McCarthy: The Well Designed Child

From: Sparks Simon (
Date: Wed Apr 25 2001 - 16:55:11 BST

Subject: McCarthy: The Well Designed Child

>The inate mental structure that equips a child to interact succesfully
>with the world includes more than a universal grammar. The world
>itself has structures, and nature has evolved brains with ways of
>recognizing them and representing information about them…What a child
>learns about the world is based on its inate mantal structure.

In my judgement, most of what is presented in this paper is a
sufficiently sound basis for considering the required mental
structure of a human child, but I consider it incomplete with regard to
one particular aspect.
I propose that an essential feature of what it means to be human, a
feature that I believe to be a constituent of a child's inate mental
structure, is the ability to want.
>From our lifelong ambitions down to our daily routine, our actions
follow our intentions; we do what we want, and if we don't we feel
unsatisfied or uncomfortable with our lives.
I believe that want leads to emotion. We are happy, sad, worried,
relieved, all because of how a situation affects the outcome of
something that we want.
Beyond the proposed components of a human's inate mental structure
that enable a child to learn from its environment, as discussed in
McCarthy's paper, I would conjecture that the drive of want actually
causes the child to interact with the environment.

>This article concerns designing adequate mental structures including a
>language of thought. This design stance applies to designing robots,
>but we also hope it will help understand universal human mental

If a robot never wants for anything, I don't think it will ever pass
the T3 Turing Test. It is a common argument that robots could never
exhibit emotion or emotive responses to situations, that they could
never have feeling.
I would argue that an understanding of how humans want for things and
how this affects their constitution would help a great deal in
designing a sucessful T3 candidate.

>Instead of building babies as Cartesian philosophers taking nothing
>but their sensations for granted, evolution produced babies with inate
>prejudices that correspond to facts about the world and babies'
>positions in it. Learning starts from these prejudices. What is the
>world like, and what are these instinctive prejudices?

It is hard to imagine that an individual inately wants for anything in
particular from this world but I would imagine that instinctively an
individual wants for personal well-being in whatever sense the
individual defines well-being. I would also say that, as the individual
develops an understanding of the world, more specific wants are
established to support the individual's definition of well-being.

>Animal behavior, including human intelligence, evolved to survive and
>suceed in this complex, partially observable and very slightly
>controllable world. The main features of this world have existed for
>several billion years and should not have to be learned anew by each
>person or animal.

A great benefit to the evolution of animals, humans in particular, has been
the ability to pass knowledge down through the generations in order to
give successive generations an advantage over their predecessors. Much
knowledge has arguably been 'hard-wired' into animals as natural
survival instincts or as environmental adaptations.
In humans in particular, it can be argued that certain codes of
conduct, certain ethics are inate in our evolved state. I'm sure, for
example, that a person would generally disagree with the idea of
killing a fellow person even having never been advised on the subject.

>[What the World is like]
>APPEARANCE AND REALITY: some properties of the world are stable even
>though their appearences change. Objects last from seconds to
>centuries, while appearences change in fractions of a second.
>Therefore, humans, animals and robots are better off representing
>information about objects in so far as it can be obtained by
>observation and inferred from past experience or is inate.

Building up mental representations of worldly objects is an extremely
importatnt part of being an interactive participant. To be able to
recall past experiences with or inate attitudes towards something is
invaluable for classifying one's environment and determining one's
conduct therin.
The reality of objects can be subjective though.That one can have
feelings about objects is an important factor in classifying them and
one's interaction with them.
It is essential that people are cautious of road traffic for example.
It is also a valuable asset for a person, or indeed a robot subjected
to the Turing Test, to be able to enjoy or dislike, for example,
particular objects or actions.
Feelings towards objects and situations are developed over repeated
encounters with them and infered experience, or indeed inate motives,
determine what one wants from them, if anything. Knowledge of what one
wants from objects or situations gives guidence towards how one should
interact with them.

>THINGS OF INTEREST: Some aspects of the world are relevant to an
>animal's or person's survival or prosperity, and others are not.
>However, notice that human and animal curiosity concerns many aspects
>of the world not related to survival.

I think that to be able to develop an interest in something new or
previously unexperienced, and to nurture some sort of feeling about it,
gives one a sense of purpose. To want new and changing things, and to
act on those wants, is to maintain an active and varied existence,
the alternative being a repetative, predictable existance.
Notice how current robot imlementations follow the second pattern much
More closely than the first.

>KINDS OF OBJECTS: Objects have kinds, and objects of the same kind
>have properties associated with the kind…
>…RELATIONS: Objects not only have individual properties and belong to
>kinds, but different objects and kinds have relations with one another…
>…ABSTRACTIONS: Kinds belong to higher kinds. Red is a color and color
>is a quality. Tis is a fact of logic rather than about the physical
>world, but its usefullness is dependant on objects being naturally
>grouped into kinds rather than being all completely different.

