• Contact humanITy
• ATcare (external)

• Home
• What We Do
• Who We Are
• Engage Our Services
• Contact
• Articles & Papers

Criteria for the Establishment of a Knowledge Management System and their relevance to the proposed London Sustainability Exchange (LSX)

Return to Information Design

This Paper was commissioned by the London Sustainability Exchange (LSX) and, subject to humanITy's payment conditions being met it is and will be the intellectual property of the LSX. HumanITy, however, reserves the rights to extract general principles from this Paper and integrate them into other work, subject to such material having no reference to the LSX. Should humanITy wish to quote from passages relating to the LSX it will seek copyright permission in the normal manner. This Paper is compatible with Michael Hayes Paper The London Sustainability Exchange: Technical Scoping Document V 0.1 of 23rd August 2001 except that this Paper is far more cautious in its assessment of the need for information interchange and educational functions.

Date: 28/08/2001

---

Article

1. Who and What is it for?

Introduction

The first and most obvious questions about a knowledge management system (KMS) are what is it, who is it for and what is it for?

A KMS is a corpus of data, rendered by organisation to present information which is then edited and organised to achieve a specific purpose.

The immediate, instinctive answer to the other two questions is likely to be that it is for as many people as possible to achieve as many outcomes as possible. Where it cannot provide this through its own resources, the system will establish one-way or reciprocal links to other information providers; it will be comprehensive, multi-faceted and yet easily navigable.

If this is the end of the matter then it will end up being none of these things. The great advantage of a digital KMS is that it can be constructed in modules and because technology is changing rapidly, because information changes and because priorities change, it should be constructed in a modular fashion. A digital KMS is provisional but forever.

It follows from this that the information, the basic building blocks, should be as high quality as possible but that the provisionality requires constant maintenance and upgrading. It is therefore vital to devote resources primarily to information quality and a clear, basic design which allows for modularity and the minimum upheaval when the KMS orientation has to be changed.

What is it for?

Modularity implies priority and, therefore, choices. In the case of the LSX, for example, a key initial choice will have to be made as to whether the primary function of the system, at the time of construction and into the next two years, is to engage academics and policy makers or to trigger grass roots activity; does LSX want to:

• convince a relatively few policy makers of what they must do in order to create the conditions for grass roots action; or
• trigger grass roots action as the result of facilities made available by policy makers?

Again, the natural answer to that is that they are symbiotic and a KMS must do both; but in establishing frameworks, coding HTML, tracking sources, relating data and maintaining all of this you either have to make stark choices or you simply pick up randomly relevant information, classify and display it.

Who is it for?

The purpose of the KMS is closely related but not co-terminal with the audiences at which it is aimed. You might opt from 1.2 for policy makers or for grass roots activity but you still have to decide whether or not the LSX wants to interest journalists in what you do; or you may want to encourage or be neutral about people outside London who might just be interested in what you are trying to do.

It is therefore vital that a primary and a small number of secondary audiences are identified. Again, because the KMS will be modular the priority can be altered as policy changes but even if you can design a KMS for all your central objectives you will not be able to design it to be equally accessible, usable and of interest to all audiences.

User Choice

It is technically possible to design a KMS so that people can allocate attributes of all sorts to themselves and in time that will lead to heuristic middleware which recognises a user's attributes because it has 'met' the user before and monitored behaviour but for the time being all these presentation, taxonomical, sequential and hierarchical options are very resource consuming.

Scaleability and Granularity

There is also a great deal of difference between theoretical and practical scaleability and granularity. In theory you can take any digital system and combine it with another, in parallel or reciprocally and repeat the process indefinitely without adding any layers of complexity but in practice digital scaleability does not work. The more complex a KMS - just like an analogue or hybrid KMS - the more layers it needs and the greater the number of connections between nodes. Granularity depends critically on the way that the information was assembled and classified. Referring a user to a Home Page of a massive site does not provide the kind of service that a specialist should offer; at the very least major references should be to site pages and major documents should be properly marked up so that portions of them can be found easily. All this, however, takes time and money.

Brokers and Portals

Even a superficial analysis of the knowledge sector shows that the creation of content is not keeping pace with connectivity. There is a great deal of value in establishing metadata and links, being a broker and facilitator, 'adding value', but the leverage industry will be one of the first casualties of any economic depression. If cash runs short it is the brokers who will go first; those making product will survive the longest. Until that unhappy day there will continue to be intense competition between would-be brokers. First it was who will portal the sites; now it is who will portal the portals?

These observations may well be counter intuitive to an organisation that wants to be open, comprehensive, inclusive, catalytic and facilitating but the Internet is no different from analogue life. In a political system the Secretariat might be very fine but what counts is the quality of the membership; taxonomy and co-ordination are admirable but a new and vital insight is much more valuable; synergy is usually the product of convergent strategy rather than 'fishing expeditions; and focused energy creates fire.

Combining the above - KMS modularity; purpose; audience; survival; and utility - they point to a sharp focus on a carefully selected cluster of issues, the relationships between them and the way in which they can command attention and action.

This short, somewhat ruthless, Section must simply reinforce the general warning. The Internet is already defaced and defiled with junk, duplication, the wrecks of enterprise and the detritus of faded dreams. Let us be pioneers in digital ecology.

