KNOWLEDGE MANAGEMENT SYSTEMS 1. INTRODUCTION # Four activities of KM: » create / discover knowledge » capture / codify / organise knowledge » share / learn knowledge » apply knowledge # We need technological and organisational systems to assist in each activity. This lecture looks at the kinds of technological systems that might help each. # Each activity presents issues that challenge us. These are discussed at start of each section. 2. TECHNOLOGICAL SYSTEMS TO HELP KNOWLEDGE CREATION/ DISCOVERY 2.1 Issues that Challenge Knowledge Creation/ Discovery » Seeking knowledge from elsewhere: Data mining and similar technologies » Helping people express complex knowledge: Mind mapping » Uncovering (tacit) knowledge: MAKE (multi-aspectual knowledge elicitation) » Stimulation to fill gaps or even create new knowledge: Assistum 2.2 Data Mining and Similar Knowledge can be obtained from external sources (usually the World Wide Web). This can be done by people reading it, assimilating it and then bringing it into the organisation's KM store. But there are also ways of getting knowledge by computer programs. Techniques include: » data mining » report mining » web mining » screen scraping » and more. To learn more: see the Wikipedia pages on each of these, and then obtain and read the references given on those pages. What follows is a brief overview of each, with some of the challenges that each meets. Data mining is the term given to sifting through large volumes of data to extracting meaningful patterns therefrom. It is often used in fraud detection, surveillance, and scientific discovery, which do not come under 'knowledge management', but it can also be used as a KM tool for obtaining knowledge from data. Data mining is good at analysing data from behaviour, and can sometimes uncover hidden patterns that which are not obvious. In this way it can sometimes result in finding tacit knowledge. Data mining can also be useful during the Knowledge Application stage. Given a set of data (which is often a pre-processed subset of the large dataset), the steps of data mining include: » Find groups in the data. This can be done either on the basis of known categories (such as spam versus genuine for email) or by a method that tries to find similarities (which can include neural nets and case based reasoning). Usually this and the next step are performed on subsets of the whole data, because the algorithms used are time-consuming. » Find a function or model that best explains the groups of data. Probably the most common method is regression analysis. Another method is association rule learning (e.g. supermarket might use this to find which products tend to be bought together). » Validate the function or model. This usually involves testing against the rest of the data (if a subset was used for finding it) to see if it fits. Often the function or model is found to be wrong. So another attempt might be made perhaps with another subset. » The function or the model given by these methods is the knowledge that has been extracted from the data. Data mining can be used on business data, scientific data, Internet traffic data, and spatial or geographic data. In the latter, it is often used in conjunction with geographic information systems. The main challenge to data mining is the volume of data to be mined. The time it takes is proportional to the square, cube or fourth power of the number of pieces of data, because the algorithms usually work by looking at each piece of data in turn and, in doing so, they compare it with all the other pieces of data, perhaps several times and in several ways. Another challenge is to find reliable ways of validating the functions generated: you want to be able to trust that what you have is real, useful knowledge. Another challenge is: is it ethical? Report mining is similar to data mining, but geared to looking at text found in reports. Web mining is similar to data mining, but its data is web pages. The beauty of this is that the HTML tags often give useful information that can guide the searches. For example header and title tags usually indicate headings that have been carefully worded to express something important, so the web mining algorithm might pay more attention to it. The pre-processing and classifying might involve finding all the pieces of text bracketed with HTML tags. Screen scraping is similar to data mining in which the data comes from screens. This is often useful for connecting with legacy systems, especially those from the 1970s, the user interfaces of which are character graphics - the characters are captured and analysed by data mining techniques. The pre-processing can involve converting a linear stream of characters into lines and columns, and screenfuls (often such UIs were 24 lines of 80 characters). But for more modern, graphics oriented screens, the data is not characters but bitmaps: pixels. This requires much more processing, with time- consuming algorithms that do visual pattern recognition and optical character recognition. See Kantardzic (2003) and Keim & Kriegel (1996) for more. 2.3 Mind Mapping Mind-mapping (or cognitive mapping) is a way of finding out what a person knows, when their knowledge is complex. A major player in the