ÐÓ°ÉÔ­´´

A STUDY TO rediscover some of the techniques devised by carpenters in
ancient Egypt may seem a long way from any of the issues involved in introducing
new forms of advanced automation. The link, according to researchers at
the Swedish Centre for Working Life (Arbetslivscentrum) in Stockholm, is
the need to understand the nature of human skill and how skills may be enhanced
or degraded by the use of different technologies.

The research at the centre, led by Bo Goranzon, is one of the fruits
of a distinctive approach to the use and design of computer systems that
has evolved in Scandinavia over the past 20 years. Researchers and trade
unionists in all the Scandinavian countries have been experimenting with
ways of involving users in selecting and, ultimately, designing computer-based
automation systems . They are particularly influenced by ideals of industrial
democracy, and believe that what they have achieved has been made possible
because of high levels of trade-union membership and a tradition of cooperation
between unions and managements in their countries.

This cooperation is reflected in the methods of the Stockholm skills
study. The researchers rely on case studies, but the skilled ‘subjects’
of these case studies, for example, photographers, nurses, and carpenters,
become part of the research team. At the same time, the researchers have
employed the descriptive approach of the humanities, rather than the analytic
techniques of the natural sciences.

The Swedish researchers turned to philosophy in particular to provide
ways of describing human skill. Goranzon describes his method as ‘a theory
of knowledge approach’, and emphasises that he is not ‘doing’ philosophy,
but using philosophical discussions of the nature of knowledge as a framework
for his research. He thereby challenges more ostensibly scientific attempts,
in particular some current attempts to lay the foundations of a science
of psychology. Many cognitive scientists, for example, take the computer
as a model for human intelligence. These researchers, often workers in the
field of artificial intelligence, assume that there is some broad similarity
between the way in which computers process programs and the way in which
brains or minds ‘process’ thought. They argue that intelligence is based
on the manipulation of symbols. This assumption provides a basis for trying
to build a scientific psychology, partly because it provides testable hypotheses.
It also gives a basis for attempting to automate human knowledge and skills,
as it implies that intelligent behaviour is ‘computable’, that is, expressible
as a series of propositions, or rules, or algorithms.

One limitation on the analogy between minds and computers is that it
seems to lead to an ‘intellectual’ view of intelligence. Researchers in
artificial intelligence provide a pragmatic, if not always systematic, test-bed
for some of the claims of the computational model of the mind. They have
been more successful in reproducing intellectual than practical skills.
Chess programs, for example, can now beat players up to the level of a Senior
Master, but the skills of a two-year-old child in making its way round a
room are more difficult to reproduce. Some researchers believe that this
illustrates the problem of what is known as ‘tacit’ or implicit knowledge.
Tacit knowledge is knowledge that is not naturally or immediately expressible
in words; in contrast, the rules of chess, for example, can be expressed
in a series of propositions, or made explicit. Even in this case, however,
some critics would argue that an advanced chess player has developed so
far beyond applying the rules that the player’s skill cannot be captured
explicitly. Practical skills such as driving a car, or even walking across
a room, are examples of implicit knowledge: they are both tasks that humans
find easier to do than to describe.

Some researchers are reacting to these limitations by turning to new
models of computation, such as designs for computer hardware or software
based on models of neural networks. Programming a ‘neural network’ computer
requires mathematical and statistical skills. It involves devising an algorithm
that will enable the machine to find its own way to complete a task, for
example, by recognising a pattern when it recurs. The programmer does not
have to find a formal description of the task itself, such as a list of
all the visual clues that indicate a particular feature. The machine will
pick up the clues by applying a statistical analysis. This approach may
help in automating some types of ‘tacit’ knowledge, such as recognising
a face. By the same token, it can no longer pretend to provide a cognitive
description of what a human being does.

The researchers in Stockholm are not particularly interested in whether
it is possible to construct intelligent machines – the field of artificial
intelligence, or AI – or how accurately pieces of intelligent human behaviour
can be reproduced by computer programs. Goranzon argues that the question
of whether machines can think is not the right question to ask. The more
important question, he believes, is the impact of automation on work and
human skills, and he is looking for a way of describing skills that allows
us to discuss this question.

