Ten or fifteen years ago, a typical honours graduate would have learnt
to program in two or three high-level languages on a mainframe computer,
used a rather crude line or screen editor, and carried out a few exercises
on an unsophisticated database package. Such a graduate would be at a loss
in a modern office, where secretaries are busy desktop publishing in full
colour, accounts clerks performing miracles on a powerful spreadsheet, and
everyone has access to vast databases of information, often stuffed with
graphics, pictures and sound, as well as text. Another change is that many
more undergraduate courses include several computing modules, and the resulting
cross-disciplinary teaching creates a need for multipurpose textbooks that
teach a topic in depth without assuming too much prior study.

A good example is Antone Alber’s Interactive Computer Systems. Were
it not for its high price, it would be an excellent text for a commercially
oriented final-year computer science module that could also be offered
to business school students. It looks at the way computers fit into large
organisations, and surveys the factors involved in the use of videotext
and multimedia, with good examples of large operational systems. The book
rightly warns that selecting the correct software for a task is difficult,
time consuming and costly, and that software vendors are under tremendous
pressure to ship products before they are fully tested.

At the other end of the application scale, Moira Monteith’s Computers
and Language will stimulate student teachers who plan to use computers in
the classroom. Many of the points it makes, however, are of wider interest
and could well lead to ideas for student projects. I liked the chapter on
collaborative writing using distanced electronic communications, while the
chapter on multimedia, hypermedia and the teaching of English gives a good
idea of what can be done. As more students come to use computers to prepare
coursework, it is worth reflecting on the conclusion of ‘What is important
in writing’:

‘All the evidence points to the tendency of pupils to develop more nonchalant
attitudes to writing when using word processors. Pessimistically, this means
that even when pupils acquire the technologically-assisted potential to
improve, they frequently fail to utilise it effectively.’

Having begun with applications, the next layer of the onion is the interface
between the user and the computer system. Alan Dix’s Human-Computer Interaction
is a valuable introduction to a subject that plays an increasingly important
part in degree courses. A first chapter on human information processing
leads to a survey of the different interface components of a computer and
how they work. This is followed by well-written sections on the design process,
including models of the user, task analysis, and the approach to interactive
help and documentation. Its coverage of groupware is good, ranging from
e-mail, through meeting and decision support systems, to shared applications.
While I would have preferred more on hypertext and multi-media, this gap
should not put you off this excellent book.

For a more advanced course, covering topics such as speech and handwriting,
and the high and low-level design issues of windowing systems, Andy Downton’s
Engineering the Human-Computer Interface, recently issued in a special student
edition, has much to recommend it.

Cliff McKnight’s Hypertext provides the kind of refreshing interdisciplinary
approach to the subject that livens up final year seminars, although the
psychological jargon in the otherwise interesting chapter on textbooks of
the future may put off some students. The chapter on using hypertext in
learning systems is a particularly well-written introduction to the limitations
of the technique, as well as its advantages. It suggests that the teaching
power of the simpler hypertext presentation systems is poor because they
do not support effective understanding and recall of the material. Different
learning situations require different tutorial approaches, and hypertext,
appropriately applied, can provide the support facilities needed in many
cases.

Software Engineering: Principles and Practice by Hans van Vliet moves
us further from the user interface and into the basic principles of modern
system design. It covers topics such as the software life cycle, cost estimation,
and requirement analysis, before going on to programming languages, testing
and software reliability. It has a very good chapter on software psychology,
which looks at the way programmers think about programs. While more could
be made of the user interface, it would make a good companion to Dix’s Human-Computer
Interaction.

While computer science courses concentrate on general design principles,
modern computer systems often involve specialist software, and the relevant
theory is often offered in the form of final-year options or postgraduate
modules. One such subject is digital image processing and Milan Sonka’s
Image Processing, Analysis and Machine Vision describes topics such as shape
representation, object recognition, image understanding and data compression.
Needless to say, the mathematics needed to understand some of the transformations
is pretty daunting. The power of the techniques is illustrated by a picture
of a car blurred by motion being transformed to a sharper image where the
number plate could be clearly read. It is a great pity that some of the
other photographic illustrations appear to have lost something in the printing.

When we dig down to the programming level, there is wide range of books
appropriate to different course needs. Pascal is a common introductory language
on mainstream computer science courses and Judy Bishop’s well-known Pascal
Precisely has been substantially revised to make it even more valuable.
The chapter structure has been improved, and there are some better examples
and new case studies. Data structures, sorting and searching are taken to
a greater depth than formerly.

For those who need to delve deeper into how software is written, particularly
for more sophisticated hardware, Alan Burns and Geoff Davies offer Concurrent
Programming. It provides a well-written and not too mathematical introduction
to a difficult subject. I found it a pleasure to read and the examples are
clearly explained with good exercises at the end of each chapter. It uses
an extended version of Pascal developed specifically for teaching concurrent
programming, which is available at a nominal charge from one of the authors.
The book should be required reading on courses on operating and real-time
systems.

In addition to computer science students who do some numerical analysis,
there are many scientists and engineers whose courses include programming
to support the mathematical calculations in their main discipline. Computing
for ÐÓ°ÉÔ­´´s and Engineers by William J. Thompson provides a workbook
approach to a number of commonly occurring tasks. The 25 programs written
in C cover topics such as computations with complex variables, the fitting
of curves through data, and Fourier transforms. To help students who are
less familiar with this language the author has confined himself to a subset
of C, and provided information on how the programs could be rewritten in
Pascal or Fortran.