field is Colin Eden. See references under 'Eden' at end. Mind mapping assists the analytic, formative and lingual aspects of a person's knowledge especially. » Analytic aspect: It involves helping the person to express distinct ideas and concepts. These are then set down on paper (or screen) as distinct entities. (In older pre-graphical versions of Eden's software, ideas were given unique identification numbers.) » Formative aspect: Ask the person for relationships between the various concepts. Draw these as lines linking the concepts. » Lingual aspect: Both the concepts and the relationships are given labels that express what they mean. The eventual map is a box-and-arrows diagram. With pen and paper, one can produce informal and yet very informative maps. With software, it is a bit more restricted, but the advantage of software is that it can be altered more easily. This is important because during the process of mapping the person will often change their mind about things. This kind of mind mapping helps the person explore what they know in relation to what has already been said. Often a lot of useful detail can be found. But it does no help them think of areas they usually overlook or take for granted - especially a lot of tacit knowledge. 2.4 Multi-aspectual Knowledge Elicitation (MAKE) To overcome this problem, Mike Winfield produced MAKE, and method for Multi-aspectual Knowledge Elicitation. See Winfield references listed at end. Whereas Eden's methods focus on the structure of knowledge of a person, MAKE focuses on its diverse meaning. MAKE generates a box-and-arrows diagram, but it seeks to group the concepts around aspects that make them meaningful. By doing this, and referring to Dooyeweerd's suite of aspects, it is then possible to see which aspects are being overlooked and to ask the person about them (or, better, to ask them if there are any other aspects they might wish to talk about, referring them to the list of aspects). This can yield an aspect map such as follows: Figure 1. Aspect map generated in Multi-Aspectual Knowledge Elicitation The steps of the MAKE process may be seen as: 1. Introduction (e.g. explanation of kernel meanings of aspects, and obtain statement of requirements) 2. Identify a few (e.g. a couple) important aspects. 3. Focus on one of these aspects and specify any laws, axioms, data, definitions, and constraints that apply to the domain. 4. Identify as many concepts as possible that lie in this aspect. (Note: May need to check the concepts at a later stage to ensure they fall within the correct aspect.) 5. Apply Low Level Abstraction to each concept, which needs, or is thought to need exploding. 6. Repeat steps 3-6 as necessary. 7. Use the aspectual template to identify any new aspects, which may apply to the concepts specified (build bridges between concepts and aspects), and return to step 3. Low Level Abstraction was a concept that Winfield developed from the 1991 edition of Clouser [2005] and refers to becoming aware of the various aspectual properties of things yet without isolating them from the things themselves. MAKE is very good at eliciting tacit knowledge. 2.5 Assistum Assistum came out of work at ICI with expert systems. These expert systems contained some human knowledge about an ill-structured area of expertise such as business strategy. Whereas mind mapping and MAKE just let the person express their knowledge, the work at ICI added some general knowledge of its own to guide them in thinking about the issues. It would force them to evaluate what they knew against the computer, and then explore why their knowledge differed from that of the computer. It encouraged them to take their own knowledge (not that of the computer) but to ask themselves reasons why it gave the results it did. In this way, it helped the user think about things they had not thought about before, and sometimes to come up with new knowledge or even new ways of seeing things. Assistum has been developed to provide graphical software help for this. It asks you to think about factors that contribute to each other, then runs them as an inference net so as to show you the outcome or implications of your knowledge. This lets you see its problems, and urges you to think of reasons why, and to alter the knowledge you have expressed. For example, if wanting to recruit a person for leadership potential, you need to think of factors important in good leadership. You might think of: » ability to create vision » ability to motivate people » ability to get things done OK, but what gives the ability to motivate people? You might think of: » passion What else? Or when might passion in the leader not be enough to motivate? For example: » when the thing they are passionate about is not relevant » or when they are not trusted. (Notice the process of knowledge clarification going on here.) Therefore, there are three things needed in order to motivate people: » passion » relevance » trust. But what makes people trust a leader? For example: » integrity » consistency. And so on. You can build up a 'tree' of what contributes to what. Assistum makes it easy to build up such tree structures on screen. Tools like Assistum also let you rearrange knowledge trees to let you look at the issue from a different angle. 