For editors Andrew Feenberg and Alastair Hannay, one of the key starting points is 20th-century German philosophy. Several papers in this collection also tackle the philosophy of the German existentialist Martin Heidegger, and their authors provide useful views on the content and significance of his reading of technology.

Terry Winograd’s paper on Heideggerian artificial intelligence contains the most startling evidence of the philosopher’s influence. Computer systems designers at MIT and elsewhere are, he says, describing their activity with concepts borrowed from Heidegger and from Dreyfus’s own Heideggerian critique of AI.

Feenberg and Hannay seem particularly interested in questions to do with the “meaning” of technology: for instance, is it a tool for increasing socioeconomic efficiency, or an alienating force?

The book’s other main point of reference is recent work in the sociology of technology. Feenberg ambitiously attempts to bring together the politico-ethical concerns of the Frankfurt school with sociological constructivism. Constructivism, as Feenberg explains it, “argues that theories and technologies are underdetermined by scientific and technical criteria”.

One of the most famous examples of constructivism, concocted by Trevor Pinch and Wiebe Bijker, concerns the social construction of the bicycle. Pinch and Bijker argue that its development from penny-farthing to the modern version was not a case of evolving technical design, but the result of solutions to a series of problems afflicting particular interest groups. For example, racing cyclists initially preferred high-wheeled cycles for speed; only the discovery that low cycles with air tyres were even quicker weaned them from the old design which to us, with the wisdom of hindsight, looks doomed from the start.

As editors, Feenberg and Hannay have made a selection of papers which places their own agenda in a rather broader context.

The issue of ‘realism’, or the theory that the extemal world is real,
is a central one in several recent introductory books in the philosophy
of science. James Brown provides one reason for this in Smoke and Mirrors.
Realism, he suggests, underwrites science; without an independent world
out there, one theory would be as good as another. Realism, he says, ‘is
a wonderfully rich and enlightening story – a true story I think – which
makes our thought and our experience intelligible. Antirealism says our
theories are ‘tales . . . full of sound and fury, signifying nothing’.’
Brown’s version of realism is deliberately paradoxical – science is ‘merely’
a story, but it is a true story.

Brown’s book is a counterblast to anti-realists and above all to relativists,
those who believe that science is just one meaningful story among others
– or, as Brown would represent them, those who believe that science is just
one meaningless story among others. For Brown, in other words, there is
only one true science.

This may seem like common sense, but Brown’s targets include representatives
of some important currents in late 20th-century thought. He is beset, on
one hand, by postmodernism with its suspicion of science as one more ‘grand
narrative’. On the other hand, he fights off the various ripples which continue
to grow from Thomas Kuhn’s story that scientific revolutions involve a paradigm
shift.

Collier’s book is an introductory account of the realist philosophy
of Roy Bhaskar. Collier and Bhaskar share many of Brown’s targets, including
postmodernism and the sociology of knowledge. In addition, however, they
also reject the hankering after a single ‘truth’ which keeps surfacing in
Brown’s version of realism. They believe (roughly speaking) that what scientists
can ‘know’ is determined by the representations available in their theories
(epistemological relativism), but there is an ‘external’ world, independent
of our knowledge of it (ontological realism). Bhaskar’s philosophy is both
original and difficult, owing much to Hegel and Marx. Collier’s book provides
a useful introduction to Bhaskar, and is worth reading especially for his
discussion of what a scientific experiment is. Experiments, for Bhaskar,
provide a way of describing what science is, in contrast to older philosophical
stories about laws and regularity.

Science students (even social science students) may find both these
books difficult, but not because of the style in which they are written.
Brown, in particular, goes out of his way to jolly the reader along with
jokes and acerbic asides. One difficulty is to do with the difference between
a philosophical and a practical discussion. Why should science students
bother with philosophy? ĐÓ°ÉÔ­´´s, after all, are not waiting on philosophers
to settle the matter of whether the world exists, or how science should
properly be accounted for, nor are the politicians who fund science.

One reply to this is that without some knowledge of the philosophical
tradition, scientists are doomed, not to avoid philosophy, but to do bad
philosophy. As an abstract argument, this may not have much weight, but
for a sense of some of the prac-tical issues involved it is worth looking
at Mary Midgley’s book, Science as Salvation.

One difference between Midgley’s book and the sort of book that philosophers
of science usually write is that its main target is scientists, rather than
other philosophers, literary theorists or sociologists. She is not looking
directly at the question of whether science is or can be ‘well founded’,
although she does address this and many other philosophical issues along
the way. Her interest is the changing conceptual framework within which
scientists have thought about their work, and she places science firmly
back within the bloodstream of myths, desires and values.