Finally we come to an area where it is almost impossible to find the
ideal book for student work. More and more applications are implemented
using powerful packages and programming tools, and students need to be exposed
to such systems and use them in project work. You only have to walk into
a bookshop to realise that there is a superfluity of self-teach and reference
books on all kinds of packages. Unfortunately, software is updated so frequently
that a conventional student textbook, written by an academic who has taught
the package for a couple of years, will be out of date before it is completed.
So it is often necessary to make do with books written for a different market.

As I am currently assessing Microsoft’s Visual C++ software development
package, I decided to try and find a book that I might recommend to a project
student who was familiar with C and knew how to use Mocrosoft Windows, and
who wanted to implement some good user interfaces. The most promising book
on the bookshelves was Mark Andrews’s Visual C++ Object-Oriented Programming.
It appealed to me initially because it introduces concepts such as object-oriented
programming in a clear way that would be acceptable in a student text. It
assumes that the reader has programming experience, preferably in C, and
the examples are based around an adventure game that might appeal to students
– if not the lecturer. Unfortunately, while the book will suit many who
want to teach themselves to program in C++, a close reading reveals that
the features of the package of greatest educational importance are not well
covered. Perhaps I was expecting too much.

Chris Reynolds is a computer consultant and part-time university lecturer.

* * *

Interactive Computer Systems: Videotext and Multimedia by Antone F.
Alber, Plenum Press, pp 453, $69.50

Computers and Language edited by Moira Monteith, Intellect Books, pp
159, £14.95

Human-Computer Interaction by Alan Dix, Janet Findlay, Gregory Abowd
and Russell Beale, Prentice Hall, pp 586, £19.95

Engineering and the Human-Computer Interface (student edition) edited
by Andy Downton, McGraw-Hill, pp 350, £35 hbk, £17.95 pbk

Hypertext: A Psychological Perspective edited by Cliff McKnight, Andrew
Dillon and John Richardson, Ellis Horwood, pp 202, £16.95

Software engineering: Principles and Practice by Hans van Vliet, John
Wiley, pp 577, £19.95

Image processing, Analysis and Machine Vision by Milan Sonka, Vaclav
Hlavac and Roger Boyle, Chapman & Hall, pp 574, £22.95

Pascal Precisely (third edition) by Judy Bishop, Addison-Wesley, pp
394, £18.95

Concurrent Programming by Alan Burns and Geoff Davies, Addison-Wesley,
pp 377, £19.95

Computing for ÐÓ°ÉÔ­´´s and Engineers by William J. Thompson, John
Wiley, pp 444, £38.50

Visual C++ Object-Oriented Programming by Mark Andrews, SAMS, pp 1098
+ disc, £36.50

Twenty years ago Gerry Weinberg published his seminal work The Psychology
of Computer Programming, and I looked forward to reading Psychology of Programming
to see how the subject has developed. However, in contrast to the exciting
and clearly expressed new ideas and often apocryphal evidence of Weinberg’s
book, we are presented with a stolid inward-looking academic tome, summarising
the findings of large numbers of carefully researched doctorate-sized projects.

Part of the trouble is that the two most important programming languages
designed to include psychological factors are virtually ignored. Cobol was
designed in 1960 around the assumption that making a program look like English
would help programmers to understand what they were doing. Despite being
considered by most academics as a decrepit dinosaur, Cobol is probably still
used by over 50 per cent of professional programmers. But it is scarcely
mentioned.

The other language omitted is Smalltalk. This language had a major psychological
input into its design, and led to the graphical user interfaces typical
of the Macintosh computers and personal computers running Windows software.
One reason for this omission is the authors’ assumption that programming
involves providing a computer with listable textual instructions. This excludes
many of the exciting new ways of instructing computers and, by taking the
historical definition of programming, gives the book a backwards looking
approach to the relevance of psychology to instructing computers.

This discouraging impression is compounded by the use of 19 different
authors, which gives the book the readability of a set of conference proceedings.

On the other hand, the detailed content of individual chapters is often
excellent. Many authors are healthily frank about the current limitations
in our understanding of how programmers think about their programs, which
explains why quick and easy solutions are not available.

The chapter ‘Programming Languages in Education: The Search for an Easy
Start’ contains interesting details of the recent work on Prolog. When this
language was first introduced into schools it was hoped that all the children
would have to do was to describe the problem that they wanted to solve –
and the answer would pop out of the ‘black box’. They would not need to
understand what the interpreter did when it translated their problems into
a form that could be processed swiftly, then retranslated the result into
an easily understandable style. Unfortunately, considerable evidence has
been accumulated that, while the language is very powerful and can be used
with great effect by some pupils, others seem to become tremendously confused.

I was pleased to see that recent psy-chological research on top-down
and bottom-up programming, which is described in the chapter ‘Expert Software
Design Studies’, confirms what I had always suspected – that most undergraduate
courses and textbooks are unreasonably idealistic in dealing with this subject.

Despite its limitations, the book will prove a valuable resource to
anyone interested in designing more effective programming languages. It
will also help educationalists to develop better programming teaching tools.

Chris Reynolds is a computer consultant and writer.

Could a computer believe that you believe that it is intelligent? The
whole subject of natural language, computing and artificial intelligence
is a minefield of contentious notions. The idea of a computer that could
believe in God, for example, is currently pure fantasy. Fortunately, Afzal
Ballim and Yorick Wilkes are well aware of the problems and do not claim
that their program ViewGen is in any way an ‘artificial believer’. They
point out that: ‘This is the only way Artificial Intelligence research actually
makes progress, very large-sounding titles, accompanied by very small, staggering
sleepwalking steps.’

