School of Computer Science THE UNIVERSITY OF BIRMINGHAM CN-CR Ghost Machine

_____________________________________________________________________________________

This file started as some thoughts on
BBC Horizon 3rd July 2013: "What Makes us Human?"
Presented by Alice Roberts
http://www.bbc.co.uk/programmes/b036mrrj

However, it has now been expanded to give an extended account of some related
research here in Birmingham, so the file name (web address) has been changed.
_____________________________________________________________________________________

Pre-configured vs Meta-configured competences,
Altricial vs Precocial Species
and why humans are apparently born premature.

(DRAFT: Liable to change)

Aaron Sloman
School of Computer Science, University of Birmingham.
(Philosopher in a Computer Science department)

Collaborating with Jackie Chappell
School of Biosciences at the University of Birmingham.
(Who has not yet checked this web page.)

Installed: 6 Jul 2013
Last updated: 11 Jul 2013; 4 Aug 2013
This paper is
http://www.cs.bham.ac.uk/research/projects/cogaff/misc/pre-meta-config.html
(Short link temporarily broken http://tinyurl.com/AltPrecComp)
A PDF version may be added later.

A partial index of discussion notes is in
http://www.cs.bham.ac.uk/research/projects/cogaff/misc/AREADME.html

Some reflections on: "What Makes us Human?"
Presented by Alice Roberts
BBC Horizon 3rd July 2013
http://www.bbc.co.uk/programmes/b036mrrj

On Wed 3rd July, Alice Roberts presented an interesting BBC Horizon Programme, about
some differences between humans and other species. In particular human infants are
born in a very underdeveloped state compared with many other species, and the size of
a normal human infant skull causes great difficulties, and risks, for mothers and infants.
There were many well presented, interesting details. Here I want to place that work
in a broader research context, shifting the emphasis from differences between humans
and other species, to differences between ways in which genomes can influence
features of organisms.

It is well known that some abilities of animals are to a large degree pre-programmed
in the genome whereas others are results of learning. For example, young humans, like
other mammals, don't have to learn to suck, although their sucking may become more
skillful with practice and muscular development.

On the other hand, a puppy learns where its food is normally placed, and a child has
to learn to crawl, and then walk, In those cases, the genes provide learning abilities.
and the environment influences what is learnt.

That simple distinction between what is pre-programmed and what is learnt does not do
justice to the variety of relationships between genes and their effects.

Human children can learn to speak, and what language they speak will depend very much
on their environment, but mechanisms of learning to speak and the mechanisms of
learning to walk are different. Many animals can learn to walk, run climb, find food,
and much else, without being able to learn to talk. A sketch of a theory about the
different ways in which a genome can produce learning, will be presented below, after
introducing some of the themes of the Horizon programme.

The Obstetric dilemma
http://en.wikipedia.org/wiki/Obstetrical_dilemma

The transition in human evolution to walking upright has long been thought to
require a decreased size of the birth-canal but the evolution of increased intelligence
required a larger brain, which led to evolution of a larger pelvic gap to avoid risks
to mother and child during birth.

For many years the question whether evolution could have reduced those difficulties
and risks of childbirth by enlarging the pelvic aperture had been answered by the
suggestion that a wider pelvis interferes with locomotion on two legs.

A wider pelvic gap might have allowed babies to grow larger in the safety of the
womb. They would then be born in a more mature state, possibly less vulnerable
to dangers in the environment

A very old suggested explanation for the early, or even premature, birth of humans,
was that it avoided the need for a larger pelvic gap, which would have reduced the
efficiency of two-legged locomotion.

For example wikipedia (on 10 Jul 2013) includes
http://en.wikipedia.org/wiki/Human_pelvis
Modern humans are to a large extent characterized by bipedal locomotion and
large brains. Because the pelvis is vital to both locomotion and childbirth,
natural selection has been confronted by two conflicting demands: a wide birth
canal and locomotion efficiency, a conflict referred to as the "obstetrical
dilemma". The female pelvis has evolved to its maximum width for childbirth - a
wider pelvis would make women unable to walk. In contrast, human male pelves are
not constrained by the need to give birth and therefore are optimized for
bipedal locomotion. ...
The Horizon programme reported recent research that challenges this idea, using
experiments on human walking and running that showed that women can perform as
efficiently as men, despite their wider pelvis.

According to the programme, the main factor controlling the time of birth is the
ability of the placenta to supply sufficient energy for the growing child. (For more
details watch the horizon programme before it becomes inaccessible -- unless someone
copies it to Youtube!).

Related research at the University of Birmingham (UK):
Species differences vs competence differences.

Alice Roberts is Professor of public engagement in science at the University of
Birmingham, a job she does very well. This web page summarises some research related
to the topic of the Horizon program, done in Birmingham, though not widely known.

Many years ago, as a philosopher working in Artificial Intelligence, trying to
understand the evolution of mathematical abilities in humans, I concluded that the
conjecture that humans were born with such under-developed brains because of the
dangers a larger pelvic gap would pose to mothers must be wrong, partly because
evolution was able to provide much larger brains than human brains, and partly
because there seemed to be a deeper explanation for humans being born "prematurely",
namely that development of the brain during interaction with the environment might
lead to greater sophistication in adults, for example more sophisticated spatial
reasoning capabilities used by mathematicians and engineers. My hope was that by
modelling that process in 'baby robots' we could provide new answers to old questions
about the nature of human intelligence, especially mathematical intelligence.

