this scene. A scene is, thus, a
generalization of restaurant experiences.
Reminding and Understanding
Reminding, understanding, learning and memory are all the same thing
(Schank 1982). New experiences come into memory and we attempt to find the
closest matching experience to this new one. That is, after all, about where it will
be placed in memory, somewhere close to similar memories (Schank 1982). This
searching to find the closest matching experience means that we touch on old
memories that are similar in some respect to the new experience. This is reminding.
We are reminded of similar cases while processing a new case. Understanding is
putting the new case in the appropriate place in memory (Schank 1982).
Understanding is finding the closest matching experience.

Memory Structure
Dynamic memory has one kind of structure that holds two kinds of
information: specific cases and abstractions of cases (Kolodner 1993 pl06). Each
set of cases and their abstractions are labeled according to the category they are in
and the way they differ from the category. The abstractions can be abstractions of
abstractions of cases to any level.
Schanks MOPs, or memory organization packets, organize situations
whose activities are similar. MOPs package scenes by listing the sequence of
scenes involved in an activity. They index instances or cases that provide
expectations about setting, characters, and sequences of scenes. Scenes index
scripts and instances of the scene that provide expectations of how the scene will
unravel in different circumstances. Scripts are structures that represent stereotypical
sequences of events that happen in one setting or scene. So a restaurant MOP
might contain scenes of sitting down, looking at a menu and ordering. It might
contain a different script for fast-food restaurants.
Case-Based Learning
A dynamic memory with indexed general and specific knowledge learns in
several ways: it acquires new cases, it re-indexes cases it already has, it creates
new generalizations, it learns what to pay attention to, and it leams new ways to
index cases. (Kolodner 1993 pl36) When learning a new domain, we pay

attention to the things our previous experiences tell us is important. As we make
mistakes and encounter failures, we try to explain to ourselves what went wrong
and how we can change our expectations to make fewer mistakes. These
expectations move us toward paying attention to the right details. They also point
to new ways of indexing cases.(Kolodner 1993 pi36) Case-based reasoning
focuses on the latter part of this process, looking at the ways in which retrieved
cases are used for adapting old solutions to new problems.
Dynamic memory theory is a theory that encompasses learning, memory,
and understanding. Case-based reasoning is a method for solving certain kinds of
problems using cases, a means of representing information, and dynamic memory
The way new experiences are understood is dependent on the state of the
memory connections already there and the context in which it is being matched.
Since the state of memory connections is personality, the personality determines
how a new experience is understood. Since the memory structure is determined by
context, personality is determined by context. If whats important is determined by
past experience, then memory personality determines whats important.
Whats important depends on the context, which depends on past experience which
is memory, which is again personality.

The memory system I built, DMP, acquires new cases, re-indexes cases it
already has, and creates new generalizations. This memory learns in three of the
five different ways a dynamic memory can. The other two ways of learning that
were mentioned: learning what to pay attention to and learning new ways to index
cases can be incorporated into the cognitive model by a mechanism for context, and
incorporated into my dynamic memory by assigning a weight to attributes when
matching cases. (Note that for the initial experiments, DMP weighs all attributes
equally.) A new case that comes in under one context may be placed in memory
under a different generalization than if it were to come in under a different context.
The more connections to a case, resulting from consideration in multiple
contexts for example, the better that concept/case is understood. For example, a
phone book might be understood in several ways, in different contexts. A phone
book is connected to an abstraction for finding resources. It may also be connected
to a concept for something to sit on to help a child better reach the table. It may
also be connected to a concept for how can I prop up this broken table leg. The
contexts assign different weights to certain attributes. The something-to-sit-on or
prop contexts may assign a greater weight to the size and shape attributes of the
phone book as opposed to color or bookishness attributes.

Permanence in Memory and Personality
The state of the memory connections is personality. Since a dynamic
memory changes its state with every input, does this mean that personality changes
with each new input? In a sense, yes. Personality is a dynamic process, a
constantly changing configuration of thoughts, feelings, and actions.(Zucker et al
1992 p8) However, for recognizable personality to emerge from a memory, there
will have to be many connections. After a certain number of inputs a structure is
built and most of the structure has acquired a permanence. Kolodner (1993)
provides an example:
At the original Legal Seafoods Restaurant in Inman Square, waiters
presented the bill after taking the order (before the food came). The first
time I went there, I found this anomalous, as most people did. I had never
been in a restaurant before where one both sat down at a table to order and
paid before getting the food. The second time I went there, I was reminded
of the first and was therefore able to expect that I would have to pay before
eating, (pi07)
Many years later, the restaurant changed their method of billing and conformed to
the usual sit, eat, then pay billing method of most restaurants of that sort. I
continue to be surprised to this day each time I go into Legal Seafoods. The
original index is still there, I suppose, and provides me with the expectation that I
will pay before eating.(pl07) There is some structural permanence in memory that
serves as the basis for the continuity of memory and personality.

DMP Memory
The category of case-based inference that includes understanding is the
interpretive sort of inference. I have cases of objects and the task will be to classify
those objects as having a certain number of attributes in common. The task is to
understand objects in terms of other objects.
Cases Represent Objects
Objects could be things like tennis balls, books, bugs, dogs, computer
chips, trees, mountains, planets, shoes, chairs, etc. My objects have six different
attributes to compare. The attributes Ive chosen are size, shape, color, being alive
or not, functionality and an attribute of being inside or outside a house. (These last
two attributes are more subjective than the other four. For any particular object, I
choose these values). New objects are understood in terms of previously
understood objects. For example, small, round, yellow objects may be grouped
apart from large, tall, brown objects.
Classifying Objects
Admittedly, if I describe mountain shapes as tall, they will generalize with
tall light poles or tall trees more readily than amorphous shapes of clay. If I classify

mountain shapes as amorphous or roundish, they will more readily generalize with
amorphous clothing or balls. So the words or attribute values that I, as human, use
to describe the objects influence the system.
If I have a memory structure in place and I input a new case that doesnt
have a color, for instance, the nearest matching case that is found can suggest the
most likely color. This scenario might occur: most mountains, since they are tall
and large, are similar to trees and most trees are brown so mountains are probably
brown. Or another scenario: most mountains are similar to balls since they are
round and are outside and the biggest ball I know of is orange so mountains are
probably orange.
Classifying a new case is dependent upon what the system has already
encountered and understood. Perhaps the unknown attribute of the new case is
whether it should be classified as an inside thing or an outside thing. If the nearest
matching case was a tree, the new case belongs outside (if thats the value of that
attribute of the tree object). If the nearest matching case was a tall staircase or a set
of bleachers in a gym, the new case belongs inside.
These examples of finding an unknown attribute by finding the nearest
match are two examples of the influence of state. A different memory structure
might classify a new object in a different substructure and give it a different attribute
filler. Different personalities might say it was a horse of a different color.
In this memory, if a sufficient number of the six attributes of two cases
match, a generalization will form with these two cases indexed from the
generalization. This memory could discover that a round, brown mountain that
belongs outside is very similar to a round, brown medicine ball that belongs inside.

The discovery that a soft, yellow, tennis ball and a soft, yellow summer day are
pleasurable may lead to a generalization that soft, yellow things are pleasurable. If
the only living things that this memory has seen have been red with black spots,
then it might be suggested that a horse of an unspecified color is red with black
Each different memory structure represents a different personality.
Different connections in memory influence what and how things are understood or
incorporated into the memory. Each different sequence of input cases may result in
a different memory structure. Each different memory structure understands new
input differently. DMP is a very small example of a personality/memory structure.
We humans know that our experiences "taint" the way we feel about new
experiences, whether that be for good or bad. (Gale 1992 p373) It is clear that old
experiences taint new experiences in DMP.

Dynamic Memory with Personality, or DMP, is a program that figures out
the unknown attribute of a new case given a memory of other cases. The value of
the unknown attribute depends on the cases the memory already has in it. The
value depends on the memory connections that have already been made; it depends
on personality. For example, if the memory is empty and a new case comes in
without a color value, the new case will still have no color value. DMP hasnt
experienced color yet. If the memory has one item in it, then that items color will
be given to the new case. Its the only color DMP knows about. In addition, DMP
adds the new case to its memory, possibly changing its personality in the process.
A case is meant to depict an object. I have given the objects attributes that
mean something to me, but mean nothing to the running program. DMP knows
only when attributes match. Objects have the attributes of function, aliveness,
shape, size, color and whether they belong inside or outside a house. Two cases

must match a certain number of these six attributes to form a generalization. For
my own convience, I have given the cases names like mountain and tennis ball.
This name is a comment field so that I can remember what I intended the case to
look like.
My code is written in Scheme, a lisp-like language. My cases are
represented using lists. The attribute is the first item in the list. The value of the
attribute is a list and follows the attribute name.
Case Example
(cl9 ;the unique name to distinguish this case from all others
(function (tool)) ;function is the attribute and tool is the value
(aliveness (non-living)) ;aliveness is the attribute name and non-living
;is the value filler of this attribute
;(attribute (filler))
;this case depicts a long shaped object
(in/out (out))
(shape (long))
(size (medium))
(color (*var*))
(name (rake))
;it can come in many colors so the value is variable
;this is the comment field useful for me and other
;humans, this field is not used for matching
;this field is used in generalizations formed from two
;cases, the two cases would form the filler values of
;this attribute
Attributes match when the values of the attributes match: (size (medium))
matches (size (medium)) but not (size (large)). Variables match everything: (color
(*var*)) matches (color (red)) and (color (green)). If enough attributes match
between two cases then the cases match and a generalization is made with those
attributes filled in and any non-matching attributes filled with *var* values. The
generalization will fill its links attribute with the two cases that matched.