The small, but engrossing, step they describe in this book relates to
the one of the enigmas of natural language understanding. When somebody
speaks, they not only have their own knowledge and beliefs, but also need
a model of the listener’s beliefs.

If a computer system is to understand language, it must be capable of
constructing such belief models. A hypothetical dialogue from the book illustrates
the point nicely: User: ‘Frank is coming tomorrow, I think.’ System: ‘Perhaps
I should leave.’ User: ‘Why?’ System: ‘Coming from you that is a warning.’
User: ‘Does Frank dislike you?’ System: ‘I don’t know, but you think he
does and that is what is important now.’

There are many problems along the way. Philosophers may argue about
‘intelligent’ computers for years to come, but language is, in effect, under
the our control. What words mean, and how they are used, changes. After
all, a successful natural language computer program will need the common
touch. If the average person describes modern computers as being intelligent,
the word effectively grows to encompass the new situation.

The book develops the concept of an environment that combines Mervin
Minsky’s theory of frames, that knowledge is divided into distinct fields,
with the models of believers. Within an environment, a ‘viewpoint’ is some
person’s belief about a topic. There is a problem in evaluating how this
should be used to update the existing viewpoints given a new proposition.
A totally zealous approach would mean examining all the implications of
the new information when it arrives, and could be called ‘the principle
of greatest effort’. A totally lazy approach would leave the relevance of
added propositions to be added on demand. The solution proposed in the ViewGen
algorithms is a sensible compromise between the two extremes.

While the approach uses a simple predicate logic formulation because
this is the lingua franca of artificial-intellingence research, the authors
point out that ‘there is (more or less) a conspiracy to ignore the ways
in which such a language can not possibly express either the complexity
of content of non-trivial beliefs or the relations between them.’

One of the resulting hurdles, the use of metaphor, is discussed in detail.
There is a question mark over the inheritance of properties and the identity
of objects. In a simple case, such as the metaphor of billiard balls for
atoms, one needs to decide which of the properties of billiard balls are
intended to be shared with atoms. In a more complex situation, a believer
may have confused Thessaly with thalassemia, thinking the disease is a province
of Greece. Given new information about ‘The cure for thalassemia’, a formal
system would detect an inconsistency because the word ‘cure’ is associated
with diseases and not places. However a human could rationalise the faulty
belief by reinterpreting the metaphor in a newspaper style as ‘the cure
for the political turmoil that beset the inhabitants of thalassemia’.

The book is well written, with excellent diagrams (apart from an unfortunate
printing error affecting Figure 1.6 and sequence), and takes a healthily
cynical stand in a field where hyperbole is far too common. I can heartily
recommend it.

Chris Reynolds lectured in computer science and is a science writer.

Pick up almost any issue of New ÐÓ°ÉÔ­´´ and you are certain to find
several items which, in various ways, demonstrate the problems that occur
with well-attached experts. For instance, the expert anatomist, Arthur Keith,
had the highest credentials, but helped Charles Dawson produce the evolutionary
answer everyone was hoping to find. It took 40 years before the Piltdown
fraud was exposed, and another 40 before Keith’s involvement was identified.

In 1989, the experts Stanley Pons and Martin Fleischmann, both from
respected institutions, produced an answer which proved too good to be true.
The result is that cold fusion has become another blind alley triggered
by genuine, but misunderstood observations, and has left an administrative
nightmare for those involved. More recently, Peter Duesberg, an expert molecular
biologist from the University of California, suggested on television that
the established theories about AIDS are fundamentally wrong – in ways which
other experts believe might adversely affect the battle against the disease.
And, of course, the Hubble telescope was put together by experts.

The trouble is that the public, and the politicians, want neat, tidy,
socially acceptable answers, but the very nature of science means that these
cannot always be supplied. In practice, much of the scientific edifice we
have built must be blemished. In the current ‘publish or be damned’ academic
rat race there is no kudos to be obtained for checking someone else’s research
results, or in publishing the reasons why an experiment failed (if you could
find a journal to accept such negative findings). As a result, much of the
fine detail of scientific research is never checked, while scientists repeatedly
waste time reinventing the equivalent of the square wheel in their particular
speciality because no warnings exist in the literature.

In many cases, the errors, misunderstandings and, in some cases, downright
fraud put about by experts have little or no effect. With the Piltdown skull,
the only real losers were the experts who were fooled by the deceit. But
in the 1930s, the fiddled results of the intelligence tests in Cyril Burt’s
study on identical twins fitted so well with the political and social climate
that they influenced the education of a whole generation of British youngsters
for good or ill.

In other cases, dangers arise from the failure of experts to ask the
correct questions, because the hazards lay in an unexpected area of expertise.
DDT was originally hailed as a wonderful pesticide because it did what was
asked of it. No one thought to ask what else it did until wildlife started
dying on a large scale.

Experts can exert unhealthy influence in other ways. The anonymous refereeing
system can mean that only conformist ideas are officially exposed in the
scientific forum. But this censorship by unnamed experts can prevent original
ideas from being openly discussed in scientific circles – and one must ask
what options the scientist with radically new ideas has to get his or her
ideas seriously discussed anywhere. For the frustrated researcher, talking
to the media must look like an obvious solution, and this may well be what
happened in the case of Duesberg. In such circumstances, it ill-behoves
the establishment, which has used anonymous experts to stifle open debate,
to object when the innovator seeks a wider audience.