3. TECHNOLOGICAL SYSTEMS TO HELP KNOWLEDGE STORAGE AND ORGANISATION 3.1 Issues that Challenge Knowledge Organisation » data warehousing » aspects of storage » representation » longevity » meta-data 3.4 Data Warehousing Suppose you wanted answers to the following questions about a certain customer (Thanks to wikipedia 24 April 2010 for this list, and for some of the other information below): » How much did we sell to customer? » How much time did sales people spend on the customer? » Was the customer happy with the service or goods we provided? » Did the customer pay their bills? » How have these customers fared over last few years? The information to answer these questions is all over the place, in different databases held by different departments - sales, customer relations, orders, etc. It is usually difficult to bring them together. Also, different databases might give conflicting information; for example, sales and finance departments might have different views of how much a certain customer contributes to the organisation, based on different information: which is correct? KM storage is concerned with how to integrate all the knowledge and information that is held. One way to do this is to use a data warehouse. A data warehouse is a repository of the organisation's information, held in an integrated, consistent and consolidated form. The importance of data warehouses is that the organisation needs data that is consolidated and integrated, and which can be relied on. When considering the topic of data warehousing, we must not only think about storage and organisation of information (this section), but also extraction and transormation, aggregation, for sake of analysis (sections on knowledge sharing and application). Data warehouses are good for analysis of performance against goals, analysis of trends (e.g. which products fared best over last few years), and exception reports. Data warehouses, therefore, often receive data from all the other databases in the organisation (and elsewhere too), and store it in a way that is consistent (see aspects of challenges to integration below). There are two main approaches to designing a data warehouse: » bottom up (Gary Kimball) - different 'data marts' are formed from data that is actually needed by the business processes, and after a time these marts are merged. Problem: shambles. » top down (Bill Inmon) - the entire data warehouse is designed, according to a top-level enterprise data model of the organisation. So the warehouse fits the stated aims of the organisation and its ways of working. The data warehouse is a 'corporate information factory' that delivers business intelligence. Top-down design has proven more robust when the business changes. But it can be inflexible when departments in the organisation change. Often hybrid versions occur. Data warehouses are optimised for speed and for analysing data. For this reason, data warehouses often hold data in denormalised form (the opposite of a relational DB that has many tables). Advantages: » Providews common data model, which makes it easlier to analyse. » Inconsistencies between information held elsewhere are detected and resolved. » It can act as an archive of the other systems in the organisations, espeically when one might be closed down and its data lost. » dAnalysis of information does not slow down the operations systems. Disadvantages: » Not so good with unstructured information: see 'free text' information below. » The information inthe data warehouse might be out of data, because of the need to copy data from other sources. » High costs, especially up front. » Duplicate data. 3.2 Aspects of Information Storage and Integration » psychic aspect: bitwise codes, e.g. ASCII, JPEG, binary, .zip. » Integration challenge: how to bring these together? » analytic aspect: knowing what data and kinds of data is in the file, what values they might have, and how to separate them separate them out. » Integration challenge: different data held; different way of separating them e.g. CSV (comma separated variables); e.g. in one database a field can take the values 'M' and 'F', in the same field takes the values "Male" and "Female". » formative aspect: syntax, structure and processing, e.g. KIF, HTML, PDF, XML; web 2.0 » Integration challenge: Data structured in different ways in different places. » lingual aspect: semantics, e.g. concept matching, thesaurus; semantic web » Integration challenge: What seems to be the same variable in two places actually means something different in each. Also there is often just different information or knowledge held in each place: how to bring it together in a way that makes sense? » social aspect: cultural connotations; 'pragmatic web' » Integration challenge: The information in each department's database has different connotations. » juridical aspect: appropriate information , and who has the right to access or change information » Integration challenge: Sales and finance departments might have different views of how much a certain customer contributes to the organisation, based on different information: which is correct? 