School of Computer Science THE UNIVERSITY OF BIRMINGHAM CoSy project

Unjamming the education pipeline
Thoughts on educational prerequisites
for an ambitious European research initiative
Aaron Sloman
Last updated: 20 Jul 2007

This is a letter written to some of those who were at the first ERCIM Interlink workshop, Nice May 2007
At the opening meeting of the Interlink workshop, I referred to the
folly of making huge financial resources available for EU research
funding without doing anything at European level about the
difficulty of finding enough high calibre researchers to fill
research posts.

Getting adequate research staff is going to be a serious problem for
successful proposals in the current FP7 exercise, and it may be even
worse in future if the recent downturn in numbers of university
students doing computer science in the UK is part of a broader

I claimed that at least a part of the cause of the shortage of
researchers is a misguided educational policy in schools,
emphasising the teaching of the *use* of tools because that is what
employers think they need.

This causes too many bright school leavers to be put off studying
computing and AI at university, because they see computing as like
cooking: a useful skill but lacking in challenging intellectual

Various people came up to me afterwards and said the point was
important enough to be part of a statement about research plans.

So I have put together a few paragraphs that might be used if a
report on future research requirements and targets is to come out of

This is just a first draft and anyone who wishes to rewrite it in
any way is free to do so.

I have an overview paragraph, suited to an executive summary and a
longer set of notes.

Executive summary

A European-wide initiative to produce ground-breaking science and technology will fail if there is not a steady stream of new young graduates who have received a broad and deep education, with much variety of content, including learning how to think creatively about very complex information processing systems. Such people need to use interdisciplinary experience and knowledge to produce new theories about how natural systems work and to use those theories in developing and testing designs for new working systems. The current educational system, far from producing such people, turns many bright youngsters away from computing as a subject for study because they are taught to treat computers as pre-packaged sets of tools (eg. word processors, browsers, spread-sheets, databases) instead of learning to design, implement, test, debug, analyse, compare and explain complex working systems. A radical change in computing education from primary schools onward is needed if Europe is to have a steady stream of new researchers who are able to move the frontiers. The movement towards uniformity in the Bologna process may make things worse by reducing the diversity of educational outcomes.

Discussion and some history

A European-wide initiative to produce ground-breaking science and technology will fail if there is not a steady stream of new young graduates who have received a broad and deep education, including learning how to think creatively about very complex information processing systems. Such people need to be able to produce new theories about how natural systems work and to use those theories in developing and testing designs for new working systems. The notion, sometimes put forward, that design can be avoided by allowing systems to evolve, or to learn, grossly underestimates the search spaces: evolution requires many millions of years and many millions of concurrently active individuals in ecosystems to produce human-like designs. Moreover, individual learning will also get lost in a combinatorial morass if the starting point is not well designed, as it is with animals: current newborn animals have the benefit of millions of years of evolution.[*] [*] @Article {chappellsloman-ijuc, author = "Jackie Chappell and Aaron Sloman", year = "(in press)", title = "{Natural and artificial meta-configured altricial information-processing systems}", journal = "International Journal of Unconventional Computing", note = "", } So even self-organising systems will need to be produced by outstanding thinkers if they are to work in a reasonable time. Producing a steady stream of broadly educated, deep and creative researchers seems to be very difficult for our current university system. In part that is because universities are not being fed with school-leavers who have already received a suitable education. Starting at age 18 or 19 is too late. And there are not enough of them who have outstanding potential and want to work in this area. About 35 years ago it was clear to some people that teaching programming in combination with other disciplines, including philosophy, could have profound educational importance in stretching minds in new ways, including providing the above competences. I once argued that 'programming' should join the three 'R's (Reading, wRiting, aRithmetic) to form the five 'R's, in a basic general education -- i.e. programming combined with analytical, explanatory and communicative competences, including designing and documenting simple AI systems (e.g. conversational systems, simple planners, simple problem solvers, simple image analysers, etc.).

Then what happened?

During the next few decades many people (including politicians, industrialists, parents, and badly mis-informed teachers) claimed that computers should be put in schools to enable children to learn to use the tools they are going to need in their jobs. So schools started using PCs running Windows (often without any programming language installed) and children had to learn to use word processors, databases, spread sheets, then later web browsers and email systems. They did not learn to analyse problems, design working solutions, implement them, test them debug them, describe their solutions, compare alternative solutions and seek ways of improving and extending working systems. Neither did they learn to represent interacting structures and processes in a mathematical formalism to support reasoning about their properties. Instead children merely learnt new ways of assembling text and pictures, searching for items of second hand information, sending messages, and presenting documents. This is like teaching people physics by teaching them to drive cars and tractors. As a result of the change towards teaching only *use* of tools, most of the bright learners decided that computing was just a collection of boring but useful tools and not a suitable subject to study at university. I first discovered that when the very bright daughter of a professor of computer science told me why she would not dream of doing CS at university, nearly a decade ago. And now a whole generation of learners has passed through the system, having been taught only the wrong things. A huge amount of human potential has been wasted. Undoing this damage will be an enormous problem, including finding ways to get get the right things taught in many thousands of primary and secondary schools where there are very few teachers capable of doing the required kinds of teaching, and where the infrastructure for collaborative learning and teaching of computing is not available. (Networked file systems allowing shared files and distributed support across schools may be essential for this to work without massive local costs.) Unfortunately national and international committees with politicians and industrialists who do not understand the problems keep recommending more of the same strategy that has failed in the last two and a half decades. Unless that is changed, no amount of planning grand research initiatives will work: there will not be enough highly intelligent highly educated researchers available to do the required work. Things are made worse by the Bologna declaration to which many nations have signed up, which emphasises uniformity of educational pathways, because that will support mobility and transparency of educational systems. That may be desirable for people who are going to do well-defined clearly circumscribed jobs, such as plumbers, nurses, medical practitioners, teachers of specific disciplines, bus drivers, accountants, factory workers, street sweepers, and many more. However, it will be disastrous for people who have the potential to learn to be outstanding researchers, pioneering teachers, innovative engineers, or business leaders. These are the people who, above all, need to be able to strike out in new directions. There is no way that governments or international committees can safely select a subset of the possible forms of education to suit such people. So educational systems with diverse contents and aims and means should be encouraged for this group of learners, leaving it to highly creative and highly motivated teachers to develop their own educational ends and means for learners whose outstanding abilities will help drive the processes. The research process, and perhaps the top level jobs in all fields, should be open to people with very diverse kinds of knowledge, provided that they have learnt to challenge, analyse, develop complex ideas, test them, compare them, explain them, and reject them when they are inadequate. For such learners, international, or even national *uniformity* in education can be compared with constriction of a gene pool and will severely harm our ability to find top class researchers with many different sorts of competence, and very different background knowledge, to push the frontiers of knowledge in new directions.

Some previous work on this.

I have written at greater length about how educational opportunities provided by computers have been wasted because of bad decisions. Education Grand Challenge: A New Kind of Liberal Education Making People Want a Computing Education For Its Own Sake A description of the missed educational opportunity and some suggestions A partial solution: a proposal for a mind-stretching interdisciplinary curriculum centred on teaching AI in the last two years of school. (But some of it could and should start earlier, e.g. between the ages of six and ten.) Why Computing Education has Failed and How to Fix it (Message sent to CPHC and UKCRC mailing lists) Aaron