Postscript: On being a ''knowledge engineer''

(1) ''He only listens to what other people have to say and writes it down''. (2)''He has that peculiar knack of making himself an expert in anything''.

Diverse opinions from the London School of Hygiene and Tropical Medicine.[lr][][+8] "He only is wise who knows that he knows nothing."


As this book was going through the Press, I was told ''You are not an ''editor', nobody sends you completed chapters. You are that scarce and precious species, a knowledge engineer''. Although I was aware of expert systems, I had not previously heard this term applied to their creators. Knowledge engineers are people who question experts and express the expertise they obtain in a computer program that can be used by non-experts. These computer-based expert systems do complex inferential reasoning based on a wide knowledge of a limited field, and are finding increasing use in such fields as ''fault-finding on 11GHz radiowave equipment', interpreting mass spectrographs for chemists (''DENDRAL'), and prescribing antibiotics (''MYCIN'). Their role in medicine is limited but is growing.

Although [f10]Primary Surgery [f09]appears on paper rather than on a screen, is less logically sophisticated, and covers a much larger field than do computer-based expert systems, the similarities between them are striking. Like other knowledge engineers, I have had great problems in getting knowledge out of experts. Real ones are hard to find, and when you have found them, they may only be master of a small field, and be so busy that they can spare you little time. If an expert is going to contribute efficiently, he has got to know his field (many do not), he has to to be able to exteriorize his knowledge (many find this difficult), and he must be willing to collaborate (many are not).

An expert often forgets what he does, and may not know what he does. There is indeed no logical reason why anyone should be aware of how a thing is done in order to be able to do it. Even when an expert can describe what he does, he can be wrong. He is more likely to be able to remember actions given conditions, than conditions given actions[md]expert surgeons know when to operate, but have difficulty listing the indications for doing so. They need cues which a knowledge engineer has to supply. An expert is often better at criticizing someone else's ideas than explaining his own, and may only express his knowledge in response to something he disagrees with. Knowledge engineers have to learn the expert's language: in doing so I became a particular kind of ''theoretical' surgeon, anaesthetist, and obstetrician.

Expert systems require constant refinement and debugging: I worked mostly by asking experts to comment on innumerable drafts assembled from tiny fragments of knowledge. As one expert said when I began, ''You will have to build it up comma by comma''.

Looking back, it is remarkable that the task was accomplished at all. Only by combing the earth was it possible to find just enough appropriate experts. The task would have been easier had it been possible to start with this postscript, instead of ending with it. Paradoxically, any merit in these manuals lies with the experts (''How could a non-surgeon know this...?'), and any faults with the knowlede engineer[md]it is his job to spot the fault and patch it with another expert's knowledge. The sixth sense that he needs to develop is to know what knowledge is useful, and when it is likely to be faulty.

From an expert's point of view, working with a ''knowledge engineer' is an efficient way of producing a book. If an expert looks at a draft and passes it with only minor changes, he might have written it. Yet he will have done so in only a fraction of the time that it would have taken to write in the ordinary way.

So, if you are an expert and read this, look kindly on any ''knowledge engineer' who approaches you; understand his difficulties, and do your best for him, since this may be the way in which the more useful knowledge sources of the future will be ''engineered'. As we go to press the EEC is sponsoring the development of a medical computer work-station that will support a variety of expert systems. Perhaps a future edition of [f10]Primary Surgery [f09]will appear on a screen? Meanwhile, there is need for a ''knowledge engineer' to work on [f10]Primary Medicine. [f09]Who would like to give a substantial part of his life to this task?

Here are two books covering the same field, written at the same time in the ordinary manner. How do the two approaches compare? The third, on which this section is based, describes the travails of us knowledge engineers.

Anaesthesia at the District Hospital. Published by WHO. Available in English and French. Price Sw fr 20; US$ 16.00. Order No 1150289. WHO Distribution and Sales, 1211 Geneva, Switzerland, and all WHO sales agents.[-3]

General Surgery at the District Hospital. Order No 1150300. Price Sw fr 30; US$ 24.00. Available in English and French. Published by WHO as above.[-3]

Welbank, M. ''A Review of Knowledge Acquisition Techniques for Expert Systems. 1983, British Telecom Research Laboratories. Martlesham Heath, Ipswich, England.[+15]

                              Maurice King. MD, FRCP, FFPHM
                                             Reader  in
                                    Public Health Medicine.
                                       University of Leeds.