3.3 Knowledge Representation # Key question: How do we represent our knowledge in symbolic form? Two main ways: » Formal languages and information structures that express a way of seeing the world ('ontologies') » Informal: text, graphics, video, sound, etc. (e.g. the web) Formal languages and structures can be more reliably searched, to find most that is relevant and exclude most that is irrelevant. But this way is usually cumbersome and even insufficient for ill- structured knowledge, especially that which includes a lot of taken- for-granted assumptions or which requires interpretation. Informal means, such as the Web, can more easily enable ill- structured knowledge to be stored. Taken-for-granted assumptions are what we expect in free-format text and pictures, and humankind is well-used to this. But free-format storage cannot be so easily searched, and you either miss a lot that is relevant, or get a lot that is irrelevant (c.f. google searches). 3.3.1 Formal Information Structures and Languages For example: » Tables » Relational databases and SQL » Object orientation » Semantic nets Example: Sports club membership. Only club members may use the facilities, not least because insurance only applies to them. The data of club membership is held in simple database tables. The following fields are held: FIELDNAME: WHAT IT REFERS TO NAME: Member's name ADDR: Address DOB: Data or birth SEX: Sex DAPP: Data of application for membership RECM: Who recommended as member APPS: Stage in process of applying to become a member START: Membership start date YEARS: Number of years a member CURR: Whether currently a member (Y or N) This is a simple structure (only one table). The idea of a formal language. This set of fields provides a very simple 'language' in which to express the fact that someone is, or is applying to be, a member. For example, the following 'sentence' tells us that a certain person is a member: (This is similar to XML format.) Now consider the following challenges. All are concerned with applying the data, but the issues raised affect storage decisions. Challenge 1: Operational/tactical search of DB: You want to send out a mailshot to all members over age of 60; can you easily do this? And what might be the problems in doing this? {*** Exercise and Discussion This can be done with a simple SQL search. What SQL query would you make? Problems: a) Look at the address in the example above: is that suitable for a mailshot? b) What about possible errors in DOB? c) Try to think of other problems. Discuss: To what extent does this include most who are relevant, and exclude most who are irrelevant? This is hardly KM; it is little more than data handling. ***} Challenge 2: Strategic search of DB: You want to know whether the club has been more, or less, attractive to those over 60, increasing or decreasing over the past 10 years. How would you do this? {*** Exercise: More difficult but still possible with SQL search. Think about how you would judge whether over 60 in each of last ten years. This is more like real KM, but is rather crude as such. ***} Challenge 3: Deduced knowledge: You want to find out how attractive the club has been to people from non-British cultures. How would you do this? And what are possible problems? {*** Exercise and Discussion: Even more difficult, but still possible. Hint: think about surnames. An investigation of surnames would probably have to be done manually, with people considering each, and asking whether a British name or not. Problems. One obvious problem is: does surname really indicate whether British or not. Another is: what do we mean be 'British' - those who are British citizens, those who have lived in UK for a time, or those who 'think' who British way, or what? This is more like real KM. It illustrates how KM requires 'judgment'. This is also an example of how you might want to use the data in ways you did not anticipate at first. ***} Challenge 4: A nearby hotel wants to negotiate short-term membership for its guests. Is this database sufficient to hold information? Or how would the structure need to be changed in order to allow this? {*** Exercise and Discussion: Think about why temporary membership might require different treatment. For example, if you just added hotel-members, your list of members would get clogged up with hotel-members, and for example mailshots would go to lots of irrelevant people. So you need a way of distinguishing 'real' from 'hotel' members. The current set of fields does not allow this. What fields might you need to add to the database? Discuss: Suppose it is impossible to change the database structure. How would you cope with the request from the hotel? e.g. could you keep paper records of hotel members? Note: The latter is part of knowledge application. In this way, you can see that the four activities of KM cannot be divorced from each other. ***} Challenge 5: You are expanding. You want to put in a range of facilities that might be of interest only to certain members. You want, therefore, to provide a way of holding information about this. {*** Exercise and Discussion: Discuss: For what reasons might you need to hold such information? The reasons indicate what kinds of information you need to hold. For example: a simple 'TypeOfInterest' flag, or a free-form description, or a flag for each facility to say whether they are allowed and/or likely to use it, ... For the latter, it is better not to use individual flags in the main record [Think: Why not?]