This was my clumsy formulation at a conference on diagrammatic reasoning around 1998,
whose proceedings were later published in a book, before biologist Jackie Chappell came
to Birmingham and helped with development of the theory:
"Other animals may have much simpler qualia, especially precocial species born or
hatched with genetically formed visual mechanisms ready for use, e.g. chickens,
deer, horses. Altricial species e.g. birds of prey, hunting or tree-climbing animals
and humans start off more helpless and grow their brains while interacting with
the environment. Perhaps this 'bootstrapping' produces a much richer grasp of
structure and motion than can easily be encoded in genes. (Contrast this with
the popular opinion that humans are born so immature because their skulls would
otherwise be too big to pass through a human pelvis. Elephants manage, so that
can't be all there is to it.)"

in Aaron Sloman, "Diagrams in the mind",
in Diagrammatic Representation and Reasoning, (Book based on the 1998 conference)
Eds. M. Anderson, B. Meyer and P. Olivier, Springer-Verlag, Berlin, 2002, pp. 7--28,
http://www.cs.bham.ac.uk/research/projects/cogaff/00-02.html#58,


One difference from the Horizon programme is that this posed a problem not about
humans vs non-humans, but about a major difference between species, where humans and
some non-human species were born underdeveloped and incompetent, whereas other
species were born (or hatched) much more developed and more competent. I don't know
whether I had originally got that idea from someone else, or merely by reflecting on
the extraordinary competences of some newborn, or newly hatched, animals, which
contrasted with other species, including humans and hunting mammals. For instance,
compare lion cubs born helpless and incompetent and a newborn wildebeest, able to get
up soon after birth, walk (or stagger) to its mother's nipple, and start sucking,
then shortly after that run with the herd. That's far beyond what the best current
robot 'brains' can support, despite much impressive progress in the last decade.
Jackie Chappell referred me to a wonderful Youtube demonstration of what can happen
when an 'altricial parent' (in this case a cat with kittens) adopts some additional
'precocial babies' (in this case three ducklings). The ducklings are highly mobile
and capable of feeding themselves at a trough, whereas the kittens are much more
helpless and dependent on the mother. The poor cat repeatedly tries and fails to
get the ducklings to join the pile of kittens. Though it all ends well when it's
time to sleep.
https://www.youtube.com/watch?v=bkXRuayRw4U
So why did evolution not produce all new mammals as mature as a wildebeest or a duckling,
including both grazing mammals and hunting mammals?

Perhaps the answer is that some animals must, for various reasons, be fully
capable at or very soon after birth/hatching (e.g. deer cannot pick up their babies
and carry them) whereas others don't have to be. So there is a tradeoff: Among animals
born less mature, some of them, the altricial species, can develop both brain
mechanisms and sensory motor subsystems in parallel and allow them to be better
adapted to their environments. Those born more mature, the precocial species, can be
much less dependent on parents (e.g. chicks and ducklings that can walk around and
peck for food). But perhaps they pay the price of a narrow set of environments they
can cope with -- closely related to the environments they evolved in.

When Jackie Chappell came to Birmingham she pointed out that I was wrong to make a
"global" distinction between species (altricial/precocial). Instead we should
distinguish competences that are well developed at birth (e.g. sucking in most
humans) and competences that develop over time partly under the influence of the
genome and partly under the influence of the environment.

We wrote a paper summarising these ideas, accepted for a major international
Artificial Intelligence conference in 2005 . That was shortly followed by an invited paper
for a journal, which added more details to the theory, in 2007.

Instead of "precocial species" and "altricial species", we now talk about "pre-configured"
vs "meta-configured" competences. The two papers that presented these ideas are
freely available here: This is still work in progress and the ideas elaborated in the second paper are still
under development.

The key idea is that instead of learning always being the same sort of process, for
instance discovery of regularities in the world, how learning happens at a later
stage can be deeply influenced by what was learnt earlier, so that there are
different routes from what is in the genome to behaviours influenced by both the
genome and the environment. To a first approximation, we can distinguish: That is not meant to be a complete, or precisely formulated, list. In general, items
later in the list depend on more stages of influence both from the environment and
from the genome. All this is roughly indicated in the following diagram, from the
2007 paper.

evodevo

(Chris Miall kindly helped with the design of the diagram.)

We hope these ideas will help to guide further research, in psychology, neuroscience,
genetics, epigenetic mechanisms, learning and good ways to organise teaching, as well
as research in robotics and AI. I am also trying to show how this can help to answer
old questions in philosophy of mathematics.

These ideas are being further developed. Changes will be added to this file.

NOTE: This is part of the meta-morphogenesis project.
http://tinyurl.com/CogMisc/meta-morphogenesis.html

Related work

To be extended...

Maintained by Aaron Sloman
School of Computer Science
The University of Birmingham