There are also double standards in inves tigating what goes wrong, as
the contrasting cases of the Maguire Seven and BSE show. In the Maguire
inquiry, the key issue was whether explosive traces were found on people
arrested in connection with an IRA bombing. This involved questioning the
reliability of crucial laboratory tests; laboratory notebooks were scrutinised,
and the experts involved clearly identified.

In research, such detailed assessment of earlier results is rare, except
where there is the possibility of a lucrative bandwagon. With both cold
fusion and high-temperature superconductors there was a rush to check the
results of the experts who announced surprising results, with the aim of
trying to beat the originators to the fame and profits of successful exploitation.
However, in more mundane circumstances, there is considerable reticence
to carry out such objective reassessment, especially when something has
gone wrong which affects the public, as for instance with the bovine spongiform
encephalopathy scare.

The answer that society requires from its experts is that BSE cannot
pass to human beings from cattle. Meat eaters want to know that they are
not already infected and that it is safe to eat beef in future. Farmers
want their profits protected. Animal feed manufacturers may well be worried
about legal liability, and the Ministry of Agriculture, Fisheries and Food
clearly wants to be seen to have been safeguarding the public interest at
all times.

Unfortunately, the very nature of the disease makes it unlikely that
any absolute assurance can be given in less than the time an infection would
take to manifest itself in a human being, which might be 10 to 15 years,
or more. It is a very brave expert who categorically states that there can
be no risk whatsoever.

Like Fred Pearce, I find John Gummer’s statements about BSE unsatisfactory,
but for different reasons. The minister relies on experts (unnamed in all
the reports I have seen) who unanimously take the absolute certainty view.
Who are these experts? Does the anonymity indicate that they are civil servants
first and scientists second? Are they the same experts who recommended (either
by explicit action or complete inaction) that the way offal was used in
herbivore feedstuffs in the 1980s was safe? Were they involved in the earlier
decisions to cut back funds for agricultural research, and if so what were
their recommendations as to the likely effects of such cutbacks on public
health and public confidence? Without such knowledge, can we be certain
that the assurances they now give are not influenced by the desire to justify
advice they gave earlier? What is needed is a Maguire-type investigation
into the expert advice given to MAFF over the years, and how it was interpreted
by the politicians. Unfortunately, such an objective public analysis of
experts’ objectivity, covering as it would, the way experts are selected
and the politicians’ ability to understand their advice, is unlikely to
happen.

Let us remember that there are many socially important questions for
which science does not yet have, and may never have, an unambiguous answer.
In such cases, beware of any experts who dogmatically state that they have
the solution, however exalted their position. Above all, beware those who
quote the dogmatic views of the unidentified expert. It may be good politics
or public relations – but it is very bad science.

Chris Reynolds is currently working with CSIRO’s Division of Information
Technology in Sydney.

While this story may be apocryphal, modern communications are not as
confidential as you might think.

In November 1988, the Colorado City Council obtained a computer system,
including portable computers, which allowed members of the council to communicate
via electronic mail. At the beginning of this year, several of the councillors
became suspicious when the mayor, Robert Isaac, appeared to be surprisingly
knowledgeable on certain issues which they had discussed over the network.

It emerged that a secretary had been printing out copies of the private
messages the councillors sent to each other. This was done to allow the
mayor to discover what was being said behind his back! Needless to say,
a bitter political debate ensued. Isaac also decided not to run for governor,
but denied that this sudden decision had any connection with the scandal.

The incident will undoubtedly lead to extra income for the attorneys
of Colorado. Of course, the 1986 Federal Elec tronics Communications Act
means that the privacy of electronic messages should be protected. But Colorado
law requires that most City Council business be conducted in public, and
the mayor claims that he only wanted to make sure that the members of the
council were not using the system to conduct illegal caucuses.

The American love of the law reveals other methods in which ephemeral
keyboard conversations can be captured and read in ways in which the correspondents
had never considered.

For example, the participants in the Irangate scandal may have put the
more sensitive of their papers through the shredder, but they overlooked
the fact that their online dispatches were recorded on computer back-up
tapes. Then there was a large company in the US which had been involved
in a pollution incident. The firm was apparently not at all happy when a
court sub poenaed the ‘I told you so’ messages that, after the accident,
had flowed between the company’s experts on the internal electronic mail
system.

As a result of such events, some American companies are now ensuring
that all their electronic mail is deleted after two weeks to avoid any possibility
of it being used against them.

While Britain has not yet reached the level of litigious frenzy that
paralyses many activities on the other side of the Atlantic, there are hazards
for the unwary user of e-mail. Write a job reference for a colleague, or
comment adversely on the scientific abilities of an author when reviewing
a paper for a conference, and you may be subject to the Data Protection
Act if you retain a copy of the text in computer readable form.

This is because every individual has a right to see data held about
them on a computer. (It might be appropriate at this point to kill a popular
myth. Word processing is exempt only if text is kept on the computer solely
for editing and printing. If the text stored will be used in any way to
analyse or select information about persons named in the text, the act will
apply.) If you are not worried about privacy you can always broadcast your
messages on a bulletin board. Bulletin boards, and the more sophisticated
computer conferences, are excellent ways of communicating information between
dispersed groups of individuals with a common interest. The public ones
appear to be populated by a trusting group of people who believe in the
free exchange of information. This is an ideal to be supported, in theory,
by every scientist. However, it is not clear if everyone who reads, but
does not contribute to, a bulletin board shares this ideal.