. Better to set up two new tables: » Facilities table: recording what facilities are available (and incidentally this allows to store much more information about the facilities, such as insurance conditions, maintenance costs, etc. » Allowance table: Holds primary keys of a member and a facility. An entry in this table shows that that member can use that facility. [Think: Or would it be better to have entries for which members are NOT allowed to use which facilities?] Problem: What about long-term existing members? Discuss: How to fill up the Allowance table to allow which members to use which facilities? The message here: KM involves expanding the knowledge available from time to time. Doing so raises storage problems. ***} Challenge 6: Now you want to hold information about the various facilities, such as invoices for purchase and maintenance, and maintenance logs. {*** Exercise and Discussion: DIscuss: How would you hold these in a way that their data fields can be searched? (i.e. merely holding scanned pictures of paper invoices etc. is not an option.) Problem: different invoices and logs have very different formats. ***} As we can see, formal structures make some searching easy, some searching possible, but also makes it difficult to use the information to inform us in ways not anticipated before, and to expand the information or knowledge held. Both these are needed. {*** Discuss: The structure was a simple table, of relational form. Would the searches and changes above have been easier if the database had been in object-oriented form? ***} During 1980s, 1990s, a lot of research went into finding better structures for information, and better languages in which to express information. XML became increasingly popular as a language. 3.3.2 Ontologies # The need to categorise knowledge [Lecture 14] » taxonomies to separate things out e.g. Dewey Decimal » Taxonomies can help web searching » multi-level: need way of separating things out at whichever level we are » these tend to be hierarchical » challenge: how something can be in several categories; e.g. library coding with ':' 3.3.3 Difference between Language and Structure The above example shows a simple structure but the sentence to set up a member was in XML, which can deal with complex structures. » The structure determines how easy it is to make sophisticated searches and to modify the types of knowledge held. » The language determines how easy it is to express the knowledge entered, and to express the kinds of searching needed. Whatever the language, the structure of the above presupposed distinct variables or attributes, each of which can be identified by name and given a value. The above methods focus on the analytic aspect of information: the ability to distinguish clearly between the different objects and attributes. But some knowledge is not amenable to such treatment, so free-format methods have become popular. 3.3.4 Free-Format Storage From the 1990s, the World Wide Web has provided an alternative way of holding a wide variety of information. Its success is that it focused not on the analytic aspect (distinguishable attributes) but on the lingual aspect of information or knowledge, i.e. what it means. To enter this knowledge, people would write it rather than design it. HTML is an excellent language in which to store such information, because: » It allows free-format text. » But it enables the author to indicate what they think are special parts of the text by means of 'markup' tags, such as headings, bullets, emphasis, citation, etc. » It provides a table mechanism for information that is more structured. » Hyperlinks enable lots of relationships between 'entities' that were not thought of beforehand. » It also allows inclusion of graphics and other media alongside text. » (If your information is linked to Internet, then that link gives extra benefits - but you might hold on an internal-only WWW, an Intranet.) » And many more. In this way, it allows free text but enables us to bring some structure to it. Searching becomes more of a challenge, because you are not sure what information is where, and in what words it is expressed. Types of search: » Lexical: Search for combinations of words (maybe allowing a bit for mis-spelling). Problem: can miss lots of relevant stuff e.g. because wrong words used. Also can get a lot of irrelevant stuff. e.g. searching on 'nirvana' will get a lot about Buddhism and a lot about a rock band of that name; Usually you want one or the other, not both. » Syntactic, Structural (e.g. google search): Take account of the structure of the information, such as the number of hyperlinks from other pages to this one. This can get more irrelevant stuff, but it can present them in a more intelligent order, i.e. things hopefully more relevant first. » Semantic: Search for things of similar meaning. This is very useful when you are not sure of the words that authors might use for a particular concept. For example, I might search for "climate change deniers", but want also "climate change sceptics", "global warming