2. Principles for the Organisation of Information

2.1 Although the concept of hypertext as a method of cross referencing and allocating a multiplicity of attributes to a single piece of information, as opposed to tree taxonomy developed by Aristotle, information was generally realised through 'frozen' artefacts such as the book, the picture or the statue. The only significant exceptions were scored music and scripted drama which both rely on a considerable degree of intermediary activity. Even films and television programmes were largely regarded as 'finished'. The first change came with the birth of multi-track sound recording. A multi-track record was the musical equivalent of a multi-stranded plait or braid. Extending the metaphor, most previous information had been rendered like an omelette or a pot of mixed paint, it could not be separated into its initial components. Sound and film editing and collage became increasingly popular in the 1960s and 1970s, creating mixtures of information from a variety of sources creating something of a fruit salad; you could separate the pieces but you did not know about the extent of the different sources. Meanwhile, to complete the picture, Leibniz's principles had at least been adopted in the analogue world for indexing and bibliographic database creation which delivered parallel strands of information of which you could trace the full extent.

2.2 Digital information differs from analogue information in the following key ways relevant to the design of a KMS:

• Style can be separated from content - A web browser interprets style but does not necessarily render it in precisely the way in which the author set it down. The exactness depends on the detail in a specified style sheet but most transactions between a creator and a consumer are not governed by a mutually employed style sheet. In a complex information array it is particularly important to agree on hierarchical and heading conventions. In this document, for instance, a deliberate decision has been taken to limit heading levels to three:
  
  • Title
  • Section x.
  • Paragraph x.x
  together with a single list classification
  • bullet *
  for ease of reading. Such conventions as x.x.x.x have been eschewed. This is a case where a balance has been struck between readability and searchability.
• All elements can be labelled and separated - As long as every piece of data is given precise attributes it can be 'lifted' from an information mix. The greatest single error in creating digital information is to make it 'flat' and label free, the digital equivalent of an omelette or paint mix. To extend the pain mix analogy, your information object should not be the flat oil painting of a child, it might be layered, as in Vermeer or pointilist style like Seurat. Conventions must be agreed on attribute specification/taxonomy/ medium/source, etc.
• Information can easily be displayed by attribute - In the analogue era it was easy to display information according to two attributes in, for example, a matrix or a graph with a horizontal and vertical axis. With a certain amount of ingenuity a third dimension could be shown aslant from a rectangular matrix; and some adventurous folk played 3-dimensional chess in a cube 8 x 8 x 8! Digital information can be oriented in a massive variety of ways as long as its attributes and their relationships are carefully defined. Even if tools are available to re-orient information in a large variety of ways, care needs to be taken so that the user is not confused. One of the key discussions in this Paper will be attribute selection or taxonomy.
• Publishing can become narrowcasting - A book or magazine, a television or radio programme, an LP or a cassette tape, were all delivered 'frozen', taking no account of the preferences of the consumer who could either take it or leave it. Digital information can be designed in such a way that users can manipulate it without overturning the original file structure and content. Tools mean that information can be allocated attributes such that it is accessible in different ways by different audiences, so the attribution is not just intrinsic, it can be externally driven. This also means, incidentally, that tools can enable information users to alter both the style and content of information, altering print size and font, foreground and background colours, frame delivery rate, audio speed and pitch, length and complexity. These will be important in the consideration of tools integrated into the data.

2.3 "Because it is there" and because information designers come much more often from an artistic/aesthetic background than from a philosophical/taxonomical/information scientist background, the default method for displaying information in digital format is moving from chronological or alphanumeric to spatial. Some of the options for a KMS are:

• 2.4 Cartographical
• 2.5 Geometric
• 2.6 Genre
• 2.7 Audience
• 2.8 Chronological
• 2.9 Alphanumeric
• 2.4 Cartographical Arrays

There are three main classes of cartographical arrays:

• Cartographical - information placed on a to-scale map, e.g., Ordnance Survey, London A - Digital maps can be simplified and expanded/contracted while keeping their scale
• Stylised Cartographical - information placed on a representation of a map where scale is sacrificed for clarity, e.g., the classical map of the London Underground which conveys information much more clearly than a map to scale with the lines following their true course
• Cartographical Ranged - information placed within a designated area like a Ward or a Borough but not in correspondence with the exact location on a map drawn to scale.

Of these, the most viable is the second. Even simplified maps drawn to scale are very confusing for orientation and navigation. The data is sometimes hyped up on map reading skills and its possible gender divide but it is very clear that many people find map reading very difficult. The problem with the third option is that jurisdictions are often not co-terminal and which ever you chose, people often do not know where they live.

2.5 Geometric

Geometric arrays have been popular since the development of calculus and the theory of sets. They have, traditionally, been lists that work in two directions at right angles but digital information allows much more complex data arrangements.

• Aesthetic - this is the simplest form of array where the information provider selects a number of attributes and arranges them aesthetically according to taste and where the physical relationship of the attributes has no significance.
• Taxonomical - this is where the information is consciously classified before attributes are assigned; again, the spatial relationship is usually not explicitly significant though information of this sort is assumed to be ranked from top to bottom and left to right.
• Matrices and Tables - Classically these are most familiar in the spreadsheet with two axes at right angles, one for dates, the other for amounts. They are also very useful for relating any two attributes such as location/sector (Does x have a y?) or quality/price (What is best value?).
• Circular - Information classes can be placed on the circumference of a circle with the unifying attribute at the centre. This is particularly useful if there are many unifying attributes with identical characteristic attributes, e.g. a large variety of towns which each have a re-cycling facility, cycle paths, pollution monitoring and water sampling.
• Proportionate - the most popular proportionate arrays are bar chars and pie charts which explain quantity.

The general rule for all geometrical arrays is that a balance must be struck between completeness and complexity. In finding a given piece of information there is also a balance between the number of options presented and the number of sets of options to be navigated before information is found, the links versus clicks trade-off. If information is to be platform independent then the maximum number of pieces of information offering a choice on any screen is limited to nine so that it can be chosen by a telephone or remote controller keypad. There will be further comment on this in Section 3 on interoperability.