Some people may well scan open bulletin boards to make money by using
the material posted in the public domain by more generous souls. Perhaps
I am guilty myself. After all, much of the information on which this piece
is based came from HICOM, a computer conferencing system, based in Britain,
for professionals in the field of human-computer interaction. My knowledge
of the Colorado incident came from the US-based RISKS bulletin board where
the facts had been copied (perhaps without permission) from The New York
Times.

Perhaps the biggest problem is that the communal approach of the healthy
public bulletin boards can be in conflict with commercially oriented operations
in which the control of information is linked with power and money. Bulletin
boards discussing topics at the frontiers of science may well have leeches
who log in, keeping their eyes open for original ideas, perhaps thrown out
informally as part of a general discussion, and for news of the activities
of their competitors. Even boards with closed memberships may be vulnerable
to voyeur hackers who are seeking ‘free information’.

Janice and her friend may well have been annoyed when their billet-doux
went astray, and the councillors of Colorado were right to be angry about
their ‘Big Brother’ mayor. The important thing to remember, whenever you
send off an electronic message, is that there are inherent conflicts of
interest: between the privacy of the writer and the rights of those written
about, and between the free exchange of scientific information and the capitalist
economy.

Bookbinder is a simple package which allows screens of text or pictures
to be linked together to form an electronic book. Making a new book is comparatively
easy. Pages are created with standard Archimedes software, such as !draw
and !paint, and are then transferred onto a plan of the book, where each
page is represented by a named box. You can then position the pages against
each other and draw links between them. For each page, up to five ‘buttons’
can be identified and associated with the links. So you can write a book
in several sections and join them up extremely easily.

Once a ‘book’ has been written, you can read it without getting tied
up with computer technology (apart from a problem with the size of the opening
page). Simply position the pointing finger over the chosen ‘button’, click
once, and another page appears. If you leave before finishing the book you
can leave a bookmark.

Unfortunately, the supplied examples are uninspiring. For instance,
the adventure story turns out to be a trip to the publisher’s shop, while
a puzzle is based on its telephone number. While the package does not have
the depth of a full hypertext system, it provides an excellent framework
for creating imaginative, illustrated electronic books in the classroom.

Genesis is a much more powerful software package than Bookbinder. It
comes with five initial examples, which include a demonstration of facilities
such as music and moving graphics, a diagram of the first two rows of the
periodic table, and a map of the European Economic Com munity. Click on
an element in the table, or a country on the map, and further information
is displayed. The Learning Curve disc (which is bundled with the model A3000
computer) contains nine further short examples that combine graphics, text
and sometimes music, including one on the planets and another on simple
environmental issues.

While Genesis has some sophisticated authoring facilities, its sophistication,
plus the lack of certain key features, limits its use for serious teaching
tasks involving a non-trivial number of hypertext pages. The trouble is
that it does not adequately separate and protect the pupil from the supporting
technology. It will delight the whizz-kids (of all ages) who like to do
clever looking things on a computer, but the aim of educational software
is to teach the chosen subject, not to pander to the potential computer
junkie. The makers seem to have given little thought to well-known hypertext
navigation problems. For instance, if you select the diagram of the inner
planets, select the Mars text, and then click the ‘next’ button – you reach
Jupiter without any warning that it is no longer an inner planet. The problem
is that, when reading a network of a dozen or more pages, Genesis passes
too much control to the Archimedes windowing package.

Hyperbook, which also runs on IBM and RML Nimbus computers, takes a
completely different approach. It does not really qualify as hypertext,
although it has some relevant features. Various hyperbook texts have been
published, and the three I examined each contained about 90 news items or
articles relating to science and technology. These were downloaded from
the Profile Information online database and excerpts reproduced with permission
from the Financial Times, Guardian, Washington Post and TASS, and (with
no evidence of permission) from New ÐÓ°ÉÔ­´´. Texts from New ÐÓ°ÉÔ­´´
include ‘Electricity: a private function that excludes research’ (29 October
1988) and ‘Designs on silk weave stronger fibres’ (29 September 1988).

These articles are presented in a most unattractive manner, with unadorned
virtually continuous text. Each book divides into about 120 pages, each
of which is an arbitrary two-and-a-half screens long. There is a single
blank line between articles, and not even an indentation to indi cate the
start of a new para graph. For all but the shortest articles, you cannot
see the title when viewing the later parts of the text. These limitations,
however, are the fault of the books, not the software, as the authoring
package will allow a more friendly presentation.

The saving grace is the powerful integrated text searching and indexing
facility that allows a hierarchy of topics to be defined – so that the heading
‘Materials’ contains the categories metals, ceramics, plastics, glass and
oxides, with metals being further subdivided with references to the text
on gold, copper, aluminium, platinum and titanium. It is possible to add
annotations at any point in the text (but not direct hypertext links) and
there is a facility for adding a glossary of terms.

Writing good hypertext teaching packages is a creative art form, communicating
ideas to pupils in a stimulating and easy to use manner. All three packages
would have benefited by the advice of a skilled hypertext author on the
human factors at the design stage. However, all have their good points,
Bookbinder is the most successful, because it does not try to be too clever,
I definitely recommend it. Genesis is ideal for linking multimedia applications,
but less suitable for easy-to-use hypertext teaching networks. Hyperbook
has some very interesting ideas but more thought is needed as to how it
will fit into the classroom environment.