sceptics", etc. For example, to get stuff on the Nirvana rock band, you can say "Search on 'nirvana' but exclude anything to do with Buddhism." » Pragmatic (not yet in full operation): Pragmatics refers to the fact that what a person means when they say or write something usually contains 'illocutionary' meaning in addition to the semantic meaning. » For example climate change deniers do not like the term 'deniers' because it sounds too like "holocaust deniers", which is a term of abuse, so they prefer to call themselves "climate change sceptics". So, whether a text uses the term 'deniers' or 'sceptics' is often an indication of which stance the author takes. » Another example: Idioms like "Were you born in a tunnel?" Such connotations and idioms are part of the 'pragmatics' of language, and can provide useful information. In fact, a wide range of terms have connotations and idioms. {*** Consider each of the challenges above and think about how each might be met using free text with HTML. » Challenge 1: finding names and addresses of all members over 60 [more difficult unless written carefully in a single structure] » Challenge 2: looking at whether the club has become more attractive to over-60s over the last ten years [much more difficult] » Challenge 3: looking at how attractive the club has been to non-British people [about the same difficulty, because the kinds of surname processing done is the same] » Challenge 4: Adding information necessary for hotel- members [easy: just write it] » Challenge 5: Allowing for different facilities and which members can use them [much easier to add the information but maybe laborious if you need to update every existing member] » Challenge 6: Can easily add scanned invoices etc. and get them converted to character text by OCR. 3.5 Longevity Knowledge and data must be expected to be held for a long time. This poses challenges: » Safety: not to lose the knowledge. Usually accomplished by good backup and archiving facilities. » Versions and changes: When you overwrite the value of a field, and when you delete information, you lose information about what it was before. You might want to find out the old value. So it is useful to keep a record of the history of changes of information. » Out of date knowledge: Some knowledge becomes out of date after a time. So it is often useful to expunge it after that time, so it will clutter up searches or mislead its users. » Knowledge gains different meaning over the years. For example the word "gay" used to mean "happy, glad, gleeful"; nowadays in the UK it means "homosexual male". » Knowledge has different connotations over the years, and across different departments. You have to try to work out what connotation the writer had when the knowledge was written. 3.6 Meta-Data # Meta-data: What you need to know (and hold) about information: Helps searchers know what to do with the information, how to interpret it, and hence how to store or retrieve it, and how far up list of results to place it. # Example: Meta tags in HTML, such as keywords. # Example: Much more sophisticated is Dublin Core metadata: » title: label assigned to data element » resource identifier: e.g. url » expiry date and time, after which the information can be removed » version: » author » cache control: maybe be cached in public, private, not-to-be- cached, can be cached but not archived » registration authority: who can register it » source: e.g. "citrus fruit wholesaler" » language: in which the data element is specified » keywords used be searching engines to index it » obligation: whether element should always contain a value, or be present » datatype: e.g. integer, text » robots: whether crawlers may index and/or follow this information Notice how some of these are to do with longevity. Metadata may be seen as information about aspects of the information that is held. For example: Registration authority is juridical (legal) aspect; author is lingual aspect; source is social aspect; version is formative aspect. {*** Exercise. Work out aspects for all the above, and then find more metadata offered by Dublin Core. ***} 4. TECHNOLOGICAL ISSUES TO HELP KNOWLEDGE SHARING 4.1 Issues that Challenge Knowledge Sharing (Example: refer to the sports club database above.) Sharing of electronically-held knowledge raises the following challenges: » Different bit-level protocols (psychic aspect of knowledge sharing). Protocols might vary across platforms or between databases. e.g. MSWord versus Powerpoint versus pdf versus Scala versus jpeg, and so on. Whether zipped or not. » Different vocabularies (lexical issue; analytic aspect of knowledge sharing): » Different variable names for same field, e.g. ADDR above might be Address or HomeAddress or Residence, etc. in other DBs. » For example, a field TypeOfMember in sports club database might have different values in each database. For example, in DB1, 1 = ordinary member, 2 = OAP member, 3 = student member, 4 = child member, 5 = hotel member, while in DB2, "O" = ordinary member, "H" = hotel member, "D" = discount member (OAP or student or low- waged). Challenge: how to translate between them. » How to deal with null or empty fields. e.g. an empty field might mean "ordinary member" in DB3 but might