The first option, the Aesthetic is almost always a wasted opportunity, particularly if the information is being manipulated without a mouse; it simply wastes time without facilitating a choice. It is the option beloved of most Web designers and is to be avoided, particularly on a Home Page.

Taxonomical arrays are particularly useful but should most strongly adhere to the limit of nine (The Power of Nine, Carey & Stringer, 2000).

Tables and Matrices are obviously extremely useful for this Project as it allows questions like: "Does this place have a re-cycling facility" as long as you can define place (as in 2.4 above) and as long as you know how to define your need clearly; you might not know what a re-cycling facility is but have friends who regularly go to the "bottle bank" which is why Section 4.7 deals with language. It also allows attributes which might be in inverse proportion (quality/price) to be balanced; in this case that might be useful for cost/benefit analysis of a re-cycling process. The key problem with tables and matrices arises when more than two dimensions are required, e.g. what? Where and How much? Here, the first resort should be to 'bundling' two attributes into one, so that Where? and What are at right angles but How Much is notionally in a third dimension at right angles to the other two but is actually shown in a matrix cell, e.g. there is a price in the cell of the matrix where the What row corresponds with the Where column. Another strategy is to try to rank the attributes on an axis which involves 'off screen' calculations which you might want to display elsewhere. It depends on what the information is for and who it is for and we have already discussed this in Section 1.

Bar charts and pie charts are excellent on their own but there is a strong temptation on the part of designers, which must be resisted, to use the segments as clickable links to other information; without good labels this is useless for non-mouse users.

2.6 Genre

Traditionally it was easy to rank information by its appearance. It was, for example, easy to distinguish between a learned book, monograph, journal, periodical, pot boiler, broadsheet newspaper or tabloid newspaper. It was also easy to assign a value too information according to its provenance, Government, broadcaster in a duopoly, advertising flyer, second hand car sales pitch.

Digital information is much more difficult to classify. Even with strong branding the source and the value to be assigned are difficult to calculate. There is a fuller discussion of the assignment of value in section qqq but, in the meantime, it is possible and right to assign information a value according to the processes it has undergone, regardless of what you think of the outcome. So, for instance, you might rank information as follows:

• Peer reviewed academic research
• Abstract of the above
• Research and analysis from an acknowledged authority
• Analysis and summary by reputable authority
• Comment by reputable authority
• Summary/comment by interested parties
• Summary/comment by general publication/broadcaster.

An information system that has to cater for a wide variety of people may wish to use such a classification system to sequence information on a topic. As academics are used to abstracts and most people want to start with a summary, I would start with Abstract unless the system is supposed primarily to report on press comment or relay a controversy. The only important note is that because an abstract is short it does not necessarily mean that it is easy to read; many only achieve condensation through the use of jargon and difficult words. Academic work or official reports without an abstract is more or less useless in a general, sectoral information system because newcomers won't read the full text and those who are highly specialised will know it already.

2.7 Audience

There is obviously a relationship between genre and audience. Academics tend to read academic abstracts and papers, policy makers will start with abstracts but will want frameworks, internal and external drivers, costings, good practice and contacts. Casual searchers will want a light touch, crusaders will want tracts and the media will want controversy.

Although tools are very good at re-purposing information and although elegant sequencing can get people rapidly where they wish to go, again the key point is to work out a primary and a couple of secondary audience segments on whom you want to use most of your resource. Rather than arrange a system according to audience so that the user can define which audience he is in and select accordingly, the system should be built with a very specific audience in mind with facilities to cater for limited sub audiences.

One of the initial papers for this Project mentioned African Caribbean elders; this is the kind of Time Out approach to social issues that gets information systems into a tangle. Cross-referencing and inter-twining is theoretically simple but very complex in practice. If the general considerations of Section 1 are accepted, the maximum connectivity should be a list of like-minded agencies such as Amnesty International, Liberty, Commission for Race Relations, etc.

2.8 Chronological

Because of our associative ability - to remember one event because of its temporal coincidence with another - chronological listings can be useful; they are particularly helpful in narratives and campaigns.

2.9 Alphanumeric

Finally, the most obvious arrangement of data is still Alphanumeric; it is the method most people most easily understand without having to think about it and it conforms with most hard copy data. Most people will be dealing with hybrid information systems for some time, including the LSX itself, and so a certain amount of information will have to be stored Alphanumerically.

2.10 Note on Trees

Even though it is the least flexible way of structuring data, the Aristotelian tree thrives, not least in metadata menus. Its fundamental problem is that an element of information can only be classed in two ways: first, as one element of all information 9which is in many ways meaningless); and secondly, as information with a single attribute, the tree equivalent of a twig. this is exactly the wrong way to classify information for the purposes of multi-attribute hypertext.

2.11 Databases

Behind every KMS lies a database. The way that the information is arrayed must strike a balance between the usability of the interface and the effectiveness of the database. Again, the theory of database effectiveness somewhat outstrips the practice. It is now theoretically possible for a data mining system (such as Autonomy) to sift huge banks of data, lift relevant pieces and drop it into a database field. In practice, the construction and maintenance of a good database is the key human resource and financial cost of any sound KMS. Emphasis is often wrongly placed on front end and interface design but the power of a system is in its back end.

Because the SLX KMS has some clearly defined parameters - London, Sustainability - it should be relatively easy to organise and manage but that will only decrease the taxonomical problems; editing, labelling and inputting will still be labour intensive.