A generalization is a case-like structure that contains the common values of
two or more cases. So if two cases have the same size, shape and color the
generalization structure will have those values filled in. The other attributes will be
variables (*var*). If a case has a variable for any attribute and a matching case does
not, that attribute in the generalization will be filled with the non-variable value.
Examples of Generalizations
Generalization Example 1.
Case 1: (function (necessity)) Case 2:
(aliveness (non-living))
(in/out (*var*))
(shape (amorphous))
(size (medium))
(color (white))
forms the generalization: (function (necessity))
(aliveness (non-living))
(in/out (out))
(shape (amorphous))
(size (medium))
(color (white))
(links (case2 easel))
These two cases match in all six attributes.
(function (necessity))
(aliveness (non-living))
(in/out (out))
(shape (amorphous))
(size (*var*))
(color (white))
Generalization Example 2,
Case 3: (function (pleasure))
(aliveness (non-living))
(in/out (in))
(shape (square))
(size (small))
(color (yellow))
Case 4: (function (pleasure))
(aliveness (non-living))
(in/out (out))
(shape (round))
(size (small))
(color (yellow))

forms the generalization: (function (pleasure))
(aliveness (non-living))
(in/out (*var*))
(shape (*var*))
(size (small))
(color (yellow))
(links (case3 case4))
These two cases match four of the six attributes.
Runs of DMP
The figures on the next pages illustrate snapshots of memory structures.
The boxes are cases with some attribute values listed. The lines connecting the
boxes are the links from generalizations to the two cases generalized. At the top is
the root memory. All memory structures are started as an empty list. There are
snapshots of memory before and after a new case is input.
In the first run, DMP knows about a white shirt and small yellow things.
Those cases are already in memory. When I input a huge, non-living, amorphous
thing that goes outside (named mountain) into the memory, the best match is the
white, amorphous shirt. So DMP concludes that the color of the mountain is white.
(See Figure 1 and transcript that follows).
On another ran, DMP knows about balls and a book and a tree. The
mountain best matches the tree and DMP concludes that the color of the mountain is
the same color as the tree, which is brown. (See Figure 2 and transcript that

Figure 4.1. Run#l. State of memory
before input of new case.
Figure 4.2. Run#l. State of memory after
new input that matched white shirt and formed a
new generalization.

Figure 4.3. Run #2. State of memory
before the input of the new case.
Figure 4.4. Run #2. State of memory after
input of new case that best matched tree. A
new generalization is formed.

Transcript of Run #1
Edscheme Interpreter. The expression following the "=>" is the input. The
expression following that is the interpreter's output. A semi-colon indicates a
=> (define ls7 '(c2 c5 c4)) ;;a sequence of cases
=> (clear-dbs) ;;memory is cleared for new run
=> (go-db! Is7) ;;cases are input into memory in the sequence of ls7
=> *db* ;;state of the memory see Figure 4.1
(c2 (*gen3218.83* (function (pleasure)) (aliveness (non-living)) (in/out (*var*))
(shape (*var*)) (size (small)) (color (yellow)) (name (*var*))
(links (c4 c5))))
=> (What-is-the-color-of c8) ;;input of new case c8
The best match is c2
So the color of the (mountain) is (white)
=> *temp* ;;state of memory after input see Figure 4.2
((*gen3238.23* (function (necessity)) (aliveness (non-living)) (in/out (out))
(shape (*var*)) (size (*var*)) (color (white)) (name (*var*))
(links (c8 c2)j)
(*gen3218.83* (function (pleasure)) (aliveness (non-living)) (in/out (*var*))
(shape (*var*)) (size (small)) (color (yellow)) (name (*var*))
(links (c4 c5))))

Transcript of Run #2
EdScheme 4.0, 1991,1993 Schemers Inc. SN #931002
Saturday, March 30,1996, 12:54 PM.
=> (clear-dbs) ;; clear memory for new run
=> (define ls6 '(c6 c4 c5 c7)) ;; a sequence of cases
=>(go-db! Is6) ;; the cases are input into memory in ls6 sequence
=> *db* ;;the state of memory before input see Figure 4.3
((*gen255434.0* (function (pleasure)) (aliveness (non-living))
(in/out (out)) (shape (round)) (size (*var*))
(color (*var*)) (name (*var*))
(links (c4 c6)))
=> c8 ;; the new case is a mountain
(c8 (function (*var*)) (aliveness (non-living)) (in/out (out))
(shape (tall)) (size (huge)) (color (*var*))
(name (mountain)) (links))
=> (What-is-the-color-of c8) ;;input of new case c8
The best match is c7
So the color of the (mountain) is (brown)
=> *temp* ;; the state of memory after input of new case
;; see Figure 4.4
((*gen255434.0* (function (pleasure)) (aliveness (non-living))
(in/out (out)) (shape (round)) (size (*var*))
(color (*var*)) (name (*var*))
(links (c4 c6)))
(*gen255448.2* (function (pleasure)) (aliveness (*var*))
(in/out (out)) (shape (tall)) (size (*var*))
(color (brown)) (name (*var*))
(links (c8 c7))))

The function attribute value of a new input can be found in a similar
manner. If the new case has a variable function attribute value, then the closest
match can provide it. (See Figures 4.5 and 4.6 and transcript of Run #3 in which
DMP suggests that the function of a mountain is to give pleasure.)

Figure 4.5. Run #3. State of memory
before the input of the new case. Balls,
books and the tree have the function of
Figure 4.6. Run #3. State of memory after
input of new case that best matched tree. A
new generalization is formed.

Transcript of Run #3
EdScheme 4.0, 1991, 1993 Schemers Inc. SN #931002
Sunday, March 24, 1996, 3:44 PM.
=> (clear-dbs)
=> (define ls6 '(c6 c4 c5 c7))
=> (go-db! Is6) ;;filling memory with the cases in ls6
=> *db* ;;state of memory before new input see Fig. 4.5
((*gen5066.97* (function (pleasure)) (aliveness (non-living)) (in/out (out))
(shape (round)) (size (*var*)) (color (*var*)) (name (*var*))
(links (c4 c6)))
c5 cl)
=> (What-is-the-function-of c8)
The best match is cl
So the function of the (mountain) is (pleasure)
=>*temp* ;;state of memory after the input of c8 (Fig. 4.6)
((*gen5066.97* (function (pleasure)) (aliveness (non-living)) (in/out (out))
(shape (round)) (size (*var*)) (color (*var*)) (name (*var*))
(links (c4 c6)))
(*gen5077.93* (function (pleasure)) (aliveness (*var*)) (in/out (out)) (shape (tall))
(size (*var*)) (color (brown)) (name (*var*)) (links (c8 c7))))

In the next run (see Run #4 and the following transcript), DMP knows
about balls and a ladybug and a book and a tree. A horse is the new case input into
DMP. The ladybug is the closest matching case to a horse, and those two cases
generalize. The color of the horse is unknown. To find a color for the horse, DMP
finds the closest matching case and decides that that color is the color of the horse,
which is red with black spots.
Figure 4.7. Run #4. State of Memory before input of
new case.

Figure 4.8. Run #4. State of Memory after input of
the horse of an unknown color.

Transcript of Run #4
EdScheme 4.0, 1991, 1993 Schemers Inc. SN #931002
Saturday, March 30, 1996, 11:33 AM.
=> (clear-dbs)
=> (define ls6 '(c6 clO c4 c5 cl))
=> (go-db! Is6)
=> *db* ;;see Figure 4.7
((*gen250550.37* (function (pleasure)) (aliveness (non-living))
(in/out (out))(shape (round)) (size (*var*))
(color (*var*)) (name (*var*))(links (c4 c6)))
=> cl 1 ;;the horse case
(ell (function (pleasure)) (aliveness (living)) (in/out (out))
(shape (round))(size (large)) (color (*var*))
(name (horse)) (links))
=> (What-is-the-color-of ell)
The best match is c 10
So the color of the (horse) is (red/black)
=>*temp* ;;see Figure 4.8
((*gen250550.37* (function (pleasure)) (aliveness (non-living))
(in/out (out))(shape (round)) (size (*var*))
(color (*var*)) (name (*var*))
(links (c4 c6)))
(*gen250576.4* (function (pleasure)) (aliveness (living))
(in/out (out))(shape (round)) (size (*var*))
(color (red/black)) (name (*var*))
(links (cl 1 c 10)))

Summary of Runs
As Ive stated before, new inputs are understood in terms of what has
already been understood. Stated another way, new inputs are interpreted in terms
of what is already in memory. If its memory contains only things that are yellow
then DMP, unless told otherwise, is left to assume the next thing will be yellow.
DMP understands a new input in terms of what is already contained in its
memory. The color of a mountain is unknown until interpreted by DMP in terms of
objects that it already knows about. DMP understands a mountain in that it finds
the closest matching object or generalization and gives the color it finds there to the
mountain. If the best match is a tall, brown tree, then the color of the mountain is
brown. If the best match is an amorphous white shirt, then the color of the
mountain is white.
If DMP understands only pleasurable things then, unless told otherwise,
new things are pleasurable. Could this be comparable to optimism in humans? If
DMP only knew about tools, then everything is a tool. Perhaps this is indicative of
an entrepreneurial slant.