Defining the content of the World in terms of Objects, Relations and
Abstractions clearly implies that an Object Oriented approach to robot
design and child psycology analysis is a useful one.
Making use of an Object Oriented approach in implementing some sort of
emotional value to interacting with the World may also proove to be
I would suggest that as a child learns about its environment, it
attaches emotional ties to different objects or group of objects. A
child may like looking a bright, colourful objects for example and so
associates the enjoyment of looking at bright colous with Objects that
share such properties. It is quite possible for a robot to develop
properties for internal representetions of objects that signify
"emotional" qualities, i.e. properties of objects, or classes of
objects, that would facilitate or hinder the process of doing what the
robot wants.

>[Human mental characteristics]
>EVOLVED FROM ANIMALS: The human was not designed from scratch. All our
>capabilities are elaborations of those present in an animal…Human pain
>is a far more complex phenomenon than an inventor would design or a
>philosopher intuit by introspection.

Similarly, as human intellect has developed, human emotion has evolved.
>From primal attraction and its antithesis, human want has produced a
range of different complex feelings for the different situations humans
find themselves in.
The refinement of human emotion as a more precise classifier of objects
and actions, I believe, goes hand in hand with the human notion of

>CURIOSITY: Humans and animals are curious about the world. However,
>this curiosity is focussed. Just how it is focussed isn't obvious.

As mentioned above, being curious about and developing an interest in
new things is a valuable asset to aid learning.
I would suggest that as a human or, to some extent an animal, develops
an understanding of the world and associates different wants and
emotions to objects and situations in the world, curiosity becomes
focussed around the subjects of more favourable emotion.
If someone develops a passion for Artificial Intelligence, for example,
that person will want do discover more about the field and will be led
by such an emotion to persue such investigation.

>SUPPOSED TO DO:Children and adults have a concept that in a particular
>kind of situation there are actions "that one is supposed to do". One
>learns what one is supposed to do and does it without reinforcement of
>the specific kind of response…The race was reinforced - or maybe it
>was our mamallian ancestors.

Again, the idea of inherited knowledge is an important aid to
accelerated knowledge aquisition, more so in humans where recorded
knowledge can be aquired without first-hand experience.
Communities tend to develop popular opinions or emotions about certain
subjects. Such consensus emotion leads to conventional actions or
reactions regarding those subjects. Such conventions are adopted
without first-hand experience, or at least used as a basis for
developing personal feeling towards the subject.

>[What abilities could usefully be inate?]
>FOCUSSED CURIOSITY: In the Shannon sense, there is just as much
>information to be obtained from the pattern of saw marks on the boards
>of my office wall as there is about what is available for lunch or
>what can be obtained by research on artificial intelligence. Curiosity
>needs to be focussed on what is potentially relevant to the baby or
>robot.Notice that human curiosity, as it ought to be, is quite broad -
>but it is also selective. Part of the answer is that curiosity is
>focussed on getting more information about kinds of object that have
>been identified.

Curiosity too could possibly be led by emotion. A human is likely to
investigate matters that will potentially offer favourable emotion, or,
a human will be curious about the things that he/she will want to know
A simplistic example would be a person researching artificial
intelligence because of an enjoyment of the field.

>GRAMMAR OF GOAL REGRESSION: The recognition that a goal is achievable
>because it is either already achieved or all the preconditions of an
>action that achieves it are achievable.

The main point I wish to make in this discourse is that I believe that
emotion comes from having goals to achieve and reacting to the
process of achieving or trying to achieve those goals. From survival
instincts to constructive, or even destructive intentions, goals
initiate emotional reactions to their progress or lack thereof.

French psycologist F. Paulhan and American philosopher John Dewey are
the most eminent ancestors of the suggestion that emotions, both
positive and negative, have as their antecedents some discrepency or
conflict between the state of the world and the expectations which the
individual brings to the situation.
My intention here is to suggest that, in designing a Turing
indistinguishable robot, to have expectations or wants or goals and to
be able to develop new and different ones as a result of learning from
the surrounding environment is an essential feature that must be

>[The well designed logical robot child]
>Ever since the 1950s, people have suggested that the easy way to
>achieve artificial intelligence is to build an artficial baby and have
>it learn from experience. Actual attempts to do this have always
>failed, and I think this is because they were based on the Cartesian
>baby model.

I agree. To base everything on sensation alone, with no inate want for
anything, with no purpose, could at best only result in a "lifeless"
substitute for a Turing indistinguishable candidate.

This archive was generated by hypermail 2.1.4 : Tue Sep 24 2002 - 18:37:30 BST