3. Technologies and Interoperability

3.1 Criteria

The audience and objectives criteria are key determinants of the kind of technologies that will be used to access the LSX KMS. It is safe to assume for the foreseeable future that policy makers, opinion formers and academics will use a hardware device not very different from the current PC or Apple Mac. It will still have access to powerful processing - though that may be server rather than client based - and as broadband becomes more common the power and speed of the system will increase. Although some people will use voice-in technology, the majority will still use a QWERTY keyboard and, therefore, they will be able to use keywords searching rather than being limited to numeric access. If the KMS is purely for this audience then it should be heavily oriented that way, designed for people with keyboards, high syntactic and lexicographic skills and able to handle a variety of navigation systems dictated more by the nature of the material than by the need for simplicity; they will also have no problem with listed items, e.g. chronological and alphanumeric listings.

3.2 The PC/Apple Audience

A major development which must be taken into account is the growth in interoperability protocols which enable devices to exchange or share information without the need for cables or even the use of software transfer protocols. This means that somebody may well keep the bulk of 'favourite' information on a server in a series of bookmarks and may wish to access it at work via a desk top device, at home via a domestic digital manager or between the two on a mobile device such as a Third Generation Mobile Telephone or a Personal Digital Assistant (PDA). At some point in this set of transactions screen size may be critical. This is not a problem if text is to be scrolled but it is a problem if data selection relies upon a set of options inside a frame which cannot be seen on a small screen but cannot be expanded on that screen. In general it is always much easier to alter the size of data than of metadata.

If the audience is limited to that cited above, then the temporary interoperability requirement for the mobile device should be accorded a very low priority.

3.3 Interoperability Criteria

Interoperability is required in geometric proportion to the forecast size of the audience and increases by another geometric factor as the income and social class of audience segments becomes 'lower'. In other words, the more people you want to reach the more likely they are to want to access information through digital televisions and mobile telephones; and the poorer they are the more likely they are to have access only by digital televisions and telephones.

If one of the primary audiences is expected to involve high usage of telephones or televisions then there are two possible interactive interfaces:

• Numeric keypad
• Voice In.

3.4 Numeric Keypad

There is a well-known psychological rule that the optimal number of elements in making a choice is 7 + or - 2. Fortunately, this accords with the familiar numeric keypad which allows a maximum of nine options plus the use of zero for "Other" and the star and hash keys for "Forward" and "Back". This interface obviously requires a taxonomy based on nine or fewer elements in any array. In terms of the effectiveness of the implementation of the optimal choice algorithm (7 - or - 2), designing screen arrays as if they were to be accessed by telephones is a very good discipline, particularly for delivering short, sharp pieces of information. It is not a good way of delivering complex or lengthy information for keypad users because most digital televisions and telephones will not have a printer.

3.5 The Power of Nine, The 6561 System

The maximum number of pieces of data that can be realised from a 'Nine' specification from three clicks is 6561. The number falls if any array presents fewer than nine items but it falls less the further down the chain the maximum number is not used, i.e. if the initial array displays eight rather than nine items, the total possible number of items in the system will be:

8 x 9 x 9 x 9 + 5832, or 88%

whereas if one of the arrays in the 'lowest' set only contains eight elements then the maximum possible number of elements is

6561 - 1 + 6560.

3.6 Data Bifurcation

It is not difficult to work out a compromise between a 6561 'narrow' system and a much wider system. Just as a retail outlet roughly runs on the 80/20 stock rule (80% of sales come from 20% of the available items), so the "Other" option at in the first array of nine should take users to a different dataset which has the less popular/current items in it. There will be occasions when items have to be moved from the 6561 frame to the "Other" class because of crowding, changing priorities, legacy data, an archiving function, etc.

In summary, whether the primary purpose of the KMS is for a limited audience or for grassroots membership, a key task is to separate the 'stock' and to render the most popular items into a 6561 system, organising the balance according to less structured criteria, particularly if many of the "other" items are part of an archive where indexing and chronology become more important. Information selection criteria for archiving may be quite separate from those for the 6561 site but this is outside the scope of this Paper.

3.7 Stock Taking

'Stock' taking should be achieved through a tool in the system so that usage dictates, to a certain extent, the bifurcation of 'stock'.

3.8 Voice In

Not for the first time in the Paper, we have to make a clear distinction between theory and practice. In theory 'Voice in' technology can deliver a lexicographic recognition of the whole of English, together with associated proper names. In practice, such systems only work with a good deal of systems training.

Most voice driven systems work on micro vocabularies; the use of these does not require systems training but it does require customer training. The smaller the vocabulary the greater constraints placed on users to work within the rules. Little is known about the 'fuzzy' aspects of voice in technologies. As long as these are absolutely isomorphic with numeric arrays it is possible to run a fairly reliable system, e.g. "Press or say". Such systems rely upon using the best elements of Web design "Go forward/go back" with telephone menu systems "You have six choices".

In summary, voice in systems should only be considered on the basis of the 6561 system above and must be isomorphic with the on-screen presentation. (Isomorphic, not identical because the screen array may have non essential detail, labelled as such, which is sloughed off when the material is automatically re-processed through XML for telephone use).

3.9 Digital Television

There is still some work to be done to create seamless interoperability between digital television and the World Wide Web but these should be viewed as background problems which will be solved by underlying standards and tools and should not affect the LSX KMS. One element, however, which needs to be watched is the customisation range offered by television compared, say, with Microsoft NT. If an information system is devised approximately according to the range of NT but television is much more rigid in its rendering (e.g. limiting text to one standard and one 'accessible' font) then design may have to be modified to meet the lower accessibility standard. this is, however, likely to be a very short term problem as broadcaster, particularly those selling on a pay per view basis, will increasingly more towards middlecasting and then casting so narrowly that entertainment menus and renderings will conform with the needs of the individual user.