In the 17th century, (male) scientists were positively obsessed with
a particular, highly gendered image of the activity of science. Modern scientists,
Midgley argues, are trying to construct a brave new world view in which
Man the scientist subdues the universe. Her exploration of the unconscious
metaphysics embodied in these myths is enlightening. One of Midgley’s achievements,
I think, is to show that it matters that modern scientists are philosophically
illiterate, at least in discussions that go beyond their narrow specialities.

Steve Fuller, in Philosophy, Rhetoric and Ihe End of Knowledge, has
a slightly different sort of interest in promoting dialogue between different
groups. In fact, this book may seem a bit daunting, because it addresses
so many audiences at once, including philosophers, social scientists, politicos
and policy makers. Unlike both Brown and Collier, Fuller is less interested
in defeating postmodernists and sociologists than in coopting them into
a perspective he has christened ‘social epistemology’. This is not altogether
an introductory book, but it is worth tackling for its panoramic view of
late 20th-century theory and politics.

Finally, this is not strictly philosophy of science, but for those who
want an introduction to postmodernism, nothing could be better than the
recently published second edition of Madan Sarap’s book, Poststructuralism
and Postmodernism.

Janet Vaux is a writer and a research student at CRICT (the Centre for
Research in lnnovation, Culture and Technology), Brunel University.

* * *

Smoke and Mirrors: How Science Reflects Reality by J. R. Brown, Routledge,
pp 224, ÂŁ35/ $59.95 hbk, ÂŁ11.99/ $17.15 pbk

Critical Realism by Andrew Collier, Verso, pp 288, ÂŁ39.95 hbk,
ÂŁ13.95 pbk

Philosophy, Rhetoric and the End of Knowledge by Steve Fuller, Wisconsin
University Press, pp 456, ÂŁ48.50/ $54 hbk, ÂŁ19.95/ $22.50
pbk

Science as Salvation by Mary Midgley, Routledge, pp 239, ÂŁ35
hbk, ÂŁ8.99 pbk

An Introductory Guide to Poststructuralism and Postmodernism (second
edition) by Madan Sarup, Harvester Wheatsheaf, pp 256, ÂŁ11.95 pbk

One day in 1966, a first-year undergraduate student at the Massachusetts
Institute of Technology had wangled access to a PDP-6 computer, a model
that had (for those days) a massive 16 000 words of memory. He was trying
to ‘teach’ a computer program to play tic-tac-toe, or noughts and crosses.
A bald-headed gentleman appeared, questioned the student on what he was
doing, and hired him to spend the summer solving the problem of connecting
a computer to a television camera and getting the computer to describe what
it sees.

The bald-headed gentleman was Marvin Minsky, a founding member of the
‘artificial intelligentsia’, a small group of researchers in artificial
intelligence (AI) who believed that human intelligence can be simulated
on a computer. The student was Gerald Sussman, now himself a professor at
MIT. The problem of getting a computer to say what it sees is still not
solved.

Daniel Crevier, who tells this story in his book AI: The Tumultuous
History of the Search for Artificial Intelligence, says: ‘The incident is
now remembered as one of the best examples of the naivete of early AI efforts.
The truth is that, after more than a quarter of a century of additional
research, the problem of computer vision has turned out to be one of the
toughest nuts to crack in AI.’

The artificial intelligentsia were, as Crevier records, skilful self-publicists
and ‘excesses of optimism seem to occur with particular frequency in AI’.
In the 1960s and 1970s, AI researchers at MIT thought they were on the brink
of creating a truly intelligent machine, and they enjoyed intense feuds
with anyone who publicly criticised them. Their most notable clash was with
the philosopher Hubert Dreyfus, to whom they ‘gave the silent treatment’;
Noam Chomsky was booed when he gave a talk at the MIT AI laboratory.

The artificial intelligentsia seemed to have a knack of attracting funds.
Minsky had money from the Advanced Research Projects Agency for a vision
research project burning a hole in his pocket when he chanced on Sussman
playing with the PDP-6. This agency also provided millions of dollars to
American AI researchers through projects such as Speech Understanding Research
and the Strategic Computing Initiative. But by and large these projects
did not deliver exactly what was wanted; most turned out to be closer to
basic research than they were to applied research.

So far, this story may seem strangely familiar to anyone who has read
Pamela McCorduck’s book Machines Who Think, which was published in 1979.
Crevier does cover much of the same ground but his study has two advantages.
First, the 14 years between the publication of the two books includes a
sobering period when AI failed to become the ‘fifth generation’ of computer
systems: Crevier’s chapter headings include ‘The roller coaster of the 1980s’,
referring in part to the rapid rise and fall of specialist AI companies.
A number of academics who had hoped to make their fortune selling AI systems
burned their fingers in the marketplace during this period.