Interpretation and understanding of a new experience depends on
personality. I have claimed that personality is the collection of connections among
experiences, or cases, in memory. The collection of connections in memory
personality determines how a new experience is understood. And personality is
determined by how things have been understood in the past.
A collection of connections among concepts or cases is dependent upon the
order of input. Different generalizations are made when cases are input in a
different sequence. A personality that has experienced only blue things assumes
that new things of unspecified color have the color blue. A personality that
experiences something new finds the closest match in memory. A personality that
has experienced many pleasurable things and a few other things would be most
likely to find a match in the larger group of pleasurable things. If most previous
cases have been tools, the match to a mountain would most likely be found in the
larger set of tool cases. A human personality who has experienced only pleasurable
things calls new things pleasurable (Myer 1992).
I have also claimed that there will have to be many connections for a rich,
complex personality to emerge, like that we find in humans. The more complex the

memory connections the more likely chaos will form. And from chaos, a more
than the sum of its parts behavior could emerge.
Complexity is a necessary component for such emergence to happen
(Penrose 1989). More complexity can mean that there will be indices of cases in
more than one generalization or that we start with more complex cases. Matching
cases according to a context would result in multiple links to a case (Kolodner
1993). This could be incorporated into DMP as a matching function that puts
emphasis on one or more attributes. Thus for a context of finding an object in the
house, mountains and light poles would come up last since they are found outside.
The in/out attribute would be weighted more heavily than other attributes. A
context of bugs would weight the attributes of size and aliveness more than
inside/outside or color. Thus, bugs could be found under the indoor generalization
and under the small, living things generalization.
DMP is a dynamic memory because it changes with each new input. DMP
is case-based because it compares and contrasts cases. A case-based reasoner is
more than the sum of its parts... [W]e must [...] examine the case-based reasoner
as a whole, looking for what emerges when th[e] component parts are put
together.(Kolodner 1993 p564) Kolodner is looking for interesting things to
Each tiny memory structure that DMP builds embodies a personality.
However, to make that personality more like a human personality, bigger, more
complex memories need to be built. DMP is an example of the studies that need to
be done. More cases and more complicated cases, context-based matching and
perhaps different kinds of cases is important to finding a more complex/interesting

personality in memory. For instance, different kinds of cases might represent
objects and activities and situations. All of these features could be added on to
DMP to form more complex memory structures.
We will look for emergent behavior. By emergent behavior, I mean any
unexpected, unpredictable behavior, for instance, the merging of disparate cases
and the creation of highly novel solutions.(Kolodner 1993 p565) Those instances
may be the birth of a human-like personality. We must allow it experience as a
child experiences, and watch it learn. Will it begin to find highly novel solutions?
Will it begin to favor some cases over others? Will this tiny personality begin to
make choices that dont seem to come from the available information?
Will it decide to dedicate its life to finding solutions to car problems? Will it
re-discover calculus or prove Fermats theorem? Will it dedicate its life to medicine
and discover a cure for cancer? Maybe this personality will find an interest in
physics and discover a safe, efficient means of fusing atoms. Maybe it will unify
field theory or take the uncertainty out of Heisenbergs principle.
Personality is necessary for any mind to dedicate itself to something.
Personality is necessary for any mind to gain the necessary expertise in some field
to discover any blockbuster ideas or invent new methods of research or better
understanding. A driving interest in something is necessary to gain the expertise
that could lead to new conceptualizations a self guiding evolution. Personality is
necessary for a mind to accomplish anything that somebody else didnt already
think about. New discoveries, the betterment of humankind, originality can only
emerge from a mind that directs itself, a mind with interest and drive, a mind with

;;Source code written in Edscheme.
;; 6 is the max any 2 cases can match (names and links do not
(define gensym ;generates a unique symbol based on runtime
(lambda () ; for this session of Edscheme
(string-append "*gen"
(number->string (runtime)) "*"))))
(define newcase
'(id (function (*var*))
(aliveness (*var*))
(in/out (*var*))
(shape (*var*))
(size (*var*))
(color (*var*))
(name (*var*))
(links )))
(define *db* '())
(define *temp* '())
(define remove-last
(lambda (Is)
((null? Is) '())
((null? (cdr Is)) ())
(else (cons (car Is) (remove-last (cdr Is)))))))
;finds the patt in Is and returns a list,
;the car of which is the stuff in
;the list before the patt and the cdr starts with the patt
(define findpatt
(lambda (patt Is) ; input a patt to find in Is
(letrec ((helper
(lambda (hold dls)
((null? dls) (list hold dls)) ; patt not found

((equal? patt (car dls)) (list hold dls))
(else (if (null? hold)
(helper (cons (car dls) hold)
(cdr dls))
(helper (append hold
(list (car dls)))
(cdr dls))))))))
(helper '() Is))))
(define go-db! ;;integrates a list of cases into the database
(lambda (Is)
(if (null? Is) 'done
(set! *temp* (integrate (eval (car Is)) *db*))
(set! *db* *temp*)
(go-db! (cdr Is))))))
; finds the bestmatching pattern to c in Is
(define bestmatch
(lambda (c Is )
(letrec ((helper
(lambda (c Is count patt)
((null? Is) (begin
;(writeln "in bestmat count count
; " bestmatch patt)
((atom? (car Is))
(if (> (length (countofmatch c
(eval (car Is)))) count)
(helper c (cdr Is)
(countofmatch c
(eval (car Is))))
(car Is))
(helper c (cdr Is) count patt)))
(else (let ((acount
(countofmatch c (car Is)))))
(if (and (> acount count)
(>= acount 6))
(helper c (cdr Is)
(countofmatch c
(car Is)))
(car Is))
(helper c (cdr Is)
count patt))))))))
(helper c Is 0 '()))))
;so the best match has been found, integrate it into *db*

(define clear-dbs
(lambda ()
(set! *db* 1())
(set! *temp* '())))
;;returns a list of the things that matched in case to case
(define countofmatch
(lambda (cl c2)
(letrec ((casematch
(lambda (lsl ls2)
((or (null? Is2)(null? lsl)
(equal? (caar lsl) 'name)) '())
((matchatt (car lsl) (car ls2))
(if (isvar? (cadar ls2))
;will match if both *var*
(cons (car lsl)(casematch (cdr lsl)
(cdr ls2)))
(cons (car ls2)
(casematch (cdr lsl) (cdr ls2)))))
(else (casematch (cdr lsl) (cdr ls2)))))))
(casematch (cdr cl) (cdr c2) ))))
;; each new case generated gets a gensym number as id
(define newcase-maker
(lambda ()
(let ((id (gensym)))
(let ((bones (cons id '((function (*var*))
(aliveness (*var*))
(in/out (*var*))
(shape (*var*))
(size (*var*))
(color (*var*))
(name (*var*))
(links )))))
(lambda msg
(case (first msg)
((type) "newcase")
((show) bones)
((getid) id)
((getname) (cadr (8th bones)))
(cdadr (cdddr ( cddddr bones))))
((putfunc!) (let ((func
(cons 'function (list
(list (2nd msg))))))
(set! bones
(cons id (cons func
(cddr bones))))))
((putalive!) (let ((al (cons 'aliveness
(list (list (2nd msg))))))

(set! bones
(cons id (cons (cadr bones)
(cons al
(cdddr bones)))))))
((putin/out!) (let ((inout
(cons 'in/out (list
(list (2nd msg))))))
(set! bones (cons (car bones)
(cons (2nd bones)
(cons (3rd bones)
(cons inout
(cddddr bones))))))))
((putshape!) (let ((shap (cons 'shape
(list (2nd msg)))))
(set! bones (cons (car bones)
(cons (2nd bones)
(cons (3rd bones)
(cons (4th bones)
(cons shap (cddr
(cdddr bones))))))))))
((putsize!) (let ((siz
(cons 'size (list
(list (2nd msg))))))
(set! bones
(append (list (car bones)
(2nd bones) (3rd bones)
(4th bones)(5th bones))
(cons siz
(cddddr (cddr bones)))))))
((putcolor!) (let ((col (cons 'color
(list (list (2nd msg))))))
(set! bones (append
(list (car bones)(2nd bones)
(3rd bones)(4th bones)
(5th bones)(6th bones))
(cons col
(cddddr (cdddr bones)))))))
((putname!) (let ((na (cons 'name
(list (list (2nd msg))))))
(set! bones
(append (list (car bones)
(2nd bones) (3rd bones)
(4th bones) (5th bones)
(6th bones)(7th bones))
(cons na
(cddddr (cddddr bones)))))))
((putlinks!) (let ((li (cons 'links
(list (2nd msg)))))
;the links has to be a list
(set! bones (append
(list (car bones) (2nd bones)
(3rd bones)(4th bones)

(5th bones) (6th bones)
(7th bones)(8th bones))
(cons li
(cddddr (cddddr (cdr bones))))))))
(else 'havent-done-else-yet)))))))
(define 2nd (lambda (Is) (if (null? Is) 'list-is-null
(if (null? (cdr Is)) 1 no-second-there
(cadr Is)))))
(define 3rd (lambda (Is) (cond ((null? Is) 'list-is-null)
((null? (cdr Is))
((null? (cddr Is)) 1no-3rd-element)
(else (caddr Is)))))
(define 4th (lambda (Is)
((null? (3rd Is)) no-third)
((null? (cdddr Is)) no-fourth)
(else (cadddr Is)))))
(define 5th (lambda (Is)
((null? (4th Is)) no-fourth)
((null? (cddddr Is)) no-fifth)
(else (cadr (cdddr Is))))))
(define 6th (lambda (Is)
((null? (5th Is)) no-fifth)
((null? (cddr (cdddr Is))) no-sixth)
(else (caddr (cdddr Is))))))
(define 7th (lambda (Is)
((null? (6th Is)) no-sixth)
((null? (cdddr (cdddr Is))) no-seventh)
(else (cadddr (cdddr Is))))))
(define 8th (lambda (Is)
((null? (7th Is)) no-seventh)
((null? (cddddr (cdddr Is))) no-eighth)
(else (cadddr (cddddr Is))))))
;;take the list returned from countofmatch and create the new case
with all of these filled in.
(define formgen!
(lambda (cl c2 )
(let ((mats (countofmatch cl c2)) (g (newcase-maker)))
(letrec ((helper