3.10 Public Access Terminals

Interoperability protocols and the growth of digital television access, not to mention the increasing use of 3-G mobile telephones, will largely render public access terminals redundant. The idea that a large minority of the population, primarily the poor, will wish to use Post Offices and other access points to transact business, particularly with the Government, was based on a number of fallacies; that:

• The interface paradigm was the PC
• Declarative transactions and financial payments would be parallel
• Communities would be willing to use terminals in public places.

As we have seen, the PC is no longer the interface paradigm. The European Union Directive on electronic currency is likely to lead to funding transfers by telephone which will, in turn, lead the Government to make payments to individuals by telephone or television terminal rather than in public places. Research has shown that the people who most need community facilities are the ones that use them least. There is also growing evidence that the public does not want to use public terminals for transactions it regards as personal or confidential such as health and finance. Public terminals will, of course, be ubiquitous but they are likely to be relatively 'walled gardens' for specific services (Government transactions at Post Offices, timetable information at railway stations, store guides) that are accessed via wireless through portable devices carried by end users. The wider 'surfing' function will be undertaken from the portable device through the telecommunications service, not through a public access terminal.

3.11 In summary, no matter which audience remit is chosen as the primary focus, a data bifurcation model (paragraph 3.5) is recommended for optimal interoperability.

4. User Requirements

4.1 Having considered a KMS from the point of view of initiators in general and LSX in particular, we need to look briefly at user requirements before returning to some factors in management in general and finally to some specific LSX issues.

There is a vast literature on user requirements and human computer interface (HCI), so this Section will simply list and comment on some key aspects. These are:

• 4.2 Customisation
• 4.3 Accessibility/usability
• 4.4 Key Words/Thesaurus
• 4.5 Navigation/Orientation
• 4.6 Hierarchy/Ranking, by supplier, by user
• 4.7 Lexicography/Language
• 4.8 Links
• 4.9 Look and Feel/House Style
• 4.10 Aesthetics.

4.2 Customisation

Customisation covers a vast range of potential alternative settings and renderings of data. The absolute minimum requirements for it are:

• World Wide Web Consortium Accessibility Guidelines (www.w3.org/wai) although these are far too oriented towards text and static graphics access by blind and visually impaired people.
• Disability Discrimination Act (1995) (United Kingdom). There is no case law on this Act which will help to establish user requirements over and above those listed above by WAI but the situation must be monitored.
• QWERTY and Numeric access (see Section 3).

The demographics of ICT customisation are complex but in general the requirement both in quantity of audience and range rise with age.

4.3 Accessibility/Usability

Accessibility and usability are not precisely interchangeable terms but they refer to the same concept. The first is generally used by campaigners to establish a rights-based case for information, the second is usually used by designers to define a market. Both are determined to a great extent by interoperability and customisation. Accessibility, however, is sometimes confined in meaning to physical and economic access to hardware and in turn that usage is frequently conflated with "Availability". Usability also encompasses broad concepts such as the ability to understand the author's intentions, orient and navigate, and transact efficiently. Accessibility will increasingly become a legal requirement and usability is a fundamental market requirement. Both should be considered optimally rather than absolutely on the basis, where legislation is not involved, of cost/benefit.

4.4 Key Words/Thesaurus

A key words searching function which allows the user to select categories of information, over-riding the designer's arrangement, is a key user requirement, particularly for QWERTY keyboard researchers. This requires, at the very minimum, the allocation of key words to all information. This requires the designer to indicate allowable key words which must be attached to all information that is to be searchable. These allowable words should be displayed at the starting point for the search. They may need to be underpinned by a Thesaurus-like function which 'translates', for example, ecological into green. On the basis of paragraph 3.5 above, the user should be able to choose between an alphabetical list and a hierarchy based on the 6561 system. To allow "other" in a relatively small KMS is a taxonomical cop-out.

4.5 Navigation/Orientation

The more complex a system the more difficult it is to navigate. A brief reference was made to this problem in paragraph 2.4 above. Many people have problems navigating in two dimensions; the classic case is the, probably apocryphal, story of the Nobel Prize wining physicists who could not walk South in a strange town with a North-oriented street map. Navigating within a website involves more than two dimensions but the size and orientation are not obvious. It is easy to get lost and vital to be able to retrace steps. This is why the structure should be simple, clear and uniform with each 'page' of a system containing a common set of links, including one which returns to the Home Page.

4.6 Hierarchy/Ranking, by supplier, by user

Every information system with more than one element requires a hierarchy, even if it is only "left/right" or "top/bottom" choice.

In digital, public information systems a hierarchy can be established in two ways. Information can be ranked according to the priorities of the producer or the consumer. The two are not mutually exclusive. A designer, working for a producer, can set access hierarchies according to major priorities or issues that are topical. He can, at the same time, use tools to monitor which topics or pages of information are most frequently accessed on a daily basis.

The LSX should primarily be producer driven with a lower level facility for user priority. This might be established, for example, through the producer establishing a 'frame' of nine major priorities with use then determining in which order the nine will be ranked. This is a useful concept in determining the order in which options are presented in a telephone menu. If an issue ranked by the designer as seventh suddenly becomes topical, as reflected in access to its pages, the designer or, better still a tool, can 'lift' the item up the rankings.

4.7 Lexicography/Language

If the primary audience for a KMS is academics and policy makers then lexicography is not an issue. It can be assumed that the user will have no problems either with a large, general lexicographic range not with specialist vocabulary. Lexicographic range, however, is an issue if the primary audience is grass roots activists and people wishing to gain information which will determine individual action, like re-cycling waste or using cycle paths.

