__________________________________________________________________________________________ Sent: 10 Dec 2013 To: Anne McElvoy Alison Woollard Kathryn Asbury Roger Kneebone Mark Elder Katy Hickman Re: excellent programme "The Building Blocks of Life and Intelligence" From: Aaron Sloman http://www.cs.bham.ac.uk/~axs I heard most of this excellent discussion just before the 10pm news on Radio 4 last night, and thought I should write to say thank you, because the views presented are so important at a time when there is much going on now that can affect syllabus development, possibly in very bad ways, for years to come. Too many of the people involved in decision-making seem to have very narrow views of the functions of education, and very shallow understanding of the mechanisms involved. I have been trying with biologist Jackie Chappell (http://jackiechappell.com/), who studies animal cognition, to develop a theory that goes beyond the ideas presented in the programme, but supports them. Many biologists make a distinction between (1) *precocial* species, whose members are born or hatched fairly well developed, relatively independent, with many of their competences genetically specified (e.g. most invertebrates, chickens that peck for food after hatching, deer whose newborn foals can find their way to the mother's nipple shortly after birth and can run with the herd soon after that) and (2) *altricial* species, born relatively underdeveloped and incompetent, with competences apparently developed over long period of time, largely influenced by the environment (e.g. corvids, hunting mammals, cetaceans, apes, and humans). Different biologists emphasise different distinctions: E.g. http://www.stanford.edu/group/stanfordbirds/text/essays/Precocial_and_Altricial.html http://web.colby.edu/mainebirds/2009/08/06/86/ Jackie and I argued in a 2007 paper that this is over-simple, and that instead of distinguishing *species* we need to distinguish *competences*: At one extreme pre-configured competences are largely specified in the genome and may be manifested soon after birth or much later (e.g. sucking and breathing in humans), whereas meta-configured competences are only schematically specified in the genome and when their development is turned on (some relatively late in life) they can take shape under the influence both of current environments and the previously acquired information and previously acquired meta- and pre- configured competences. Different species can have different combinations of precocial and altricial competences. We conjectured that high intelligence in adults (e.g. hunting mammals, corvids, elephants, apes, humans) requires an extended, structured learning process, involving qualitatively different layers. Evolution seems to have 'discovered' that, for some of the competences, delaying their initiation could bring great benefits because they have so much previously acquired material to build on. For instance, important aspects of mechanisms of language development are delayed until they have a lot of material to work on including knowledge of things to communicate about, desires whose fulfilment needs communication with others and a host of physical, perceptual, and other competences required for language production and for activities in which language can play an important role. For example, language learning processes can make use of previously acquired social and game-playing competences. More subtly, the development of grammatical competences seems to be delayed until a lot of pattern-based linguistic competence has already been acquired, and then sophisticated mechanisms, apparently unique to humans, reorganise what has been learnt into more powerful forms, with much wider scope -- i.e. the grammar of English, French, Urdu or whatever the child has been learning. If the grammar formation started too soon it would have too little evidence and could waste time and resources going down false trails requiring later correction. But at first those new grammatical competences cannot cope with exceptions that need special handling (especially in a language like English), though the exceptional cases were previously used fluently because they were not exceptions, just examples. (E.g.: a child who had previously learnt patterns like 'Johnny hit me at school', starts saying 'Johnny hitted me at school' when grammatical competence is building, and before the exception-handling mechanisms have developed (a sophisticated software engineering task). Musical competences seem to me to have partly similar requirements. More abstract competences (e.g. some communicative and social competences, some teaching and learning competences, and many competences related to doing jobs in the community) have to develop at a still later stage since they have to deploy complex previously acquired competences. It's as if evolution discovered the importance of various timed delays for many meta-configured and meta-meta...