Bookbinder costs Pounds sterling 55, from Musbury Consultants, 8 Fairhill,
Helmshore, Rossendale, Lancashire BB4 4JX. Tel: 0706 216701.

Genesis costs Pounds sterling 86.95 or Pounds sterling 69.95 with education
discount (exclusive of VAT). Learning Curve is bundled with the A3000 computer.
Software Solutions, Broadway House, 149-151 St Neots Road, Hardwick, Cambridge
CB3 7QJ. Tel: 0954 211760.

Hyperbook Reader costs Pounds sterling 49, Science and Technology hyperbook
Pounds sterling 25, single volume hyperbooks Pounds sterling 15 each.

Hyperbook Authorising System costs Pounds sterling 15 (prices exclusive
of VAT). Also available for IBM PC and Nimbus computers from Longman Logtron,
Dales Brewery, Gwydir Street, Cambridge CB1 2LJ. Tel: 0223 323656.

Chris Reynolds is a computer consultant and writer, currently based
in Australia.

GEOFF SIMONS has written books ranging from the strictly practical Introducing
Word Processing, through his provocative use of multi tudinous references
to the published computer literature in The Biology of Computer Life (Review,
21 March 1985), to the highly philosophical Is God a Programmer? (Review,
7 July 1988). So when I received his latest book I was not certain what
I would find. However, Viruses, Bugs and Star Wars contains elements of
all three of Simons’s approaches.

He begins by surveying the ubiquitous role of computers in Western society.
Factory automation and programmed financial dealings in the City are just
two of the examples he employs to show how human beings are becoming progressively
marginalised in important situations.

Simons then deals with the design and management of secure and reliable
mainframe computer systems in a methodical manner. Errors always occur in
the design and operation of computer software so he emphasises the importance
of establishing the criteria for both quality and for testing the products.
Fire and explosion are the commonest major disasters to computer systems.
Anyone reading the book will realise that the fire this March at the DEC’s
Customer Services Headquarters at Basingstoke was not an isolated event.

In the chapters ‘Malice and Mischief’ and ‘Virus, Worms et al’, Simons
covers the human threat to computer systems. But he offers short snippets
of information about many incidents in the style of Reader’s Digest, when
what we need is a more selective discussion of a few key cases. And we need
these cases put into context. For instance, the Equity Funding Life Insurance
Company fraud, involving $1.6 billion and 63,000 bogus insurance policies,
surely deserves more space than the account of a Devon bank manager who
destroyed Pounds sterling 737 of spoilt cheques in 1978.

Computer hackers are a growing army of mischievous people who treat
the security precautions designed to keep them at bay as no more than an
obstacle to overcome. While the author mentions the best-known cases of
hacking, and quotes examples from well into 1989, the extremely significant
1987 Appeal Court ruling in the Gold and Schifreen case has been inserted
as a brief footnote to his report, which is comparatively extensive, of
the original sentence that the appeal overturned.

Clifford Stoll’s work on tracking down a hacker who was looking for
military secrets, which he recounts in his book The Cuckoo’s Egg (Review,
3 March 1990) is mentioned only in the final ‘update’ section, although
an extensive account had been published in Communications of the ACM in
May 1988.

While I would be the first to agree that computers can have a significant
impact on both society and individuals, it remains important for commentators
to select suitable examples and present them responsibly.

It may be true that, in 1986, two patients died in the US because a
computer-controlled therapeutic radiation machine gave them the wrong dose.
Of course, this was a horrible tragedy for those concerned, but surely such
events should be put into context by pointing out that accidental deaths
from incorrect treatments were with us long before computers were invented,
and that modern treatment methods save large numbers of lives every day?
Simon fails to do this.

The problem with Viruses, Bugs and Star Wars is that it attempts to
cover too much. So the book falls between many stools: it is too lightweight
for a manual on computer security. The section on viruses is little more
than a journalistic rewrite of press cuttings and the social aspects need
to be put over with more passion to be convincing. I think it is unlikely
to appeal to the serious scientist.

Chris Reynolds is a director of Codil, a computing consultancy.

I clearly cannot take it all with me – but I really ought to get it
into some kind of order before I leave. But where shall I start? My files
of studies of local history should be relatively simple to sort out, so
they could be a good place to begin. Within minutes I find I have been distracted
into reading the notes I made nearly 15 years ago from great great-uncle
Robert’s archives, fortuitously saved from destruction by being deposited
in a county records office. His account of his grandfather’s involvement
in the bribery scandal of the 1802 Aylesbury election is great fun.

Suddenly I regain my sanity. I cannot read everything as I go. At this
rate it will take me at least five years to tidy up . . . I hurriedly stop
reading and resume the task in hand, this time on the materials relating
to my computer research.

A length of paper tape beckons and I try to work out what it was once
used for. It could be the program I wrote in 1971 to run on that horribly
primitive timesharing service whose name I have forgotten. Or perhaps it
came from that unsuccessful attempt to use the Modular 1 computer in 1976?

I open a cupboard. Some punch cards spill onto the floor. The large
cylindrical object is the replaceable disc pack from a 1970 period mainframe,
and all the information it ever contained should, in theory, be in one of
those reels of magnetic tape. But which tape is which? They all have numbers
on them – but where is the list? One has a label with my name on it, but
nothing else. Assuming the magnetic patterns on it are still readable, which
is unlikely, what computer could I use to read it? I don’t think it is from
the first computer I ever worked with, which was a Leo 3 in 1965. It could
be from a System 4 computer and hence be in an IBM format which may still
be recognisable. If it came from the ICL 1900 system I first used when starting
my university research I might be able to bribe the Science Museum to find
out what it contains, using the exhibit in their computer gallery. Unfortunately,
even if they gave me the opportunity, Sod’s law would ensure that their
magnetic tape deck is seven track, and my tape is nine track (or vice versa).