Goranzon draws on the later writings of the philosopher Ludwig Wittgenstein
to devise a description of tacit knowledge and the nature of skill. Another
researcher, Ingela Josefson, used Wittgenstein’s work to help a group of
nurses to express why they believe practical skills are valuable in their
work. Both Goranzon and Josefson focus on Wittgenstein’s notion that to
follow a rule is to grasp how to do something in a practical way, on the
basis of examples, rather than to apply a set of instructions.

In one study Goranzon observed the work of a professional photographer,
Peter Gullers. Gullers describes what he is doing when he assesses the light
conditions for a photograph, without the benefit of an automatic light meter.
The judgment he makes is based on years of experience, but ‘all of these
earlier memories and experiences that are stored away over the years only
partly penetrate my consciousness . . . The thumb and index finger of my
right hand turn the camera’s exposure knob to a setting that ‘feels right’,
while my left hand adjusts the filter ring. This process is almost automatic.’
The rules he follows are expressed directly in action; they are not a set
of propositions, not even a set of formulae for calculating the f-stop.

There is another important implication of Goranzon’s study with Gullers.
A skilled professional photographer who judges the light conditions by eye
may not be behaving like a computer; nonetheless, the task is computable.
An automatic light meter does something functionally equivalent to the operation
of the skilled photographer. This brings us back to the point of the Stockholm
study, whose primary aim is not to ask what skills can be automated, but
what happens when a skill is automated. We should probably expect a photographer
who habitually uses a light meter to lose the ability to judge light conditions
by eye. But, Gullers suggests, they lose more than that – their sensitivity
to shades of light becomes blunted, which affects the quality of the pictures
they take. The machine cannot compensate for this: no light meter can, to
take one of Gullers’ examples, appreciate how in a rolling mill ‘the light
coming through the sloping skylight contrasts with the sooty heat of the
air in the brick building’.

Gullers’ study may seem to be a special case, perhaps involving artistic
or aesthetic skills. Josefson’s study with a group of nurses, however, helps
to illustrate that what are at issue are peculiarly human skills. Josefson
used passages from Wittgenstein’s Philosophical Investigations to help the
nurses to analyse what they felt to be valuable about their own practical
skills. Among other things, the nurses seem to have arrived at a firm sense
of where the value of their skills lay. For example, a passage in the book
on ‘the knowledge of mankind’ provided the subject of repeated discussions:

‘Can one learn this knowledge? Yes; some can. Not however by taking
a course in it, but through ‘experience’. Can someone be a man’s teacher
in this? Certainly. From time to time he gives him the right tip. This is
what ‘learning’ and ‘teaching’ are like here. What one acquires here is
not a technique; one learns correct judgments. There are also rules, but
they do not form a system, and only experienced people can apply them right.
Unlike calculating-rules.

‘What is most difficult here is to put this indefiniteness, correctly
and unfalsified, into words.’

Passages like this provided ways of understanding some of the experiences
that the nurses brought to the group. One nurse, for example, described
how she learnt to deal with violent patients by observing closely the behaviour
of an older and more experienced nurse. Another described her own ability
to ‘see’ when a patient was not recovering properly from an operation, although
all the vital signs were normal and she could not explain why she was concerned;
she believed that her experience was a decisive factor.

The implications of this study, as with that by Goranzon and Gullers,
go beyond simply describing a particular skill. In an essay called ‘The
Nurse as Engineer’, Josefson contrasts the results of her study with the
concerns of other researchers who wish to give nursing a sounder theoretical
and scientific base. Judy Ozbolt of the Center for Nursing Research at Ann
Arbor in the US, for example, hopes to bring in expert-systems techniques
to remedy the idiosyncratic, unreliable and untested practice of nurses.
But for Josefson, practical skills that can be learnt only on the ground
and that resist easy formulation in a set of rules are the essential basis
of nursing. For Ozbolt, the reliance on practical experience and on apprenticeship
methods of learning represent a shortfall in scientific standards and professionalism.
The differences between the two researchers reflect different theories of
knowledge, with profoundly different implications about how nurses should
be trained and what constitutes a ‘good’ nurse.