(lambda (Is)
((null? Is) g)
(case (caar Is)
((function) (g 'putfunc! (caadar Is)))
((aliveness) (g 'putalive!
(caadar Is)))
((in/out) (g 'putin/out! (caadar Is)))
((shape) (g 'putshape! (cadar Is)))
((size) (g 'putsize! (caadar Is)))
((color) (g 'putcolor! (caadar Is)))
((name) (g 'putname! (caadar Is)))
((links) (g 'putlinks! (caadar Is)))
(else 'error-in-case))
(helper (cdr Is)))))))
(helper mats)
(g 'putlinks! (cons (car cl) (cons (car c2)
(g 'getlink))))
(list (g 'show)))))))
;(getnewlinks (formgen! c3 c2))
(define getnewlinks ; get the links of a generalization
(lambda (Is)
(if (8th Is)
(cadadr (cdddr (cddddr Is)))
;; the bestmatch of each list is found at the top level
(define integrate ; integrates the case into the database
(lambda (easel Is )
(letrec ((helper
(lambda (als)
(if (null? als) (cons (car easel) als)
(let ((newls (findpatt
(bestmatch easel als) als)))
;(writeln "newls is newls)
(let ( (lsl (cadr newls)))
((null? lsl) (append (car newls)
(cons (car easel) lsl)))
((list? (car lsl))
(if (matchgen easel (car lsl))
(let ((gensanslinks
(remove-last (car lsl)))
(getnewlinks(car lsl))))
(append (car newls)
(append gensanslinks

(list (cons 'links
(helper newlinks)))))
(cdr lsl))))
(append (car newls)
(cons (car lsl)
(helper (cdr lsl))))))
((atom? (car lsl))
(let ((oldcase (eval (car lsl))))
(if (> (length
(countofmatch easel oldcase))
(append (car newls)
(formgen! easel
(cdr lsl)))
(append (car newls)
(cons (car lsl)
(helper (cdr lsl)))))))
(else 'unknown-integrate))))))))
(if (null? Is)
(cons (car easel) Is)
(helper Is)))))
(define What-is-the-function-of
(lambda (cas)
(let ((proc 2nd))
(let ((*temp* (set! *temp* (integrate cas *db*))))
(let ((best (bestmatch cas *db*)))
(writeln "The best match is best)
(display "So the function of the ")
(display (cadr (8th cas)))(display is ")
(cond ((null? best) (proc cas))
((atom? best)(if (isvar? (cadr (proc (eval best))))
(cadr (proc cas))
(cadr (proc (eval best)))))
(else (cadr (proc best)))))))))
(define What-is-the-color-of
(lambda (cas)
(let ((proc 7th))
(let ((*temp* (set! *temp* (integrate cas *db*))))
(let ((best (bestmatch cas *db*)))
(writeln "The best match is best)
(display "So the color of the ")
(display (cadr (8th cas) )) (display is ")
(cond ((null? best) (proc cas))
((atom? best)(cadr (proc (eval best))))
(else (cadr (proc best)))))))))

;;input sequences and test procedure calls
(go-db! Is6)
(What-is-the-function-of c8)
(What-is-the-color-of c8)
(set! *temp* (integrate c9 *db*))
(set! *db* *temp* )
;(countofmatch c8 (eval (4th *db*)))
(define lsl
(define ls2
(define ls3
(define ls4
(define ls5
(define Is6
(define ls7
'(cl c2 c3 c4 c5 c6 c7))
'(c7 c6 c5 c4 c3 c2 cl))
'(c2 cl c3 c4 c5 c6 cl))
'(cl c3 c2 c4 c5 c6 c7))
'(c5 c6 c7 cl c2 c3 c4))
'(c6 c4 c5 c7))
'(c2 c5 c4))
;; the various match functions for matching case to case
and case to generalizations
(define matchcases ;sends matchatt 1 attribute at a time
(lambda (patt const)
(letrec ((helper
;if matchatt returns true then put that pair in a list
(lambda (lsl ls2)
((equal? (caar lsl) 'name ) '())
((matchatt (car lsl)(car ls2))
(if (isvar? (cadar lsl))
;put the filler not the *var*
(cons (car ls2) (helper (cdr lsl)
(cdr ls2)))
(cons (car lsl) (helper (cdr lsl)
(cdr ls2)))))
(else (helper (cdr lsl)(cdr ls2)))))))
(helper (cdr patt) (cdr const)))))
;cdr doesn't include the case or gen name
;;matches two attributes returns true or false
(define matchatt
(lambda (lsl ls2)
(letrec (( matchfillers
(lambda (pattern constant )
((or (null? constant)
(null? pattern)
(equal? pattern constant)) #t)
((or (isvar? constant)
(isvar? pattern)) #t)
((equal? (car constant) (car pattern))

(Matchfillers (cdr pattern) (cdr constant)))
(else #f)))))
(matchfillers (cadr lsl) (cadr ls2) ))))
;/matches a case to a generalization
(define matchgen
(lambda (pat const)
(letrec ((helper (lambda (11 12)
((equal? (caar 11) 'name) #t)
((and (raatchatt (car 11) (car 12))
(helper (cdr 11) (cdr 12))) #t)
(else #f)))))
(helper (cdr pat) (cdr const)))))
(define header
(lambda (exp)
(car exp)))
(define isvar?
(lambda (x)
(and (pair? x)
(equal? (car x) (quote *var*)))))
(define writeln
(lambda args
(for-each display args)
(define cl
'(cl (function (necessity (clothing)))
(aliveness (non-living))
(in/out (*var*))
(shape (amorphous ))
(size (med))
(color (white))
(name (shoe))
(links )))
(define c3
'(c3 (function (necessity))
(aliveness (non-living))
(in/out (out))
(shape (amorphous ))
(size (large))
(color (white ))
(name (shoe))
(links )))

(define c2
'(c2 (function (necessity))
(aliveness (non-living))
(in/out (*var*))
(shape (amorphous ))
(size (med))
(color (white))
(name (shirt))
(links )))
(define c4
'(c4 (function (pleasure))
(aliveness (non-living))
(in/out (out))
(shape (round))
(size (small))
(color (yellow))
(name (ball))
(links )))
(define c5
'(c5 (function (pleasure))
(aliveness (non-living))
(in/out (in))
(shape (square))
(size (small))
(color (yellow))
(name (book))
(links )))
(define c6
'(c6 (function (pleasure))
(aliveness (non-living))
(in/out (out))
(shape (round))
(size (med))
(color (orange))
(name (ball))
(links )))
(define c7
'(c7 (function (pleasure))
(aliveness (living))
(in/out (out))
(shape (tall))
(size (large))
(color (brown))
(name (tree))
(links )))

(define c8
'(c8 (function (*var*))
(aliveness (non-living))
(in/out (out))
(shape (tall))
(size (huge))
(color (*var*))
(name (mountain))
(links )))
(define c9
'(c9 (function (tool))
(aliveness (non-living))
(in/out (out))
(shape (tall))
(size (large))
(color (*var*))
(name (lightpole))
(define clO
'(clO (function (*var*))
(aliveness (living))
(in/out (out))
(shape (round))
(size (small))
(color (red/black))
(name (ladybug))
(define ell
'(ell (function (*var*))
(aliveness (living))
(in/out (out))
(shape (round)) ;barrel shaped?
(size (large))
(color (*var*))
(name (horse))

[Adler 1992] Adler, Mortimer J. The Great Ideas: A Lexicon of Western Thought.
New York: MacMillan Publishing Company, 1992. A collection of 102 essays that
collectively define Western thought. Each essay treats each idea as if the original
authors, whose writings the ideas are drawn from, were sitting around a table, deep
in conversation. The essays include War and Peace, Fate, Justice, Custom
and Convention, etc. The original authors include Homer, Freud, Virginia Woolf,
and others (from the flap blurb)
[Aronoff et al. 1987] Aronoff, Joel; Rabin, A.I.; Zucker, Robert A., eds. The
Emergence of Personality. New York: Springer Publishing Company, 1987. The
published lecture series on personality named for Henry A. Murray. This is one of
four published books. This one focuses on the processes through which
personality emerges and is maintained across life.
[Baron 1982] Sternberg, Robert J., ed. Handbook of Human Intelligence.
Cambridge: Cambridge University Press, 1982. The contribution of Jonathan
Baron is Personality and Intelligence. Barons theory of personality depends
heavily on Deweys theory of good thinking. Baron discusses the cognitive
styles and development stages of thinking, or learning to think, or habits of
thinking, and then considers implications of some issues in the study of personality
for the study of intelligence, including correlations between personality and IQ,
intelligence as a perceived trait, the social context of intelligence and personality
change and its relation to training of intelligence.
[Bergson 1991] Bergson, Henri.(1859-1941) Matter and Memory. New York:
Zone Books, 1991. Translated by N.M. Paul and W.S. Palmer from the fifth
edition of 1908. This book affirms the reality of spirit and the reality of matter,
and tries to determine the relation of the one to the other by the study of a definite
example, that of memory.p9 The aim is to show that it is a mistake to reduce
matter to the perception which we have of it. Matter is an aggregate of images,
which are more than representations but less than things.
[Bernstein 1982] Bernstein, Jeremy. Science Observed. New York: Basic Books,
Inc., Publishers, 1982. The first part is a profile of Marvin Minsky. The second
part is essays of various aspects of science like time, fusion, speculations on the
change in the world due to the atomic bomb and a little history of the Nevada test
explosion, little biographies of Schrodinger, Oppenhiemer and Einstein. The third
part is discussions of subjects that interest Bernstein: chess, identifying crank
theories of science, the interrelatedness of science and mysticism...