I glance towards the spot where I have buried the transcript of my great
great-uncle’s notes, and remember the happy days I had spent sieving through
the originals. His notebooks, the earlier engravings and later photographs,
the letters and the newspaper cuttings were all immediately intelligible.
In fact, apart from some minor yellowing, they looked much as they did when
he owned them in Victorian times. I had never thought of it before, but
I was very lucky that he had lived in a pre-computer age, when records were
all in human-readable form. Historians in a hundred years’ time will not
be so lucky. Anyone searching through the remaining records of late 20th-century
scientists will face some major problems.

The new dimension is due to the multimedia effect. Virtually all the
Victorian and older records are paper-based, and can be used by the ‘unaided’
historiographer. Then along came the gramophone, first with wax cylinders
of various types, to be followed by discs, which standardised on 78 rpm
for a while before moving on to 45 and 33, to be succeeded by cassette tapes
and compact discs. The moving picture shows a similar progression from primitive
films to video tapes and laser discs in various formats at various dates,
some of which proved ephemeral. At least these media have recorded information
that would otherwise be totally inaccessible, so I suppose historians cannot
grumble.

The introduction of the computer has been accompanied by an even greater
variety of media and equipment of an even more complex and fugacious nature.
In addition, much of the information is of a type that would, in pre-computer
days, have been preserved for posterity on paper. The scope for confusion
this causes is enormous. Over the past 25 years I have worked with main-frame
computers from eight different manufacturers, minicomputers from another
four companies, and microcomputers made by yet four more. As I am a computer
scientist this is higher than average – but the typical computer-using scientist
is unlikely to be using a single system for more than 10 years, and at the
present rate may well use a dozen or more different computers in a working
lifetime.

Let us imagine that the archives of some current scientist are put in
a trunk in an attic – to be rediscovered a hundred years after his death
– which is the way many interesting documents have survived in the past.
The antiquary of the 21st or 22nd century will be faced with an assortment
of punched cards, paper tapes, magnetic tapes and floppy discs in many different
sizes and formats. These will contain a chaotic collection of text files,
programs, commercial software, and coded data, and undoubtedly will have
been produced on a variety of different machines. For example, will the
interested biographer ever know whether the winchester on that broken portable
contained the hesitant drafts of a 20th-century magnum opus? In fact, it
seems likely that the only usable personal research records of present day
scientists will be those which have been printed out – assuming that the
paper originally used was good enough quality to survive.

I glance at my wastepaper basket. While I have been thinking about the
future I have continued to look through my own files. I notice that I have
just thrown away a pile of crumbling 20-year-old teletype listings (why
didn’t the university supply acid-free paper?) and some even more unreadable
offerings from a thermal printer I used in 1980.

Chris Reynolds a computer consultant who is about to take ‘sabbatical’
with the CSIRO in Australia.

The interface between people and computers is a key issue, and more
and more courses are including modules on the subject. Paul Booth and Yvonne
Waern take contrasting views of the subject in their two books. Booth’s
An Introduction to Human-Computer Interaction represents the technological
attitude to the interface. It starts with a good survey of the interdisciplinary
nature of the field, which has links with cognitive and social psychology,
artificial intelligence and software engineering. Booth then looks at the
hardware used at the interface, and the nature of dialogues.

Yvonne Waern’s book, which really falls into the field of cognitive
ergonomics, looks at the subject from a very different angle. Cognitive
Aspects of Computer Supported Tasks starts with the basics of cognition,
and then looks at the concepts of long-term and working memory in humans
– subjects not covered by Booth. The treatment of mental models is far deeper,
in the context of people learning computerised tasks. Waern concludes with
the proposition that we may have to revise our concepts of ‘thinking’ and
‘communicating’ as computers develop their capacity for ‘information processing’.

If I had to choose just one of these books for a course, I would select
Booth’s volume for teaching psychologists; while Waern’s text would be excellent
for making the technology-oriented computer scientist realise that it is
important to understand the people who use computers.

One of the most readable introductions to the field of artificial intelligence
is Computers and Thought by Mike Sharples et al. The authors assume no mathematical
knowledge and illustrate the problems with excellent examples. I particularly
like the thought-provoking discussion about the Eliza package that carries
out ‘conversations’ with people. Elsewhere, the authors use the design of
an automated tourist guide of London to illustrate many of the ideas being
discussed. For example, in the chapter on vision they look at the way in
which a computer could recognise a sign for London Transport. An optional
disc contains the software.

I can heartily recommend Ken Pedersen’s Expert Systems Programming as
a well presented account of the use of expert system packages, with the
minimum of hype. He covers the basic building blocks and the inference process;
then des cribes handling unknown and multivalued information, to gether
with the limitations in various ways of handling uncert ainty. Finally,
he looks at topics such as knowledge acquisition, prototyping and fine-tuning.

These last two books provide easy, practically oriented introductions
to artificial intelligence and expert systems. Many final-year courses,
however, will need a more mathematical treatment, and guidance on how to
build the relevant software. Artificial Intelligence and the Design of Expert
Systems is a possible answer.