The size of a KMS will determine the way in which lexicographic range is handled. In a small system information can be made accessible to various categories of users through the manual re-purposing of the basic text through providing abstracts, summaries, simplifications, FAQs and short lists of bullet points. In a larger system the information must be labelled while it is being created. In standard word processing headings can be labelled for tables of contents and key concepts can be labelled for indices. This process needs to be extended in optimising access to texts by designating passages in a ranked order, e.g.:

• Crucial
• Explanatory
• Illustrative
• Incidental.

Précis engines can shorten and simplify files but their rules are more syntactic than value driven.

This is a resource hungry process unless information providers create their content according to a recognised template that gives a common set of priorities and procedures to the operation.

Ideally, too, language engineering tools should be developed which rank words according to degrees of difficulty but this is so complex that it is much better to run files through complexity checkers to establish the 'reading age' of the content. These tools are not particularly reliable where specialist, micro vocabularies are invoked but they are a cost-effective general guide.

Ironically, although language engineering tools for operations within a language are relatively primitive, tools for translating between languages are becoming much better developed. On the Home Page it would be possible to provide a link to language options which would then allow a user to find a document and send it to a remote site for translation. This is not ideal because it requires the non English language user to resort to an intermediary in the first instance but it is the only viable option at this point.

4.8 Links

As we shall see in Section 5., one of the key parameters of a KMS is the kind of 'frame' which is put round it. No KMS built for the web can be totally self contained and insulated if it has a single link because that link may provide other links which ultimately provide access to the whole mass of web content. As links are both desirable and unavoidable, the credibility of the KMS and user orientation both require reciprocal linking so that LSX receives more visitors and so that users who start out with the LSX can find their way back to it from a site at first remove. More than that cannot be expected.

Although individual files in the LSX KMS may provide automatic links to other sites, LSX should provide a listing of officially accredited linked organisations with which it has reciprocal arrangements.

4.9 Look and Feel/House Style

Far too much is made of the 'look and feel' and house style of the digital KMS. As will be seen in paragraph 4.10 below this is largely a marketing question. What is important, however, is that the KMS does have a uniform layout, as far as that is possible given the heterogeneity of the information, and should have common on-page navigation and orientation facilities.

Where possible, repetitive, non-essential information should be excluded so that, at the very least, it is automatically excluded from small screen access. Clearly the page reference is important but as long as that contains the "LSX" address, an additional banner and logo is not needed on every page.

4.10 Aesthetics

The aesthetics of a digital KMS should depend upon whether it is primarily intended to gain 'sales' from loyal customers and referred customers or whether it wishes to attract 'passing trade'. Most website designers, for instance, assume that they are attracting passing trade. This arises because of the myth of surfing. Most web users:

• 'click through' to a site from a related site
• initiate a search; or
• log on to a specific site.

In all three cases the 'passing trade' metaphor and the design that goes with it are not relevant.

5. Managing the KMS

5.1 So far we have looked at criteria’s for constructing a KMS but the way in which the design affects operation and maintenance requires some consideration.

A vital distinction in managing a KMS is the extent to which organisational users and external users are ranked. Many systems start out being oriented towards outside users but over time the KMS narrows to organisational needs.

In the case of the LSDX this is not such a critical problem as organisational and outside users will have a great congruence of interest and purpose. Nonetheless, it is important to be careful of organisational, as opposed to sectoral, jargon/argot. Nowhere is this so bad as in general statements of vision, mission and values. These may be required in strategic planning and form filling but you will be known by your brand rather than your prospectus.

In managing the KMS, therefore, it is vital that clarity of purpose and outcome is kept at the forefront. The following topics have been chosen in order to assist with KMS focus:

• 5.2 Communicating Purpose
• 5.3 Establishing Rules
• 5.4 Maintenance
• 5.5 Selection Criteria
• 5.6 Templates
• 5.7 Excluding
• 5.8 Adding and Upgrading.
• 5.2 Communicating Purpose

Following on from comments about general statements of vision, mission and values, it is important for a KMS to tell people what its purpose is, what makes it a KMS rather than a database or an information deposit.

The whole theme of this Section will be that a KMS must be dynamic and that applies to parts of it which are generally thought to be static, such as why it is there. Are you trying to:

• Communicate
• Educate
• Persuade
• Archive
• Generate
• Broker.

the mix will surely alter over time; so must the statement of your purpose.

5.3 Establishing Rules

Contrary to popular opinion shared, apparently, by social and free market liberals, people are not good at handling open-ended choice. They need to know what the rules are so that they feel comfortable in your KMS. However well you and a neighbour design a KMS they will not be identical, so people need to know what the important rules and features are of your KMS. These should at the very minimum include:

• Navigation/route maps
• Restricted areas and reasons
• Charges
• Copyright
• Parameters for inclusion and exclusion of data
• Weighting and value of data
• Status of recommendations
• Status of partners
• Etiquette
• Contact details and restrictions
• Feedback, comment and complaint procedures.

Try to avoid saying what your policy is and indicate what you provide. Don't, for example, say you have an information accessibility policy is, point to the link to customisation.

5.4 Maintenance

The most important criterion for assessing the viability of a KMS is its maintenance requirements. The budget for this needs to be agreed and broadly allocated before the establishment and early stages criteria are finally agreed.

Broadly, the following algorithms apply:

• There is a direct relationship between quality and control, not because information provided by third parties is not valid but because without a degree of control it will not comply with important aspects of the KMS such as its taxonomy and attribute labelling.
• Equally, there is a direct relationship between control and cost; the more control, the higher the cost.
• Conversely, there is an inverse relationship between control and size; the less control, the bigger a phenomenon can be for the same cost.