mean "information not supplied" in DB4. » How to indicate "information not yet available". » How to indicate uncertainty in information. For example, in answer to "Which party will you vote for in election", many people will answer "I don't know" or "I'm not going to tell you". These are NOT the same as "Information not yet available". » Different information structure (syntactic issue; formative aspect of knowledge sharing). For example: » One database has information that another might not have. » One database holds data or birth as a single field, while another holds it as three fields. » One database holds information on which members can use which facilities as 'tick fields' while another holds it as a separate 'Allowances' table. » Different meaning to the information (semantic issue; lingual aspect of knowledge sharing). For example: » DBs might indicate 'child member' by 4, but in one this means up to 12 years of age, in another up to 14, in another up to 16 years of age. See also lingual aspect of longevity above. » Different connotations etc. (pragmatics issue; social aspect of knowledge sharing). Different groups of people have different connotations. For example: # Different languages. Especially for web pages. This covers most of the levels or aspects above. 5. TECHNOLOGICAL SYSTEMS TO HELP KNOWLEDGE APPLICATION When applying knowledge that is in your KM system, think of it, not as 'stuff', but as a communication from those who wrote and stored it. 5.1 Issues that Challenge Knowledge Application » Access » Understanding » Validity of Stored Knowledge » Visualisation 5.2 Access to Knowledge This is covered by the above. See 'Aspects of Information Storage'. 5.3 Understanding The knowledge must be understood before it can be used properly. Otherwise it might mislead its user. Or it might be misused. 5.4 Validity of the Stored Knowledge For example, is it out of date? Is it true? Is it reliable? Jürgen Habermas suggests that the validity of any rational communication can be of several types. If you see the stored information as something communicated to you, then you can consider to what extent it is valid. His validity types are: » Intelligble: Does it make sense? » Truthful: Does it accord with reality? (including out of date) » Sincere: Was the writer sincere? (or were they joking? Or even deceitful?) » Appropriate: Is the knowledge appropriate to the use to which it will be put? Especially, is it ethically appropriate or socially appropriate? 5.5 Visualisation of Information It is very helpful to put knowledge into graphical form. This helps understanding. Especially of complex knowledge. Some examples will be given. # Napoleon's March # The challenge of knowing what's in hundreds of websites, or navigating them: » e.g. multi layer tiles, to navigate down » e.g. 3D scene: each skyscraper represents a different website; fly through » e.g. Themescape: peaks show lots of documents about a given topic (topics chosen by abundance of words) » Hyperbolic trees (a graph showing connections) REFERENCES Clouser R, (1991, 2005 2nd ed.), The Myth of Religious Neutrality; An Essay on the Hidden Role of Religious Belief in Theories, University of Notre Dame Press, Notre Dame, Indiana, USA. Eden C, (1988), "Cognitive Mapping", European Journal of Operational Research, 36:1-13. Eden C, (1989), "Using cognitive mapping for strategic options development and analysis (SODA)", in Rosenhead J (ed.), Rational analysis for a problematic world: problem structuring methods for complexity, uncertainty and conflict, John Wiley, Chichester, UK. Eden C, (1992), "On the nature of cognitive maps", Journal of Management Studies, v.29, n.3, pp.261-265. Eden C, Ackermann F, Cropper S, (1992), "The analysis of cause maps", Journal of Management Studies, v.29, n.3, pp.309-324. :: nyr Habermas, j. (1984). The Theory of Communicative Action; Volume One: Reason and the Rationalization of Society, tr. McCarthy T, Polity Press, Cambridge, UK. (1984). Inmon, W.H. (1995) Tech Topic: What is a Data Warehouse? Prism Solutions. Volume 1. 1995. Kantardzic, Mehmed (2003). Data Mining: Concepts, Models, Methods, and Algorithms. John Wiley & Sons. Keim D. A, Kriegel H.-P. (1996), Visualization Techniques for Mining Large Databases: A Comparison , IEEE Transactions on Knowledge and Data Engineering, Vol. 8, No. 6, Dec. 1996, pp. 923-938. Kimball, R., Ross, M. The Data Warehouse Toolkit Second Edition (2002) John Wiley and Sons, Inc. Winfield M (2000) Multi-Aspectual Knowledge Elicitation PhD Thesis, University of Salford, U.K. Winfield M J, Basden A, Cresswell I, (1996), "Knowledge elicitation using a multi-modal approach", World Futures, V.47, pp.93 - 101. Winfield M J, Basden A, Cresswell I, (1995), "A Revised Multi- Modal Approach to Information Systems Design", Proceedings of the Thirty Ninth Annual Meeting of the International Society for Systems Thinking, Free University, Amsterdam, July 24-28, 1995. Winfield, M.J. & Basden, A. (2006). Elicitation of highly interdisciplinary knowledge. pp. 63-78 in S. Strijbos, A. Basden (eds.) In Search of an Integrated Vision for Technology; Interdisciplinary Studies in Information Systems. Springer. Andrew Basden. All rights reserved. 17 April 2010, 24 April 2010.