-configured complex competences that build on others. The following figure shows different routes from genes (at top) to behaviour in the environment at bottom, followed by feedback from the environment. (From Chappell and Sloman 2007, IJUC. Chris Miall helped with the diagram.) http://www.cs.bham.ac.uk/research/projects/cosy/papers/#tr0609 The simplest, most inflexible, routes downward are on the left side of the diagram. More complex routes develop later, using information acquired by use of competences developed via routes further to the left, using feedback from the environment, crudely depicted in the diagram. This (admittedly still sketchy) theory could explain the initially puzzling phenomenon mentioned in your programme: some features that develop later are more strongly influenced by the genome than some that develop earlier. Example: the desires and competences required for procreation have very strong genetic influences, but in many species their development has to be delayed until a host of other competences have developed in addition to physical size and strength required for child-bearing and rearing. So by the time strongly pre-configured mating mechanisms are beginning to have a powerful influence, they can build on and make use of results of meta-meta...configured competences, including linguistic competences, artistic competences, athletic competences and others that can play a role in the mate-influencing and selecting process -- which by then vary enormously across individuals and cultures. Writing love letters may be strongly influenced by powerful genetic mechanisms but make use of meta-configured linguistic knowledge and competences acquired over an earlier extended period of development, leading to huge differences in the script, vocabulary, syntax, and other details of the letters, even though the fact of writing is a product of common genetic mechanisms related to procreation. A consequence of the theory is that there can be very many different routes to common end states (which may differ in ways that reflect the trajectories). So teenage linguistic, story-telling, mathematical, musical, geographical, philosophical, engineering, and scientific competences can take many different forms if allowed to build on different life histories in different individuals and different social groups. An enlightened educational system needs to allow for this, and even encourage it, since diversity brings many benefits. But, as your programme indicated, there are increasing pressures towards regimentation and uniformity, partly because of the worthy but ill-considered desire for high standards, and partly because of the reliance on league tables as a management mechanism, which requires use of common measures, and therefore common content to be measured, denying learners and social groups opportunities to build fruitfully on individual life histories and regional sources of cognitive content. Such policies favour those whose extra-curricular learning trajectories match some implicitly required standard pattern. Even if the standard pattern could be made available nationally, this would lead to paucity of resources for future developments in the community, including restricting scientific, artistic, and other adult potential. Some captains of industry, and other employers, *think* they need clearly defined qualifications so that they have a good source of potential employees and reliable indicators of achievement, and in the process they lose something they *actually* need even more, namely diversity of talents and ideas and high levels of creativity. I am involved in discussions about changes that need to be made in computing education; and, unfortunately, the same old mistakes are about to be made by well-meaning but misguided teachers, computer scientists, administrators and politicians who focus too much on apparently economically significant aspects of computing and computer science and not enough on the deep relevance of ideas about computation to many sciences and many aspects of life. (They are products of the broken system they are trying to mend.) We need a system that provides strong support for unplanned, individually tailored, diverse, learning trajectories, and I think your programme gave some of the reasons. A better understanding of the mechanisms of cognitive development, and the diversity of developmental trajectories, will help to support those reasons. A deeper understanding requires extending the work of developmental psychologists such as Piaget and Vygotsky, in combination with research into animal cognition, and experiments on various mechanisms of cognitive development in intelligent machines, all enriched with a healthy dose of analytical philosophy. === Note: I was particularly pleased that this programme had none of the distracting acoustic junk so often added by patronising BBC radio producers who think listeners cannot understand what is being said without illustrative or mood-setting background noises or music -- making many of the words uttered hard to recognise, for people of my age with presbycusis, now in their millions because people live much longer than they used to. The paper that originally presented the diagram above (in a slightly different form): Jackie Chappell and Aaron Sloman, 2007, Natural and artificial meta-configured altricial information-processing systems, International Journal of Unconventional Computing, 3, 3, pp. 211--239, http://www.cs.bham.ac.uk/research/projects/cogaff/07.html#717__________________________________________________________________________________________
A PDF version may be added later.
A partial index of discussion notes is in
Maintained by Aaron Sloman
School of Computer Science, University of Birmingham.
(Retired philosopher in a Computer Science department)
Installed: 10 Dec 2013
Last updated: day