Maintenance should not be considered as a technical adjunct not integral to a KMS operation. The cost saving in contracting maintenance out of house is rarely real. The savings invariably go on core staff time spent in briefing and correcting the external contractor who is not an organic part of the organisation's growth. This is why so many externally managed systems become 'flat' and formalistic; they lose their muscularity. An in-house manager should be much more concerned to maintain the system in accordance with the spirit of the organisation and not simply with the data specification. If the personnel involved in the KMS consider themselves to be information scientists or technicians apart from the central mission of the organisation then the system will fail. This is much more often said than believed and much more often believed than enacted.

Maintenance should, equally, be an organisational responsibility, discussed both at management and Trustee/board level.

5.5 Selection Criteria

The selection criteria for information within the KMS should, as noted in paragraph 5.2, be clearly stated. These should be of three sorts:

• Conformance with the KMS framework
• Quality
• Topicality/Archive.

In this context, conformance is judged against the taxonomical framework and the target audience(s). Quality refers to whether the material achieves the purposes of the author, it does not refer to how valuable the author or user thinks the information is which is a matter of genre, discussed in paragraph 2.6 above. Certain material may also be required because, regardless of the other two considerations, it is topical or fits a quite separate archiving criteria, referred to in paragraph 3.6 above.

5.6 Templates

The most important aspect of a template is taxonomy/labelling which conforms as closely as possible to the host KMS. Style and extent is much less important. Genre is desirable.

Of course there is a trade off between simplicity and completeness in template design. With a small KMS simplicity is the most important of the two because in a small system the intense labelling will be carried out in house. In a large KMS completeness is more important because the organisation cannot afford a huge labelling exercise.

5.7 Excluding

No KMS can be unlimited in scope and so a first and important criterion is the scope for the KMS as a means of excluding information. This, needless to say, should not be invoked dishonestly to exclude data which would fall inside the scope if it met other criteria.

Inclusivity and brand identity and integrity present a potential conflict. Unless there are very good reasons for an exception, brand and integrity should always be given priority over inclusion. There are almost 200,000 Registered Charities in the UK, not to mention unincorporated associations, NGOs not Registered as charities and millions of businesses. LSX, even if it is not paying for information, should consider itself to be a buyer not a free archive because of the resource it will need to prepare and maintain all the information in its KMS. Even if it is paid for including information which falls inside the scope of the KMS it should remember to charge full not marginal cost, particularly if continuous addition adds to the layers of maintenance and organisational complexity.

There should be a written, public policy about scope (paragraphs 5.3, 5.5) which should include a generous but firm statement on exclusion.

5.8 Adding and Upgrading

Better to dismantle a site than leave it to rot. Because competition is intense, attractive, user friendly, high quality, focused KMS web sites with new things to say are going to have an advantage over the competition.

The cardinal rule for any system that is required to grow is that it must be designed in such a way that the growth can occur without upheaval. For example, if you have a ranking system that is numeric and you want to add a new classification between 3 and 4 then all the numbers from 4. to the end have to be altered. The system should therefore be designed to be organic and not to fit inside a container. Admittedly, this goes against the recommendation in paragraph 3.6 but the 6561 system should be the specially accessible, cross platform, first port of call but even there the addition of a single clock would increase the potential units of data by a factor of 9 from 6561 to 59049. The 'back' part of the system, on the other hand, can be indefinitely open.

There is a current fashion, partly dictated by cash-flow limitations, to phasing development. In the area of KMS design this is an unhelpful idea because the phasing is normally an anticipatory exercise whereas modularity is an organic idea. Each time a module is added it will have implications for the whole KMS, triggering 'vertical' alterations and at some point the modular additions will trigger a 'vertical' layer of metadata and some reorganisation. If, however, the initial taxonomy is sound there will be a great deal of room for entering data into fields without any alterations of the structure of the KMS.

If there is a God she is a taxonomist.

6. Specific LSX Issues

6.1 The following issues have been raised in Papers submitted to the LSX or in the Steering Group Meeting on 30^th July 2001. These are:

• 6.2 Economic Models
• 6.3 Brokering/Adding Value
• 6.4 Features of Sustainability Model
• 6.5 Audience; the Critical Question
• 6.6 Elaborating and Simplifying
• 6.2 Economic Models

Although it is strictly outside the remit of this Paper, may I offer a few thoughts on economic models.

The economic model is not only affected by the cause or product, it is also affected by the audience or audiences. The less wealthy and/pr influential an audience, the less the likelihood either of advertising, commercial sponsorship or revenue from usage. This may turn out to be yet another reason why the primary audience for the LSX should be academics and policy makers. A grass roots advisory service might be started in a separate and somewhat cross-subsidised fashion. Certainly, if a grass roots activist service is funded by an NGO for a period it will have to become sustainable through its own activities or cross-subsidy. I suggest the latter.

Personally, I believe that the Internet advertising model is extremely fragile. Even in the analogue world the advertising budget is the first thing to be cut in an economic downturn. The alternative is sponsorship but the major corporates who are capable of finding budgets for activities like the LSX are very few and most of them are engulfed with requests for assistance. The competition is fierce and in spite of any protestations about 'sustainable' capitalism, my guess is that they will mostly want to promote capitalism with a slightly less restrictive definition. Some might even regard the LSX as anti capitalist.

The picture is not as bleak as it might seem. A large number of major organisations are publishing Corporate Social Responsibility (CSR) reports alongside their Annual Reports. They might well wish to be associated with LSX in some way but it seems to me that it is much better if this is reciprocal, halfway between sponsorship and consultancy, where CSR statements are assessed against LSX standards; the company gets a mark for good behaviour in the LSX KMS and might even be demonstrated as a model of good practice and the LSX receives sponsorship. I recognise that this leaves the organisation open to some suspicion but as long as it only takes sponsorship from those who pass its tests (and it is hardly likely to receive offers of any from those who don't), it should be reasonably secure from adverse comment.