[Boden 1990] Boden, Margaret A. ed. The Philosophy of Artificial Intelligence.
Oxford: Oxford University Press, 1990. A collection of fourteen papers
commenting on the philosophy of AI. Some of the papers are: Computing
Machinery and Intelligence by Alan M. Turing, Minds, Brains, and Programs by
John R. Searle, Motives, Mechanisms, and Emotions by Aaron Sloman, The
Connectionist Construction of Concepts by Adrian Cussins, and others.
[Bock 1993] Bock, Peter. The Emergence of Artificial Cognition: An Introduction
to Collective Learning. Singapore: World Scientific Publishing Co. Pte. Ltd.,
1993. The first chapter of Bocks book is a description of a futuristic computer
intelligence named Mada. Mada is turned on and is expected to learn and develop
much as a human child does, over the course of about twenty years. Bock asserts
an effective knowledge base of 1014 units and programming that leams. He favors
the paradigm of collective learning which is evaluation after a synthesis of a
collection of responses. He discusses costs, in terms of time and effort, stability,
performance and parameters of collective learning.
[Boorstin 1994] Boorstin, Daniel J. Cleopatras Nose: Essays on the Unexpected.
New York: Vintage Books, A Division of Random House, Inc., 1994. A
collection of essays that explore some of the surprising novelties and unexpected
continuities in our recent past. How has technology opened new realms of
ignorance and given our times a claim to be the age of negative discovery?
[Callahan 1994] Callahan, Gene. Excessive Realism in GUI Design: Helpful or
Harmful?. Software Development, Vol.2, No. 9 September 1994, pp36-44.
This is about the interfaces of computers and how they hide the computer too well
and limit the effectiveness of a computer for its user. The user cant get under the
hood to see the computer at work and no matter how well youve designed your
desktop its still not a desktop and you have to acknowledge that users know this
and allow for the computemess to come through. Semiotics: the study of signs
which is grouped into index, icon and symbol, is discussed here. Symbols come
down to the language used to describe something and is the most useful for
discussions of abstractions and in self-referencing.
[Caprara 1993] Caprara, Gian Vittorio. Barbaranelli, Claudio. Borgogni, Laura.
Perugini, Marco. The Big Five Questionnaire: A New Questionnaire to Assess the
Five Factor Model. Personality and Individual Differences, Vol.l5,No.3, 1993,
pp281-288. This article presents a new questionnaire for the measurement of the
Big Five Factor Model. The Energy (Extraversion) dimension is organized into
Dynamism, which refers to expansiveness and enthusiasm, and Dominance,
which refers to assertiveness and confidence. The Friendliness dimension is
organized into Cooperativeness/Empathy, which refers to concern and
sensitiveness towards others and their needs, and Politeness, which refers to
kindness, civility, docility and trust. The Conscientiousness dimension is
organized into Scrupulousness, which refers to dependability, orderliness and
precision, and Perseverance, which refers to the capability of fulfilling ones
own tasks and commitments. The Emotional Stability dimension is organized into

Emotional Control, which refers to the capacity to cope adequately with ones
own anxiety and emotionality, and Impulse Control, which refers to the
capability of controlling irritation, discontent, and anger. The Openness dimension
refers to Openness to Culture, which is the broadness or narrowness of ones
own cultural interests, and Openness to Experiences, which in openness to
novelty, tolerance of different values, interest toward different people, habits and
[Carr 1994] Carr, Caleb. The Alienist. New York: Bantam Books, 1994. This is a
novel about the hunt for a serial killer in New York City in 1896. The team tries a
new technique that creates a profile of the killer in order to find him. The leader of
the team is a psychologist who believes in a radical philosophy that every mans
actions are to a very decisive extent influenced by his early experiences, and that no
mans behavior can be analyzed or affected without knowledge of those
experiences. p57 A theory referred to as context. Whilst part of what we
perceive comes through our senses from the object before us, another part (and it
may be the larger part) always comes out of our own mind. William James from
The Principles of Psychology as quoted by Carr on page 2.
[Chalmers 1995] Chalmers, David J. The Puzzle of Conscious Experience.
Scientific American* Vol 273, No 6 December 1995, pp80-86. Chalmers divides
the problem of consciousness into easy and hard, where the easy problems concern
the objective mechanisms of the cognitive system like how sensory stimuli is
integrated into a brain. The hard problem is why this physical process is
accompanied by conscious experience. Since physics, at this point, does not
explain conscious experience, Chalmers proposes that conscious experience be
considered a fundamental feature, irreducible to anything more basic.p83 (Like
mass and charge are considered irreducible in physics). This fundamental property
will have its own fundamental laws associated with it, and the concept of
information will be central to these laws. It may even be that a theory of physics
and theory of consciousness could eventually be consolidated into a single grander
theory of information.p86 Systems with the same organization will embody the
same informationp85 and systems with the same organization have the same
conscious experience.p86
[Chamiak 1985] Chamiak, Eugene and Drew McDermott. Introduction to Artificial
Intelligence. Reading, Massachusetts: Addison-Wesley Publishing Company, Inc.,
1985. This book includes chapters on internal representation, low-level vision
research, language parsing and comprehension, search algorithms, logic and
deduction, memory organization, expert systems and learning. Histories of the
fields of AI and intellectual filaments that connect AI to related disciplines(pix) are
covered in the book.
[Crichton 1995] Crichton, Michael. The Lost World. New York: Alfred A. Knopf,
1995. This novel is the sequel to Jurassic Park. Extinction, chaos theory, self-
organization and dinosaurs. The time-honored scientific approach of reductionism
- taking the watch apart to see how it worked didnt get you anywhere with
complex systems, because the interesting behavior seemed to arise from the

spontaneous interaction of the components. p4 Complex systems tend to locate
themselves between the need for order and the imperative to change, the edge of
chaos. [T]here is enough innovation to keep a living system vibrant, and enough
stability to keep it from collapsing into anarchy p4 Stray too far from this delicate
balance and the living system becomes extinct.
[de Grasse Tyson 1995] de Grasse Tyson, Neil. Cosmic Windows Natural
History Magazine, Vol 104, No 6 June 1995, pp22. Published by the American
Museum of Natural History, New York. This article describes what we can see in
the universe beyond the visible light spectrum; the stuff we can see using infrared
light, ultraviolet light, microwave radiation... De Grasse Tyson also briefly
discusses the telescopes wed like to build that extend our vision to neutrinos and
gravity waves. Until computers can simulate human curiousity and insight, robots
will remain tools designed to discover what we already expect to find.p22
[Drescher 1991] Drescher, Gary L. Made-Up Minds: A Constructivist Approach
to Artificial Intelligence. Cambridge, Massachusetts: The MIT Press, 1991. This
book presents the schema mechanism. The schema mechanism is a general
learning and concept-building mechanism intended to reproduce aspects of
Piagetian cognitive development during infancy.(preface) The mechanism
discovers correspondences among its sensory perceptions (for instance vision and
touch) and eventually learns a representation of an object independently of how, or
if, the object is currently perceived.(preface) The mechanism practices a kind of
learning that uses almost no a priori knowledge of the world.p3 Regularities in
the world perceptions are represented as schemas, each of which predicts some
effects of an action under specified circumstances. The schema mechanism is
principally concerned with empirical learning and with concept invention.p4
[Dreyfus 1986] Dreyfus, Hubert L.; Dreyfus, Stuart E. Mind Over Machine: The
Power of Human Intuituition and Expertise in the Era of the Computer, with Tom
Athanasiou. New York: The Free Press, A Division of Macmillan, Inc., 1986.
Too often, computer enthusiasm leads to a simplistic view of human skill and
expertise.pxi Dreyfus and Dreyfus think it is a misguided effort(pxv) to attempt
to endow electronic circuits with skills involving judgement as well as calculation.
Each of us has, and uses every day, a power of intuitive intelligence(pxiv) that
cannot be captured in formal rules and therefore cannot be mechanized.
[Dreyfus 1992] Dreyfus, Hubert L. What Computers Still Cant Do: A Critique of
Artificial Reason. Cambridge, Massachusetts: The MIT Press, 1992. Organizing
and representing commonsense knowledge is incredibly difficult. Either a way of
representing and organizing everyday human know-how must be found, or AI will
be swamped by the welter of facts and beliefs that must be made explicit in order to
try to inform a disembodied, utterly alien computer about everyday human life.p3
Dreyfus mentions degrees of understanding but takes it away by saying that
understanding is part of an interrelated set of terms for talking about behavior such
as ask, answer, know, etc. some of which have an all-or-nothing
character.p8 Dreyfus points out the failures of a lot of researchers.

