12.07.2015 Views

Knowledge Management in an E-commerce System

Knowledge Management in an E-commerce System

Knowledge Management in an E-commerce System

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

<strong>Knowledge</strong> <strong>M<strong>an</strong>agement</strong> <strong>in</strong> <strong>an</strong> E-<strong>commerce</strong> <strong>System</strong>M<strong>in</strong>or Gordon ‡ , J<strong>an</strong><strong>in</strong>a Jakubczyc*, Violetta Gal<strong>an</strong>t* <strong>an</strong>d Marc<strong>in</strong> Paprzycki ‡* Wrocław University of EconomicsDepartment of Intelligent <strong>System</strong>sWrocław, Pol<strong>an</strong>d{jakubczy,gal<strong>an</strong>t}@m<strong>an</strong>ager.ae.wroclaw.pl‡ Oklahoma State UniversityComputer Science DepartmentTulsa, Oklahoma, US{m<strong>in</strong>org,marc<strong>in</strong>}@cs.okstate.eduAbstractThere are a number of ways to conceptualize the activities of <strong>an</strong> e-<strong>commerce</strong> system. In this paper we consider theknowledge m<strong>an</strong>agement aspects of e-<strong>commerce</strong>, <strong>an</strong>d demonstrate that knowledge m<strong>an</strong>agement is <strong>an</strong> essential part of thesystem’s capabilities. Thereafter we present a detailed discussion of how the three ma<strong>in</strong> knowledge m<strong>an</strong>agement functions(creation, tr<strong>an</strong>sfer <strong>an</strong>d application) take place <strong>in</strong> various parts of the system. F<strong>in</strong>ally, we defend the claim that knowledgem<strong>an</strong>agement is the base of adaptivity <strong>in</strong> <strong>an</strong> e-<strong>commerce</strong> environment.Keywords<strong>Knowledge</strong> <strong>M<strong>an</strong>agement</strong>, <strong>Knowledge</strong>, Meta-knowledge, E-<strong>commerce</strong> <strong>System</strong>, Ontologies, Adaptivity1. INTRODUCTIONBecause of its immense <strong>an</strong>d grow<strong>in</strong>g <strong>in</strong>fluence on all parts of the market <strong>an</strong>d the commercial org<strong>an</strong>izations with<strong>in</strong> it,knowledge m<strong>an</strong>agement has become of the primary focuses of m<strong>an</strong>agement sciences [22]. As contrasted to otherm<strong>an</strong>aged resources such as assets, capital <strong>an</strong>d people, knowledge is characterized by perpetual regeneration: the moreoften knowledge is used, the more knowledge is produced. What is this knowledge? Researchers <strong>an</strong>d practitioners<strong>in</strong>volved <strong>in</strong> knowledge m<strong>an</strong>agement usually conceptualize it as the highest level of reflection over the mass of availabledata. It is the end product of the tr<strong>an</strong>sformation from data to <strong>in</strong>formation to knowledge. This process is the <strong>in</strong>itial step ofknowledge m<strong>an</strong>agement, which is usually comprised of three ma<strong>in</strong> functions [2, 6, 11, 16, 29, 30, 31]:1. knowledge creation, which encompasses the further sub-processes of acquisition, storage, validation, process<strong>in</strong>g ofnew knowledge <strong>an</strong>d <strong>in</strong>tegration with exist<strong>in</strong>g knowledge2. knowledge tr<strong>an</strong>sfer, which addresses the questions of how, what, when, to whom, <strong>in</strong> what form <strong>an</strong>d to whatpurpose knowledge is delivered; <strong>an</strong>d3. knowledge application, which def<strong>in</strong>es the mech<strong>an</strong>isms by which knowledge is employed.These knowledge m<strong>an</strong>agement functions create a cont<strong>in</strong>uous knowledge feedback loop: when exist<strong>in</strong>g knowledge isapplied, a set of new possibilities for further knowledge creation materializes. When this knowledge is acquired, it must<strong>in</strong> turn be tr<strong>an</strong>sferred <strong>in</strong> order for it to be applied aga<strong>in</strong>. This spiral process progresses through a series of <strong>in</strong>cremental <strong>an</strong>dcont<strong>in</strong>uous adv<strong>an</strong>cesThe functions of knowledge m<strong>an</strong>agement are carried out <strong>in</strong> various forms <strong>in</strong> both hum<strong>an</strong>- <strong>an</strong>d Internet-basedcommercial org<strong>an</strong>izations. Much of the execution of these functions is <strong>in</strong>gra<strong>in</strong>ed <strong>in</strong> the basic hum<strong>an</strong> mode of thought, <strong>an</strong>dneed not be consciously considered when mak<strong>in</strong>g decisions – each day we create, tr<strong>an</strong>sfer <strong>an</strong>d apply knowledge <strong>in</strong><strong>commerce</strong> without much awareness of it. In this paper we are <strong>in</strong>terested <strong>in</strong> knowledge m<strong>an</strong>agement as it materializes <strong>in</strong> e-<strong>commerce</strong>. In the electronic environment knowledge m<strong>an</strong>agement is a synthesis of both economic <strong>an</strong>d technicalknowledge processes. Most theories of economic knowledge take for gr<strong>an</strong>ted the medium of knowledge exch<strong>an</strong>ge –International Conference on Electronic Commerce Research (ICECR-5)