Another important strategy is to look at major organisations which might be near or just outside the standard, indicating to them how little they need to do to come within it. There is no point being critical without offering a route to success. Such a route must be concrete not abstract.

It is best to see consultancy and sponsorship as the extreme ends of a spectrum of relationships.

If none of the above is successful then charging is inevitable if the intention is to run a KMS over a sustained period in order to provide public access. From the point of view of sponsors and donors, it is not the role of an NGO to make indiscriminate subsidies to the whole planet. The ideal system, particularly for regular, small-scale use is to invoke a micro-payments system similar to that used by long distance telephone companies but this is complex. A system split between 'tasters' and pay-per-use is another option but the best available at the moment is what we might think of as Membership. This requires a degree of maintenance but if it can be marketed and managed effectively it will be more cost effective than seeking outright donations from NGOs and corporates.

Finally, a possible source of start-up finance, though a messy source over a sustained period, is the grant making trust sector.

6.3 Brokering/Adding Value

With reference to paragraph 1.6, we have to sort out some terminology. As I understand it, a broker does not change the nature of a product but simply collects and distributes it. This is a valuable service in the data field because sorting and structuring is entailed. This is one of the things LSX may well do. Adding value requires more pro-active packaging and, in this case, enabling users to see structured information in the context of a highly specified framework. The brokering function requires systems and tools but adding value requires a high degree of creative intelligence. It involves "Ways of Seeing". In a KMS the primary way of seeing is the taxonomical function which implicitly connects by applying selection criteria. This, I understand, was the major outcome of the discussion on 30th July 2001 even though it was imprecisely articulated.

Although there is intellectual property in the framework, the prism, the filter through which information is analysed, the main source of added value is in its application. Unless the framework is copyrighted rigorously it will be regarded as public domain information. Adding value will have to be dynamic and it may mean adjusting the framework. This is tiresome but in a world, for example, where the use of nuclear energy has shot up the political and ecological agenda within the space of a year, analysis will change.

I understand the attempt to separate the framework from outcomes which must be determined by people but simply through the processes of its construction a framework is not a neutral artefact.

The difficulty of this discussion points to the need for an absolutely clear what it is we are trying to do.

Michael Hayes, in his Paper of 23^rd August 2001 Technical Specification V0.1 raises two additional, pertinent issues for consideration: first, whether LSX should be a 'light touch' broker of interaction (described colloquially as a 'chat room' system); secondly, whether LSX should have a specific educational function.

I have no doubt that the first should be rejected because unmediated interaction tends not to achieve any objective. Mediated transactions, on the other hand, are costly in direct proportion to the quality of the mediation. With respect to his second suggestion, an educational function should be kept separate for exactly the opposite reason, that is, it is much easier to raise funds for specifically educational purposes and, therefore, these should not be entangled with campaigning or simply information provision.

6.4 Features of Sustainability Model

The Features of Sustainability framework is set out as a tree with sustainability as its trunk, five branches with respectively, 3 3 4 2 and 1 sub branches.

My analysis of the 12 elements is:

• Natural 1 - Fundamental
• Natural 2 - Fundamental
• Natural 3 - Fundamental
• Human 4 - Derivative
• Human 5 - Tangential
• Human 6 - Tangential
• Social 7 - Tangential
• social 8 - Tangential
• Social 9 - Fundamental
• Social 10 - Derivative
• Manufactured 11 - fundamental
• Financial 12 - Fundamental*

(*This last is fundamental but in a weaker sense than the other Fundamental factors.)

This produces six fundamental factors, to derivative and four tangential.

I would be inclined to prioritise these as fundamental, derivative and tangential in any hierarchy; and as the whole tenor of this Paper is that the KMS should start modestly and identify a highly focused brand, I would, with some regret but firmly, list the tangential factors in associated values but no more.

The justification for retaining the derivative factors is that they allow people to associate behaviour with consequences which do not all flow from the fundamental to the derivative; although poor health is caused by poor ecology, poor government causes poor ecology. this last way of looking at the problem, however, would force a hierarchical ranking with governance above ecology. This may be logical but all agencies have to decide what they can change and what they cannot. The LSX might, in a total framework, put itself in a structural hierarchy below London Governance and put that below UK Governance and so on but that is hardly an effective brand attribute. This is why 'framing' the KMS is so important and why the central activity must be at the head of any hierarchy.

If these principles are broadly accepted then the other elements of the taxonomy can be settled, though I would be inclined to look at concrete applications no further than first 'click' level; people who are opinion formers but not academics and policy makers think of factories, trains and rivers rather than emission, consumption and pollution.

6.5 Audience; the Critical Question

After this discussion, we are back to the critical question from Section 1, what is the primary audience for which the KMS is initially to be designed? I think that the answer, conclusively, has to be for academics, policy makers and opinion formers. This, it seems to me, presents the best starting point from every angle, not least the economic.

The LSX should then consider a mutual arrangement with an action-oriented organisation to work with it to construct an action-oriented structure reflecting the initial LSX classification of good practice, etc. It should also consider a parallel initiative to finance an educational function.

These conclusions, I recognise, are painful and are in considerable conflict with models with which the LSX is familiar and/or comfortable. I am particularly aware of the commitment to education/training and the development of human capital but that interest is not unique to LSX.

6.6 Elaborating and Simplifying

If there is any consolation, it is that it is much easier to move from the complex to the simple than to elaborate the simple and render it complex. That is both a reason for adopting the recommended course of action and an encouragement that it provides a better path for expansion than starting with the activist audience and somehow transforming the KMS to suit academics, policy makers and opinion formers.

Return to Information Design