[Dyer 1983] Dyer, Michael George. In Depth Understanding: A Computer Model
of Integrated Processing for Narrative Comprehension. Cambridge, MA: The MIT
Press, 1983. Understanding a narrative in depth involves recognizing the moral
or point of a narrative. This is like characterizing the theme with an adage.
Thematic Abstraction Units arise when expectations fail. Associating themes with
adages (for instance: the pot calling the kettle black) allows recognition of themes
across contexts.
[Elliot 1981] Elliot, Donald. Lambs' Tales from Great Operas. Harvard,
Massachusetts: The Harvard Common Press, A Gambit Book, 1981. An
introduction to opera for those reluctant "to savor the toothsome, if occasionally
absurd, delights that lie waiting, "pvii "The March Hare, the bustle, the romantic
bellowings of an operatic diva of truly heroic proportions are absurd and
meaningless to those who believe that reality is of only one shape, "pxiv
Impressions, delivered to man through his senses, "come to him in the form of
symbols, for he can never really take into his mind the actuality of the things
symbolized, and whatever truth he sees or thinks he sees he develops himself,
hoping that his concepts somehow accord with a universality that is probably
forever beyond him."pxi To those who realize that truth is revealed in many forms,
opera "is a door, in short, that is surely closed to those whose concept of reality
always requires the objectively and drearily real, "pxiv
[Eysenck 1984] Eysenck, Michael W. A Handbook of Cognitive Psychology.
Hillsdale, New Jersey: Lawrence Erlbaum Associates, Publishers, 1984.
Virtually all those interested in perception, learning, memory, language, concept
formation, problem solving, or thinking call themselves cognitive
psychologists^ 1 This book covers a huge variety of topics: perception, attention,
memory structure and processes, imagery, language, cognitive development,
problem solving and reasoning and a few others. While processing is
substantially affected by the nature of presented stimuli, it is also affected crucially
by the individuals past experiences, expectations, and so on.p3
[Eysenck 1977] Eysenck, Michael W. Human Memory: Theory. Research and
Individual Differences. Oxford: Pergamon Press, 1977. Eysenck presents the ideas
of both the psychologists who are interested in the processes of human learning
and memory and the psychologists who are interested in individual differences.
There have been some researchers who have combined the two, among them
Spence, Spielberger, and Eysenck (pere). Eysenck has chapters on the processes
of memory and chapters on individual differences including arousal and memory,
anxiety and memory, age and memory, and intelligence and memory.
[Fetzer 1990] Fetzer, James H. Artificial Intelligience: Its Scope and Limits.
Dordrecht: Kluwer Academic Publishers, 1990. In the last chapter, Fetzer
compares and contrasts Winograd and Flores Understanding Computers and
Cognition (1986), Schank The Cognitive Computer (1984) and Minsky Why
People Think Computers Cant (1982). Winograd and Flores are critical of the
capacity of computers for cognition while Schank and Minsky support the view
though with qualifications that computers could become cognitive though it would

be different than human cognition. The discussion revolves around language,
symbol systems and semiotic systems and whos got what and what that means for
cognition. Humans are semiotic, computers are symbol systems. Semiotic
systems can possess minds, symbol systems cannot. p302
[Forsythe 1989] Forsythe, Richard, ed. Machine Learning: Principles and
Techniques. New York: Chapman and Hall, 1989. Essays that explore machine
learning; inductive learning, intelligence, creativity, learning and memory,
knowledge bases. Forsyth introduces the concepts of induction within which
modem efforts to computerize the act of induction can be understood. He explains
terminology and talks about what philosophers and psychologists think about
induction. He builds a framework for induction, a model for discussing its pieces:
parameters, attributes, language, training and testing sets, algorithms.
[Franklin 1995] Franklin, Stan. Artificial Minds. Cambridge, Massachusetts: A
Bradford Book, MIT Press, 1995. A synthesis of work in artificial intelligence
and natural sciences of the brain. A tour through the contemporary work being
done in AI, cognitive science, cognitive neuroscience, artificial neural networks,
artificial life, and robotics. Franklin rejects the distinction between mind and non-
mind in favor of a continuum from less to more mind, and for the central task of the
mind to be to choose the next action.
[Freedman 1994] Freedman, David. Brainmakers: How Scientists are Moving
Beyond Computers to Create a Rival to the Human Brain. New York: Simon &
Schuster, 1994. High-level intelligence includes commonsense reasoning,
problem solving, and creativity. There are also the capabilities of perception, motor
control, reflexes which are really difficult to reproduce on machines.
Neuroscientists have speculated that the processing of conscious information takes
up as little as a thousandth of the human brains computing power; most of the rest
goes into dealing with the lower-level aspects of survival. This suggests that AI
has been trying to construct a tower starting from the top.p20 Learners need
general knowledge to fall back on of they cant learn. General knowledge means
knowing things about chairs and counting and things to eat and animals and how
things work in the world and language to communicate all that. Thats a lot of stuff
to know and we dont really know how to cram it all into a computer yet.
[Froehlich 1984] Froehlich, Werner D. ed. Psychological Processes in Cognition
and Personality. Washington: Hemisphere Publishing Co., 1984. This book takes
stock of the variety of conceptualizations and methods applied to the study of
perceptual and cognitive processes. There are contributions from nineteen authors
about process-oriented research in psychology. The effect of individual
differences on the processes involved in the development of a percept has been little
studied. Francine A. Lastowski Individual Differences and Microgenesis p95.
[Gale 1992] Gale, Anthony and Michael W. Eysenck, eds. Handbook of
Individual Differences: Biological Perspectives. New York: John Wiley & Sons,
1992. This book, one of five in a series, "focuses on individual differences in
temperament, personality and intelligence and the ways in which individuals both

vary in the expression of general processes and reveal a failure of biosocial coping
mechanisms, "px Included are "consideration of genetic, biochemical,
electrocortical and peripheral mechanisms in the determination of individual
differences, "px
[Gardner 1993] Gardner, Howard. Creating Minds: An Anatomy of Creativity
Seen Through the Lives of Freud. Einstein. Picasso. Stravinsky. Eliot. Graham
and Gandhi. New York: BasicBooks, A Division of HarperCollins Publishers,
1993. A new approach to the study of human creative endeavors. There are three
foci for creative accomplishments: the individual talent, the domain in which they
do their work, and the group of knowledgeable people around them who judge the
quality of their work. There are seven figures studied in this book, seven areas of
human intelligence. Gardner hopes to illuminate the nature of their own particular,
often peculiar, intellectual capacities, personality configurations, social
arrangements, and creative agendas, struggles, and accomplishments.p6
[Gleick 1987] Gleick, James. Chaos: Making a New Science. New York: Penquin
Books, 1987. The modem study of chaos began with the creeping realization in
the 1960s that quite simple mathematical equations could model systems every bit
as violent as a waterfall. Tiny differences in input could quickly become
overwhelming differences in output.p8 Physicists and mathematicians look for
equilibrium, convergence. In chaos, there are little spikes of order, [fjleeting
bits of periodic behavior.p77 Chaos is ubiquitous; it is stable; it is
[Greenfield 1995] Greenfield, Susan A. Journey to the Centers of the Mind:
Toward a Science of Consciousness. New York: W.H. Freeman and Company,
1995. The purpose of the book is to harness what we know of brain chemistry
and brain electricity to help see how we might, one day, formulate the physical
basis of the phenomenological sensation of consciousness.px
[Guterl 1995] Guterl, Fred. Reinventing the PC. Discover Magazine, Vol.16,
No.9 September 1995, p42-47. Five ways of incorporating the computer into our
world invisibly: computers in your clothing that communicate with computers in
the environment, using the small electric fields generated by electronics to detect
movement of hands to manipulate things, every person would have their own video
channel for conferences or communication or companionship, lip-reading
computers increase the bandwidth of a computers speech input and decrease the
amount of errors in communication, software agents that complete tedious or
difficult tasks like sifting through e-mail messages or scheduling appointments.
[Hitchhiker 1995] Hitchhikers Guide to Artificial Intelligence. AI Expert
Magazine, San Francisco, CA: Miller Freeman Inc., 1995. Articles on AI: where it
is hidden in current systems, explanations of concepts of AI like neural networks,
case-based reasoning, fuzzy logic.

[Hoc 1995] Hoc, Jean-Michel; Cacciabue, Pietro C.; Hollnagel, Erik eds.
Expertise and Technology: Cognition and Human-Computer Cooperation.
Hillsdale, NJ: Lawrence Erlbaum Associates, Publishers, 1995. This book is a
collection of essays related to expertise and expert systems. Amain concern of
this book is the cooperation between people and machines.p7 Topics include man-
machine interaction, human error and reasoning bias, ergonomic system design,
cognitive simulation and theories of expert cognition. We are not really experts if
we only are able to use the system when everything works as it should, but unable
to do so if something go wrong.[sic] p2
[Hoeg 1994] Hoeg, Peter. Borderliners. Trans. Barbara Haveland. New York:
Dell Publishing, a division of Bantam Doubleday, 1994. This novel talks of time
and the different perceptions of space and time. An example used is a spiders web
that is only thirty inches across.p250 The web is an extension of the spiders
senses. In every movement of the web the spider can tell how big something it is,
and how far off. The web can not help the spider sense the whole world, just the
part the web is built for; direction, distance, an approximate size and weight. It
misses the colors, the smells and sounds. Even if the web were bigger, it would
still miss these parts of the world. Eventually, if even bigger, the web would
collapse under its own weight. It might bring down the trees it was connected to, it
would change the world. And the spider would not be able to cope with the
number of signals, the number of insects or other things coming into contact with
its web. Then the abnormally large web and what it brought back would come
into conflict with the essence of the spider, with its nature.p252 Hoeg compares
this to Mans web, his exploration of the world. Mans web changes the world.
We have built submarines and computers with what we have learned.
[Hofstadter 1985] Hofstadter, Douglas R. Metamagical Themas: Questing for the
Essence of Mind and Pattern. Toronto: Bantam Books, 1985. Hofstadter covers
lots of different subjects in this book: self-reference, patterns of music, the
Prisoners Dilemma, Rubiks cube, the Turing test, strange attractors lots of math
stuff. [L]ife itself is a mixture of things of many sorts, little and big, light and
serious, frivolous and formidable, and Metamagical Themas reflects that
complexity.pxxvii Many people turn off when faced with issues that are too
big. Hofstadter tries to make these big issues graspable and fascinating, enticing
people with the beauty of clarity, simplicity, precision, elegance, balance,
symmetry. The artistic qualities central to science and life are the things Douglas
Hofstadter explores and celebrates in Metamagical Themas.
[Jung 1961] Jung, C.G. Memories. Dreams. Reflections. New York: Vintage
Books, A Division of Random House, Inc., 1961. Recorded and Edited by Aniela
Jaffe, Translated by Richard and Clara Winston. A sort-of autobiography. Myth
is more individual and expresses life more precisely that does science. Science
works with concepts of averages which are far too general to do justice to the
subjective variety of an individual life.p3 Jung tells his personal myth. The
only question is whether what I tell is my fable, my truth.p3