The Stockholm researchers are not opposed to new technologies on principle,
but they argue that they must be regarded as tools rather than as replacements
for people. They also argue that those using the tools must be more closely
involved in developing them. This brings us back to the study of carpentry
techniques in ancient Egypt. A cabinet maker called Thomas Tempte, who has
been working with researchers at the Centre in Stockholm, saw Tutankhamen’s
chair at an exhibition. He began to study the woodworking tools used in
ancient Egypt, and then decided to try to reconstruct the chair. This turned
out to be an exercise in understanding the constraints imposed by the tools,
as well as learning respect for the craftsmanship of woodworkers for whom
there were no short cuts, no templates or standard instruments, and whose
skills had to range from developing an eye for proportion to the ability
to choose wood. Tempte points out that craft workers have in the past invented
and modified the tools that they use, and that they can still be involved
in the design of their tools. A contemporary example of this comes from
a project undertaken by Swedish and Danish researchers in collaboration
with the Nordic Graphics workers union to design a computer system.

The lessons of the various different case studies making up the Stockholm
skills study cannot be directly translated into a recipe for creating or
introducing computer systems that make the most of human skills. One of
the lessons, indeed, is that such methodologies will probably be developed
and proved only in practice. But the systems that are built in this ‘human-centred’
way should lead to better ways of working.

* * *

The Swedish Centre for Working Life

THE Swedish government set up the Centre for Working Life in Stockholm
in 1977. The centre receives government funding and includes representatives
from trade unions and employers’ organisations on its managing board. Its
researchers study aspects of working life, and under this remit they have
undertaken several case studies and projects relating to the impact of new
technology.

In the late 1970s, Bo Goranzon led a study of the likely impact of introducing
new technology at Sweden’s national insurance (social security) offices.
This is known as the PAAS study (from the Swedish words for ‘Perspective
on Analysis, Tools and Working Methods in Systems Development’). More than
10 000 staff at many different offices participated in study circles and
answered questionnaires. The PAAS study covered issues such as methods of
consultation for finding appropriate systems, preserving or creating job
satisfaction, and education and training – all in the context of introducing
new technology into a particular organisation.

Goranzon and other researchers moved from the PAAS study to the skills
study. They have also been involved in other related projects, including
a series of seminars organised jointly with the Swedish Research Council,
called the Dialogue Seminars, staged at Stockholm’s national theatre, the
Royal Dramatic Theatre. The seminars explored the history of the concept
of dialogue in philosophy, the theatre and the natural sciences, as a way
of giving an account of a practical skill.

The Working Life Centre has also attrac-ted a number of professional
philosophers. For example, Kjells Johannessen, a lecturer at the Department
of Philosophy at Bergen University, has undertaken a study of the concept
of practice in Wittgenstein’s later philosophy, which lays the foundation
for the research at the Working Life Centre.

The philosopher Allan Janik from Brenner Archive in Innsbruck has also
been associated with the Centre for some years. He has been exploring the
notion of tacit knowledge, inspired by the Austrian writer Karl Kraus and
Ludwig Fleck, a philosopher of science.

The Working Life Centre was also involved in an EEC-funded research
programme called COST-13 on ‘AI-based systems and the future of language,
knowledge and responsibility in the professions’. Others involved in the
study included the Norwegian Research Institute for Computers and Law at
the University of Oslo and the Austrian Research Institute for Artificial
Intelligence in Vienna.

Now Goranzon and other researchers have begun a new international research
programme, which will extend the philosophical and historical scope of the
Stockholm approach. This is called the Diderot project, after the 18th-century
French editor of the encyclopedia who attempted to propagate the scientific
and political ideals of inquiry and tolerance now characterised as the Enlightenment.
Researchers in Sweden, Austria, Britain and France are involved in the Diderot
project. This project will, again, bring a humanistic approach to bear on
the continuing issue which underlies all the Stockholm skills research:
the impact of computers on work and human skills.

