Book of the week: Not Exactly: In Praise of Vagueness

John Gilbey is reassured that not always having an answer is normal

In an increasingly complex digital world, it is tempting to get sucked into the precision - often a wholly spurious precision - that seems endemic to this culture. I recently caught myself doing it, despite my own distinctly analogue leanings. When asked for the time at a bus stop last week, I found myself replying "Seven forty-eight" - instead of the answer I would usually give: "Just after quarter to eight." I put this momentary lapse of good taste down to early morning low blood sugar and the disabling effect of public transport, but it is a symptom of something much bigger - a subject that has captured the (not insubstantial) imagination of computer scientist Kees van Deemter.

In this engaging and approachable book, we are asked to consider the importance of the concept of "vagueness" in our lives. Take a look at the previous paragraph, for example: what did I mean by "early morning" or "low blood sugar" or a "disabling effect"? These things are "vague" in the sense that they are tricky to define adequately and could be subjected to a wide range of interpretations. Yet they can have an extraordinary and profound impact on our lives and our environment, especially when we try to build software systems that can take account of them.

After an introductory chapter - in which, interestingly, this magazine gets a namecheck on the first page - van Deemter divides his work into three sections. He begins by discussing areas in which vagueness seems hard to avoid, pointing his philosophic spotlight into some of the darker areas of measurement and description such as obesity, poverty and intelligence - all areas where it is difficult to arrive at explicit definitions and where it is even harder to avoid being contentious. He even ventures into the vexed question of "Munro bagging" - the sport of climbing all the Scottish peaks over 3,000ft high - as an example of an arbitrary measurement and boundary.

A welcome thread running through the book is the use of entertaining dialogues - of the Platonic style - to illustrate some of the more convoluted concepts. This works well and adds significantly to the clarity of the argument in a number of cases. One particular example that will doubtless strike a chord with academics everywhere is the post-interview discussion at the (apparently ubiquitous) University of Poppleton.

The complex relative merits of three candidates are being discussed, with "Human Resources breathing down our necks", around a whiteboard. Different mechanisms for the evaluation of vague criteria are attempted and abandoned with what to many will be a heart-warming familiarity. It is always reassuring to be shown that there is not an obvious answer to many common problems; or, as the author puts it through the voice of one of the protagonists, "when things differ along several dimensions, comparisons between them get tricky".

Vagueness in identity is also addressed, in a dramatic extension of the "grandfather's shovel" concept - where the apparently historic shovel has had two new blades and four new handles. The example van Deemter uses is that of a 1920s racing car that gloried in the name of Old Number One. This car became the subject of a High Court case in 1990, after it was bought - as original - for a large sum of money. The purchaser, we are told, questioned whether this was really the original car, and expert opinions were sought. This opened up all sorts of interesting questions regarding just how much something can be changed - for whatever reason - before it ceases to be "authentic", "genuine" or "original".

In the next section he moves on to tease out some logic from our love of vagueness - bringing in some of the bigger guns of reasoning to discuss the true meaning of vague expressions. In this he uses the powerful toolkit of paradox as an enabling force, which puts the discussion firmly into the realm of philosophy but does so in such a relaxed and genial manner that the lay reader can generally be carried along, rather than scared off, by the matter presented. The layout of the chapters is such, however, that if you do feel you are losing ground mentally - or possibly losing the will to live in some of the more strenuously argued sections - you can skip between passages without missing out on the overall message.

There are big concepts discussed in this text, all of which impinge on our everyday experience. This is perhaps most obvious in the context of processing vague information, such as the perception of tallness and the understanding of colour. Van Deemter draws us through those areas where precise reasoning can perhaps ultimately be achieved - and prepares us by considering the importance of context in such analysis - before moving forward into discussions of the role and drawbacks of such mechanisms as fuzzy logic. Along the way we are treated to crash courses in classical logic and context-based reasoning.

In several of these areas, I was grateful for the comprehensive notes appended to the main text. At some points, however, the notes started to become a text in their own right, and I wondered whether it may have been better to include more of the additional material in the main body to avoid the to-ing and fro-ing that inevitably results from such a construction.

The final section builds on the author's journey to look at the role of vagueness as a strategic element in communication, and how the emerging artificial intelligence (AI) research community has approached the management of vagueness in the first half-century or so of its existence. This adds significantly to the discussion, by drawing into focus just how complex conceptually are the behaviours we are trying to mimic through the use of AI. Perhaps inevitably, the Turing Test is used as the measure of whether human responses have been successfully emulated - but van Deemter uses the device to good effect, which is far better than the mere mention it sometimes receives.

I would suggest that the concept of vagueness is set to be of huge significance in the next few years, as developers and providers vie for a share of the imminent augmented-reality software market. I suspect that the products and services that prove ultimately successful will be the ones that have fully accommodated vagueness as discussed here by van Deemter - thereby delivering a truly human-friendly interface between the virtual and the real world.

The style throughout is amusing, persuasive, conversational and engaging, but this does not detract from the thoroughness with which van Deemter approaches his thesis. Irrespective of the way the material is packaged, however, this text builds into an interesting mix of reasonably hard-core philosophy and computer science. While this realisation may initially slightly unsettle the non-specialist, it is a combination that should be applauded - and is one that you should expect to see a lot more of in future.

Not Exactly should appeal equally to students of philosophy and computer science, as well as any lay reader who wishes to develop more than a casual understanding of some of the factors influencing the development of artificial intelligence applications. I found it especially reassuring to realise that vagueness is an inherently human attribute - a badge I will henceforth wear with pride.

THE AUTHOR

Kees van Deemter is a reader in computing science at the University of Aberdeen. He is interested in getting computers to speak and write, and in the logical, linguistic and philosophical questions that this raises. Paramount among these are the puzzles posed by the real or apparent flaws of human languages, namely ambiguity and vagueness.

Originally from the Netherlands, Dr van Deemter did his PhD at the University of Amsterdam and was a postdoctoral researcher at Stanford University. He was a tenured research scientist at Philips Electronics and a Principal Research Fellow at the Information Technology Research Institute (University of Brighton) before taking up his current position at Aberdeen in 2004.

When not at work, he is a devout follower of FC Barcelona, which he says is "unambiguously the most watchable football club in the world".

Background information about the book can be found at www.csd.abdn.ac.uk/~kvdeemte/NotExactly

Not Exactly: In Praise of Vagueness

By Kees van Deemter Oxford University Press

357pp, £16.99

ISBN 9780199545902

Published 28 January 2010

  • Print
  • Share
  • Save

Related images

  • Cover
  • Print
  • Share
  • Save
Jobs