[Kolodner 1993] Kolodner, Janet. Case-Based Reasoning. San Mateo, California:
Morgan Kaufmann Publishers, Inc., 1993. Kolodner proposes case-based
reasoning as a means to an intelligent machine. [L]eaming happens as part of the
process of integrating a new case into memory. pxiv Case-based reasoning is
problem-solving. It provides a way to deal with uncertainties. This book
represents the state of the art in case-based reasoning.pxiv It provides current
answers to questions: how to represent knowledge in cases, how to index cases,
how to implement retrieval processes for efficiency, how to adapt old situations to
fit new ones, etc.
[Kossowska 1994] Kossowska, Malgorzata and Edward Necka. Do it Your Own
Way: Cognitive Strategies, Intelligence and Personality. Personality and
Individual Differences Journal, Vol 16, No 1 January 1994, p33-46. A verbal
analogy task is used to investigate cognitive strategies. Analogy is a basic pattern
of human inductive reasoning. p35
[Kreitler 1990] Kreitler, Shulamith and Hans Kreitler. The Cognitive Foundations
of Personality Traits: Emotions. Personality, and Psychotherapy. New York:
Plenum Press, 1990. A new definition of traits, an explanation of trait dynamics in
terms of meaning, a specification of trait characteristics that allows us to compare
traits or to determine whether a given personality disposition is a trait or not. The
Kreitlers apply their psychosemantic theory of cognition to the trait concept. A
meaning approach to personality traits.
[Lebowitz 1980] Lebowitz, Michael. Generalization and Memory in an Integrated
Understanding System. Dissertation presented to the faculty of the graduate school
of Yale University for degree of doctor of philosophy. University Microfilms
International, December 1980, Ann Arbor, Michigan. Lebowitz developes an
Integrated Partial Parser which reads in stories taken in from newspapers and the
UPI news wire about international terrorism, parses them and incorporates them
into memory. IPP can build generalizations and uses these generalizations to
understand future stories and incorporate them into a dynamic memory. It is a
robust and efficient understanding system.(abstract).
[McLeish 1993] McLeish, Kenneth, ed. Bloomsbury Guide to Human Thought:
Ideas that Shaped the World. London: Bloomsbury Publishing Limited, 1993. A
collection of 2500 ideas that the contributors think have had the greatest impact and
influence on our development through the ages. The subject of this book, as stated
in the introduction, is what we think about while we are surviving. Humans
observe some phenomenon and begin asking questions about it. These are the
start of an ever-expanding web of thought which spreads both in our minds and in
those of everyone we share it with. These edifices of thought are built not solely on
the millions of facts we have inside our heads, but also on the ideas and knowledge
of other people, of our contemporaries, our predecessors and our progeny.pviii
Alphabetical arranged shorts on subjects like chaos theory, personal construct
theory, personhood, mind, memory, computing, consciousness, emotivisim,
ethical relativism, endocrinology, fa?adism, faith, farce, fascism, Galois theory,

game theory, Germ Layer theory, Godel's Incompleteness theorem, idealism,
induction, kitsch, metals and alloys, phylogeny, robots and rock music, etc.
[Minsky 1982] Minsky, Marvin Why People Think Computers Cant. TheAI
Magazine Vol.3, No.4 Fall 1982, p3-15. When we talk about computers thinking,
and being conscious and being self-aware, we are talking about them being like us
and maybe thats not what we should be doing. Minsky feels that perhaps when
we define these terms we are putting limits on them that shouldnt be there, that we
dont know how to define these terms anyway. Minsky feels that a computer could
be self-aware in its own way, or conscious in its own way. No one knows if
machines will learn to learn because we simply do not know enough today of
either men or possible machines.pl5
[Minsky 1985] Minsky, Marvin The Society of the Mind. New York: Simon &
Schuster, 1985. Lots of little essays on how the minds work. Each mind is made
of smaller processes called agents. Each agent only does a certain very special
thing. The agents are joined together in tangled web-like societies that lead to
[Murphy 1993] Murphy, Cullen. The Lay of the Language. The Atlantic
Monthly, Vol. 275, No. 5 May 1993: p20-22. Language evolves, linguistic rules
change, specifically the distinction between the verbs lay and lie. Lay has been
laying siege to lie with growing success.
[Murray 1938] Murray, Henry and the Harvard Psychological Clinic. Explorations
in Personality: A Clinical and Experimental Study of Fifty Men of College Age.
New York: Oxford University Press, 1938. All the experimenters studied the same
series of individuals with the same concepts actively in mind, and then to report
their findings and collaborate in accomplishing the common purpose of formulating
the personality of each subject. Psychology must construct a scheme of concepts
for portraying the entire course of individual development, and thus provide a
framework into which any single episodenatural or experimentalmay be
fitted.p4 Personology encompasses psycho-analysis (Freud), analytical
psychology (Jung), individual psychology (Adler). The intention of this book
was to construct a theory of personality; to devise techniques for getting at some of
the more important attributes of personality; and by a study of lives of many
individuals to discover basic facts of personality.
[Myer 1992] Myer, David G. Psychology. Third Edition. New York: Worth
Publishers, 1992. This is a psychology textbook. This book covers foundations
of psychology, development of a life, motivation, learning, social behavior,
personality, etc., the principles and processes of psychology, pxvii
[Nass 1995] Nass, Clifford; Moon, Youngme; Fogg, FJ; Reeves, Byron; Dryer,
Chris. Can Computer Personalities Be Human Personalities? CHI 95 Mosaic of
Creativity, May 7-11, 1995, pp. 228-229. This short paper demonstrates that (1)
computer personalities can be easily created using a minimal set of cues, and (2)
that people will respond to these personalities in the same way they would respond

to similar human personalities.p228 It has become a commonplace idea that
believable agents must have personalities.p228 This paper states that there are two
major interpersonal personality dimensions: dominance/submissiveness and
affiliation (warmth/hostility). [T]he present study presents evidence that human-
computer interaction is fundamentally social and interpersonal.p229
[Natarajan 1991] Natarajan, Balas K. Machine Learning: A Theoretical Approach.
San Mateo, CA: Morgan Kaufmann Publishers, Inc., 1991. This book involves
rigorous mathematical analyses of programs that learn to imitate a class of
input/output behaviors. There are two classes of these input/output behavior
learning programs: concept recognition and learning functional evaluation (getting a
value from another value because of some relation). Natarajan focuses on Golds
paradigm, which is identification in the limit, which is an algorithm that tries to
figure out the set when presented with a stream of elements of the set; and the
paradigm of Valiant which is the probably approximately correct model of
learning (PAC). PAC characterizes streams of examples by their statistical
properties and measures the error in the hypothesis of the learning algorithm with
respect to the same statistical properties. PAC is a measure of complexity of Golds
[Nollman 1990] Nollman, Jim. Dolphin Dreamtime: The Art and Science of
Interspecies Communication. New York: Bantam Books, 1990. Jim Nollman
communicates with animals using music. He communicates with whales, dolphins,
orcas, turkeys, buffalo and other animals. This book addresses the intelligence of
animals as a natural wisdom of the interconnectedness of all life. He speaks of
particular animals as individuals with an individual animal consciousness. Nollman
talks of the isolation of the western culture, the separateness of humans from the
rest of the Earth. He compares those aspects of western culture that separate us
from dolphins and whales with the aspects of other cultures, like the American
Indian or the Australian Aboriginies, that engage the human as a part of a bigger
part, of a community of animals, or a community of life. He says were missing
out by excluding ourselves from this community.
[Norman 1993] Norman, Donald A. Things That Make Us Smart: Defending
Human Attributes in the Age of the Machine. Reading, Massachusetts: Addison-
Wesley Publishing Company, 1993. |I]t is primarily our social structures that
determine both the direction that technology takes and its impact upon our lives.
Thats why I call it a social problem, and thats what this book is about.(pxii)
Norman says that we serve technology. He says that technology should serve us;
that we need to reverse the trend of a machine-centered point of view.
Technological artifacts are tools that can make us smart. They also limit the way in
which we can use them and learn from them. They shape the way we think and, at
times, what we value. [T]his book illustrates the degree to which we can be
helped, hindered, or even manipulated by various representations of
information.(book jacket)

[Penrose 1989] Penrose, Roger. The Emperors New Mind: Concerning
Computers. Minds, and the Laws of Physics. New York: Oxford University
Press, 1989. Penrose believes there is more to creating an artificial mind than mere
computations. He believes that quantum mechanics and perhaps even deeper
mysteries are essential for the operation of the mind. In this book, Penrose
discusses complex numbers, complexity theory, Godel undecidability, paradoxes
of quantum mechanics, formal systems, consciousness, the possibility of matter-
transmission, etc... Perhaps when computations become extraordinarily
complicated they can begin to take on the more poetic or subjective qualities that we
associate with the term mind. Yet it is hard to avoid an uncomfortable feeling that
there must always be something missing from such a picture. p447
[Penrose. 1994] Penrose, Roger. Shadows of the Mind: A Search for the Missing
Science of Consciousness. Oxford: Oxford University Press, 1994. Penrose
discusses his belief that consciousness is dependent upon something that is not
directly computational. There is something in awareness that eludes both top-
down, algorithmic and bottom-up, leam-by-experience computational simulations.
A thorough examination of Godels theorem is central to his argument. The second
part of his argument includes some puzzles and paradoxes of physics and biology,
and the mysteries of quantum physics, presuming that a program is written that is
complex enough to have a chaotic dynamic evolution that could manifest itself as an
[Pinker 1994] Pinker, Steven. The Language Instinct. New York: William
Morrow and Company, Inc., 1994. Pinkers thesis is that language is an instinct to
humans. Apes dont have that instinct and can never truly grasp a human natural
language. He says that there is a window in childhood where children must be
surrounded by and use their language. Adults who have missed this window have
difficulty using the language with any personal evolution. The infiniteness of
language composition has, in the end, to come from instinct.
[Sayre 1963] Sayre, Kenneth M. The Modeling of Mind. Edited by K. Sayre and
Frederick J. Crosson, New York: Simon and Schuster, 1963. Early essays on the
possibility of creating an artificial mind.
[Schank 1984] Schank,Roger C. with Peter G. Childers. The Cognitive Computer:
On Language. Learning, and Artificial Intelligence. Reading, Massachusetts:
Addison-Wesley Publishing Company, Inc., 1984. Schank attempts to set the
computer in perspective, to see it as a machine, a machine with fantastic
possibilities.px He considers computer understanding in light of human
understanding. He addresses three questions: What do we have to know about
computers in order to live in a world that is full of them?, What can we learn
about what it means to be intelligent through our development of computers that can
understand? and How will intelligent computers affect the world we live in?
This book is an introduction to AI according to Schank.