* * *

Human-centred systems have their roots in industry

THE ROOTS of the Scandinavian approach to designing computer systems
go back to a series of projects carried out in the 1960s, called the Norwegian
Ind ustrial Democracy Programme. This was organised jointly by the Norwegian
Federation of Trade Unions and the Norwegian Employers’ Federation. The
programme used techniques developed at the Tavistock Institute in London
for dealing with social and technical change, known as the ‘sociotechnical’
approach. The aims of this approach include creating work structures that
are nonalienating and provide job satisfaction. The Industrial Democracy
Programme also organised a number of experiments on the organisation of
work.

The programme seems to have had a poor record of creating interest among
workers. The broad approach was revived, however, largely thanks to a computer
scientist called Kristen Nygaard. In 1970 a new project was jointly organised
by the Norwegian Iron and Metalworkers Union (NJMF) and the Norwegian Computing
Centre. They were critical of sociotechnical methods and looked for ways
of getting workers genuinely involved in the programme, for example by setting
up local union clubs so that union members could learn about planning, control
and data processing and ultimately initiate new ways of using new technology.
A series of projects in all the Scandinavian countries followed the path
beaten by the NJMF study.

For example, Swedish and Danish researchers undertook the Utopia project
with the Nordic Graphics workers’ union. As with earlier projects in the
tradition, researchers and workers spent a great deal of time learning from
each other. They concentrated on the task of page make-up (designing the
layout of a page) on a newspaper. Together, the researchers and graphics
workers carried out lengthy experiments, using paper and plywood models
of a graphics work station to decide just what sort of tools would best
support the process of page make-up. A system called TIPS, based on the
Utopia work station, was installed as a pilot study at the Swedish newspaper
Aftonbladet. However, this turned out to be a test-bed only for the computer
system and not for some associated ideas about new working practices.

Elsewhere in Europe, other small groups of researchers have explored
similar or related ideas. They include Enid Mumford at the Manchester Business
School, who has continued to develop the sociotechnical approach, Howard
Rosenbrock at UMIST, Mike Cooley at the London Enterprise Board, and researchers
at the University of Bremen in West Germany.

Some of these groups were brought together in a recently completed project
on advanced information, which was funded by the European Esprit programme
and led by Mike Cooley. Researchers and companies in Britain, West Germany
and Denmark cooperated on a project to build human-centred systems for Computer
Integrated Manufacturing (CIM). Systems that this European initiative developed
have now been installed at two British companies, Rolls-Royce and BICC.

Rolls-Royce Aeroengines, for example, has installed a human-centred
lathe system. An operator from Rolls-Royce has been trained in modern computer
skills as part of the project. He is responsible for programming the machine
and can use it for turning, milling and drilling (tasks which are usually
done on three different machines). Managers at Rolls-Royce are enthusiastic
about the flexibility of the system. They point out, however, that the human-centred
system is just one of many approaches that Rolls-Royce is exploring.

Danish researchers in the Esprit project have developed a particularly
innovative tool for industrial designers. The electronic sketch pad allows
designers to continue using pencil and paper: they will be able to copy
a drawing or sketch into the system and then produce precise designs with
the help of CAD (computer-aided design) techniques, or simply use the system
as a sort of electronic folder for storing the drawings, or as a sketch
pad for sending ideas to other departments.

Janet Vaux is a freelance journalist specialising in the field of artificial
intelligence and advanced computing.

Further reading: Knowledge, Skill and Artificial Intelligence, edited
by Bo Goranzon and Ingela Josefson, Springer-Verlag, pp 193, Pounds sterling
20.50; Artificial Intelligence, Culture and Language, edited by Bo Goranzon
and Magnus Flavin, Springer-Verlag, pp 290, Pounds sterling 28.50 (published
in March); and Dialogue and Technology, edited by Bo Goranzon and Magnus
Flavin, Springer-Verlag (to be published in 1990).