Secondly, Crevier, who teaches electrical engineering at the University
of Quebec, supplies a counter-theme to the story of the new Frankensteins:
he describes the history of the programming ideas. He gives a basic introduction
to AI as a programming style or set of styles. He explains the points at
issue in the design of AI languages such as Lisp and Prolog, and in fields
such as knowledge representation, natural language processing and artificial
neural systems, in a way that is not too taxing for a lay person. At its
best, Crevier’s book is both readable and informative.

In my view, the book deteriorates a little in the last few chapters,
where Crevier is dealing with more recent history. This may be partly because
the mainstream of AI research has become more humdrum, and Crevier continues
to focus his attention on the ‘dream’ of intelligence. So he concentrates
on two oddities: Minsky’s recent work on the ‘society of mind’, and Doug
Lennat’s project to write a program that will have ‘common sense’. At this
point, Crevier ceases to act as a reliable guide to the concerns of most
academic researchers in AI, or of any programmers applying AI techniques
in the real world.

The final chapter has a few gems in store, however. I particularly like
Sussman’s description of how he uses ideas of computer programming, based
on the AI language Lisp, to explicate complex algorithms to electrical engineering
students. He suggests this represents ‘a breakthrough in how people can
express complex ideas’ that is comparable to the development of calculus
for expressing ideas of motion, or the development of algebra for expressing
generic ideas about numbers.

Janet Vaux is a freelance writer, specialising in artificial intelligence.
She is currently at Brunel University, researching a thesis on technology
transfer in the AI industry.

DAVID HAMLYN was, until he retired in 1988, professor of philosophy
at Birkbeck College, London. His new book, In and Out of the Black Box,
is a philosophic work and will be read mainly by researchers and students
of philosophy. Its main subject, however, is another academic discipline,
psychology.

Hamlyn looks, in particular, at the project by cognitive psychologists
to construct a model of human thought as an information processing system.
Many cognitive psychologists see this project as the most likely basis for
developing a properly scientific psychology, and some may regard Hamlyn’s
book as the latest counterblast from the humanist or antiscience camp. While
there are several grounds for describing his ideas as humanist, however,
I do not believe they are antiscientific.

Hamlyn sketches out his own approach to the terrain when he says, in
the introduction: ‘Surely . . . psychology, whether scientific or not, must
have some relevance to human beings and animals and such; it must not be
concerned merely with organisms, machines or abstract systems.’ The claim
seems minimal enough: notions such as belief and desire, percep tion and
action, may not be objectively determinable, but a psychologist needs to
be able to give some account of what we are talking about when we use these
sorts of words to describe ourselves and others.

Those familiar with the information-processing model may recognise this
as a reference to the problem of intentionality, or the question of how
an idea or proposition is meaningful for a subject. According to the information-processing
model, the realm of subjectivity is a place equivalent to a transducer,
which transforms incoming data (input) into some appropriate response (output).
This realm of transformation or subjectivity is a ‘black box’, in the sense
that we cannot directly ‘see’ what goes on there, but can only hypothesise
its functions on the basis of what is needed to turn input into output.
Most of us would probably happily accept this as a metaphor for the way
in which intelligence has its place out of sight, inside people’s heads.

How satisfactory is this metaphor? Hamlyn argues that we must broaden
the discussion and turn our attention to the other two terms in the model:
input and output. Input and output are simple concepts which, he suggests,
are quite inadequate to the task of explaining what goes on in, say, perception
or action. Perception, for example, is not simply the reception of data
but itself depends on knowledge and other less obvious requirements, such
as our social existence.

Much of this will be more or less familiar stuff to those who know Hamlyn’s
writings (his basic approach is much the same as it was 20 years ago, when
I was a student at his department in Birkbeck). The importance of this book
is that he applies these arguments to what has become a very compelling
metaphor of thinking, even for many outside academic psychology.

Many professional philosophers of mind might share Hamlyn’s reservations
about the information-processing model, but are content with a more cursory
knowledge of the debate as it is rehearsed in psychology, rather than philosophy,
depart ments.

Hamlyn’s purpose is not to say whether a science of psychology is possible,
or what such a science might be. He comments in the final paragraph of the
book: ‘. . . In science it is not always easy to predict what will come
from what, and it is wrong for a philosopher to get in the way’.

His purpose, rather, is ‘to try to point to conceptual error and to
present what may be a bet ter system of concepts’. His sense of territory
is, if anything, overdeveloped (or perhaps partly disingenuous). I wish
he had been prepared to indulge in more speculation, for example, about
how developments in the neurosciences might (or might not) facilitate a
science of psychology.

Janet Vaux is a science writer specialising in artificial intelligence
and advanced computing.