[Schank 1991] Schank, Roger C. The Connoisseurs Guide to the Mind. New
York: Summit Books, 1991. Learning is the accumulation and indexing of cases
and thinking is the finding and consideration of an old case to use for decision-
making about a new case. To make thinking beings, we must encourage
explanation, exploration, generalization, and case accumulation.p248 An
interesting exploration of the authors learning processes about the expectations of
culturally different eating establishments.
[Schank 1982] Schank, Roger C. Dynamic Memory. New York: Press Syndicate
of the University of Cambridge, 1982. People have dynamic memories. This book
explores the processes and structures a computer program might have to develop a
dynamic memory. [T]here is a kind of selection process at work that picks some
memories for special treatment by retaining them for long-term memory. It is the
nature of this selection mechanism and the nature of the types of memories that do
not fade quickly that is the key problem before us.pl3 [RJeminding occurs when
we have found the most appropriate structure in memory that will help in
processing a new input.p21 Understanding means being reminded of the closest
previously experienced phenomenon.p24 [M]emory structures for storage and
processing structures for analysis of inputs are exactly the same structures.p25
All things are in some ways similar and in some ways different from all other
things. So it is with memory structures. That is how a memory can come to be
dynamic.p223 Everything is connected to everything else in memory.p223
[Schank 1986] Schank, Roger C. Explanation Patterns: Understanding
Mechanically and Creatively. Hillsdale, New Jersey: Lawrence Erlbaum
Associates, Publishers, 1986. If we can find a set of processes that machines can
slavishly follow and if by so doing, these machines can come up with creative
thoughts, what would that tell us about human beings?pv This book discusses the
mechanisms of mind, to define an apparatus that underlies our ability to think.
Schanks premise is that the simplest of mental tasks and the most complex of
mental tasks are all interrelated. Memory search, learning, creativity are part of the
same process. Understanding is a creative act in itself.
[Schank 1994} Schank, Roger C.;Kass, Alex; Riesbeck, Christopher K. Inside
Case-based Explanation. Hillsdale, New Jersey: Lawrence Erlbaum Associates,
Publishers, 1994. Part of the Artificial Intelligence Series. Explanation is
understanding. The difficult part of understanding [...] is developing creative
hypotheses about why the events that the story describes took place.pxv
Understanding is not an all-or-nothing affair.p6 We are reminded of an
experience because the structures we use to process new experiences are the same
structures we use to organize memory. We pass through old memories while
processing new input. Finding the right one is one of the things we mean by
understanding.p8 The more that goals, beliefs, and prior experiences and
memories are shared, the more complete the level of understanding that can take

[Singer 1993] Singer, Jefferson A. and Peter Salovey. The Remembered Self:
Emotion and Memory in Personality. New York: The Free Press A Division of
Macmillan, Inc., 1993. Memories are an important window into ones life story,
revealing characteristic moods, motives, and thinking patterns. Maintenance of
equilibrium and reduction of discrepancies between desires and present
circumstances is a huge goal of the individual. In brief, we suggest that one aspect
of what makes an individual unique, different from all others, is his or her set of
memories and, in particular, what we call self-defining memories.p3 Memory
content can be influenced by mood states, attentional processes, and biases of the
[Smith 1984] Smith, Anthony. The Mind. New York: The Viking Press, 1984.
Smith talks about the anatomy, growth, and physiology of the brain; how the
autonomic and willed impulses work on the whole body through the nervous and
glandular systems. He discusses consciousness, memory, senses, right and left
dominance, abnormal abilities, and brain damage.
[Spear 1994] Spear, Norman E. and David C. Riccio. Memory: Phenomena and
Principles. Boston: Allyn and Bacon, 1994. This book takes the psychobiological
view of the memory process. Memory is divided into two phases; the acquisition
phase (learning, storage) and the later phase of retrieval and expression. Animals
as well as humans are studied in this study of memory. The authors emphasize
the behavioral characteristics of memory: how forgetting proceeds in various
circumstances; what factors or events induce forgetting or affect the speed or
amount of forgetting; and how forgetting can be prevented or alleviated p2
[Staniszewski 1995] Staniszewski, Mary Anne. Believing is Seeing: Creating the
Culture of Art. New York: Penguin Books, 1995. What is Art? How do we as a
world culture relate to art? The evolution of art as a means of expressing ourselves
in our culture. Artists prepare the mind and the spirit for new ideas-new ways of
seeing.p289 Art is a reflection of ourselves. What is at issue now is reflected in
[Thornton 1992] Thornton, C.J. Techniques in Computational Learning: An
Introduction. London: Chapman and Hall Computing, 1992. The question of how
particular learning algorithms actually work is addressed in the main body of the
book.pl2 There are three areas of computational learning: machine learning
(symbolic learning techniques), connectionism (neural networks), genetic
algorithms. This book focuses on the first two. A learners goal is to produce a
representation of the target mapping when provided with examples that show what
output should be for a given input. Thornton compares and analyzes learning
algorithms in two of the three areas of computational learning.
[Waldrop 1987] Waldrop, M. Mitchell. Man-Made Minds: The Promise of
Artificial Intelligence. Walker Publishing Company, Inc. 1987. Artificial
intelligence is the art of making computers do smart things. This is a book about
the meaning and promise of AI. p6 Intelligent computers, with capabilities rivaling
humans, means discussing other political questions about their roles in society, like

how much responsibility and rights will be given to these computers. This book is
a discussion of questions like: Can a machine feel? be aware? The pursuit of
artificial intelligence in general has taught us a great deal about intelligence itself.p8
[Weiss 1991] Weiss, Sholom M. and Casimir A. Kulikowski. Computer Systems
that Learn: Classification and Prediction Methods from Statistics. Neural Nets.
Machine Learning, and Expert Systems. San Mateo, CA: Morgan Kaufmann
Publishers, Inc., 1991. This book is meant as a practical guide to the application
of classification learning systems .(p.v) The learning methods are divided into
three groups: statistical pattern recognition, which has been studied the most; back-
propagation neural networks and machine learning methods, which centers around
decision trees. How to estimate the true performance of a learning method, and
which technique is best in which situation are two practical considerations when
choosing a learning system. In the last chapter, expert systems are contrasted with
learning systems and the potential advantages of combining them are discussed.
[Wrench 1969] Wrench, David F. Psychology: A Social Approach. New York:
McGraw-Hill Book Company, 1969. The various areas of psychology that can
contribute to our understanding of human behavior on a social level is the major
subject of this book. There are three major themes; man as an information
processor, man as a motivated organism, and man in relation to social groups.
Personality is viewed as made up of conflicting processes rather than as being
unitary.pl82 Material is more easily learned if it can be made more
[Wright 1995] Wright, Robert. Can Machines Think?. Time. Vol. 147 No. 13
(1996): p50-56. This article was provoked by the chess match between Garry
Kasparov, the world champion, and Deep Blue, a computer chess program. A
discussion of what Chalmer describes as easy programming problems, like
playing chess, and hard programming problems, like making small talk, or
playing Trivial Pursuit. There is also a discussion of consciousness the
existence of pleasure and pain, love and grief [...] a fairly central source of lifes
meaning p54 Materialists feel that there is no such thing as soul or mind or
consciousness. It is all a culmination of the physical processes of the brain. Then
there are the mysterians who feel there is an extraness to consciousness; the
physical processes can explain behavior and the feelings, subjectivity is extra.
Chalmer says maybe consciousness is a nonphysical property of the universe
vaguely comparable to physical properties like mass or space or time. And maybe,
by some law of the universe, consciousness accompanies certain configuations of
information, such as brains.p56 In this view, a thermostat might have
[Wyer 1989] Wyer, Robert S. Jr and Thomas K. Srull. Memory and Cognition in
its Social Context. Hillsdale, New Jersey: Lawrence Erlbaum Associates,
Publishers, 1989. Intended to be a flow diagram of the stages of processing, this
book is really a new model of social information processing (and specifically, how
we model other people in our minds). Encoding, organizational, storage, retrieval,
and inference processes are covered. There is offered new conceptualizations of the

representations formed from person and event information and the way these are
used to make judgments. The role of affect in information processing and the
content and structure of self-knowledge are also covered.
[Zucker et al. 1992] Zucker, Robert A.; Rabin, A.I.; Aronoff, Joel; Frank, Susan,
eds. Personality Structure in the Life Course: Essays on Personologv in the
Murray Tradition. New York: Springer Publishing Company, 1992. This is a
publication of a periodic lecture series on personality whose goal is to stimulate the
type of theoretical exploration that is not usually encouraged in more traditional
forums.(pvii) This book focuses on personality over a lifetime.