In this, George Luger and William Stubblefield introduce the basic concepts
needed, such as predicate calculus and heuristic search. They then examine
Prolog and Lisp as languages for problem solving, and follow this with chapters
on rule-based expert systems, knowledge representation, natural language
and automated reasoning.

Computer graphics is another area where spectacular advances are being
made. Peter Burger and Duncan Gilles have written a heavyweight text that
is aimed at final-year undergraduates with a sound knowledge of mathematics
but which should be of value to anyone programming advanced graphic applications.

Interactive Computer Graphics starts by looking at the basics of graphics
hardware and software, then moves on to fundamental algorithms for raster
graphics, including anti-aliasing used to minimise the jagged edge effects
that can occur on plotting straight lines. The next three chapters delve
into the analysis of two- and three-dimensional figures.

One of the most impressive features of modern graphics is the use of
colour shading to produce realistic looking surfaces. Burger and Gilles
give the appropriate mathematics, including that for ray tracing. Finally,
they examine special effects, animation and hardware for interactive graphics.

Hypertext is the name given to databases of texts, and sometimes pictures,
which contain active links that allow the reader to explore the subject
along a variety of routes. Such databases can be of particular value in
teaching, Hypertext Hands-On! represents a printed version of such a database,
by Ben Schneiderman and Greg Kearsley.

The package includes an IBM-formatted disc containing the database,
which allows the reader to use hypertext itself to learn about hypertext.
This works well (apart from a bug in the set-up routines under certain circumstances)
and the student can compare the merits of the printed page and the interactive
database. My own reaction is that the printed page is more appropriate for
systematic learning to an examined syllabus, while the computer version
is better for casual self-instruction. But Hypertext Hands-On! is an excellent
example of the way textbooks may develop.

Computer Systems Development: History, Organization and Implementations
by A. L. Friedman with D. S. Cornfield, Wiley, pp 437, Pounds sterling 35

An Introduction to Human-Computer Interaction by P. Booth, Lawrence
Erlbaum, pp 283, Pounds sterling 9.95 pbk

Cognitive Aspects of Computer Supported Tasks by Yvonne Waern, Wiley,
pp 352, Pounds sterling 29.95

Computers and Thought: A Practical Introduction to Artificial Intelligence
by M. Sharples, D. Hogg, C. Hutchinson, S. Torrance and D. Young, Bradford
Books, pp 414, Pounds sterling 22.50 hbk, Pounds sterling 13.95 pbk, disc
Pounds sterling 14.25 + VAT.

Expert Systems Programming: Practical Techniques for Rule-Based Systems
by Ken Pedersen, Wiley, pp 315, Pounds sterling 20.15

Artificial Intelligence and the Design of Expert Systems by G. F. Luger
and W. A. Stubblefield, Benjamin Cummings, pp 680, Pounds sterling 19.95

Interactive Computer Graphics by Peter Burger and Duncan Gilles, Addison-Wesley,
pp 520, Pounds sterling 19.95

Hypertext Hands-On! An Introduction to a New Way of Organising and Accessing
Information by Ben Schneiderman and Greg Kearsley, Addison-Wesley, pp 182,
Pounds sterling 22.45 + VAT, including disc

Chris Reynolds gave up lecturing in computer studies to become a director
of CODIL Language Systems, a computer consultancy

XSEL’s Progress: The Continuing Journey of an Expert System by E. Mumford
and W. B. MacDonald, Wiley, pp 257, Pounds sterling 19.95

A WORKING party of the Council for Science and Society produced Benefits
and Risks of Knowledge-Based Systems. It alerts the public to the potential
benefits and possible dangers of advanced information technology, in a very
readable manner. The book identifies seven common characteristics of expert
systems, including solving difficult problems, heuristic reasoning, functioning
on data that contains errors, explaining what they are doing and justifying
their conclusions.

The authors then point out that this is really a list of wished-for
properties rather than a description of actual systems. These may perform
well only in limited domains, and show only some of the desired characteristics.
However, most existing expert systems contain at least a knowledge-base
of facts, a set of rules that operate on the facts, an inference system
that chooses the rules to be applied and a facility to explain to the human
user why it has made a particular decision.

In assessing attitudes to the technology the authors split people into
three categories. The ‘believers’ see huge potential in the approach that
will provide us with better ways of sailing the ever-rising sea of information
that threatens to flood over us. Expert systems will manage certain crises
where humans have difficulty in coping, such as emergencies in a nuclear
reactor.

The ‘social critics’ do not deny that there can be benefits but point
out the dangers. For instance, a future system ‘could appear to have a fairly
subtle command of natural language within the limited exchanges for which
it was designed. Many users might therefore assume that it has a complete
command of that language’. Last, there are ‘disbelievers’, the best known
of whom are the Drefus brothers, who argue that, while such systems might
perform competently in some narrow domains, they will never become the equal
of human experts.

The working party concludes that there are benefits, but dangers also
exist. It recommends strongly that codes of practice are necessary.

In contrast, Enid Mumford and Bruce MacDonald’s book, XSEL’s Progress,
relates to a single expert system called XSEL, short for eXpert SELling.
DEC Computers developed it as an aid for sales staff to ensure that individual
orders, involving many components, can easily be drawn up to meet each customer’s
requirements. It has proved extremely successful in simplifying the ordering
and supply problems of marketing customised computer configurations. The
book describes how DEC succeeded not in terms of hardware or software but
by reviewing the people who made up the company, and the way that they were
managed.

For anyone interested in the management (or history) of the development
of complex systems, this book provides a useful case study of how one successful
– and frequently cited – system came into being.