<strong>Knowledge</strong> <strong>M<strong>an</strong>agement</strong> <strong>in</strong> <strong>an</strong> E-<strong>commerce</strong> <strong>System</strong> 2hum<strong>an</strong> <strong>in</strong>teraction. In e-<strong>commerce</strong> the primary medium of knowledge acquisition, exch<strong>an</strong>ge <strong>an</strong>d application is thecomputer; this fundamental difference raises m<strong>an</strong>y issues <strong>in</strong> the tr<strong>an</strong>slation of economic theories of knowledge to thedoma<strong>in</strong> of e-<strong>commerce</strong>. Here we wish to survey some of these issues, as well as speculate on some possible approaches toconfront<strong>in</strong>g them. S<strong>in</strong>ce e-<strong>commerce</strong> is a relatively new field, questions of how (or if it is possible) to tr<strong>an</strong>slate thetheoretical <strong>an</strong>d practical experience of hum<strong>an</strong>-driven economies ga<strong>in</strong>ed from years past <strong>in</strong>to a form that is applicable <strong>in</strong>the e-<strong>commerce</strong> doma<strong>in</strong> are still open. Before attempt<strong>in</strong>g to address a few of these questions, we will first look at some ofthe differences between e-<strong>commerce</strong> <strong>an</strong>d traditional <strong>commerce</strong> <strong>in</strong> the context of knowledge m<strong>an</strong>agement functions, <strong>in</strong>Section 2. The process of knowledge creation <strong>in</strong> the context of e-<strong>commerce</strong> will be summarized <strong>in</strong> Section 3. Typicalknowledge acquisition techniques result <strong>in</strong> what we will name “low level knowledge”, a form of knowledge that must becoerced <strong>in</strong>to useful frameworks for particular knowledge consumers <strong>in</strong>side <strong>an</strong>d outside the system. This process will bediscussed <strong>in</strong> Sections 4 <strong>an</strong>d 5. F<strong>in</strong>ally, even though the processes of knowledge acquisition <strong>an</strong>d knowledge tr<strong>an</strong>sfer mayseem difficult, that of knowledge application is considerably more so. We will address this application <strong>in</strong> Section 6.2. COMMERCE VS. E-COMMERCE AND THE KNOWLEDGE PROCESSESThe tr<strong>an</strong>slation of knowledge m<strong>an</strong>agement functions from the doma<strong>in</strong> of <strong>commerce</strong> to that of e-<strong>commerce</strong> is largely aprocess of explicat<strong>in</strong>g what is now implicit – <strong>an</strong> end-goal which has much <strong>in</strong> common to that of the classical approach toartificial <strong>in</strong>telligence, which seeks to comprehensively formalize hum<strong>an</strong> knowledge <strong>an</strong>d knowledge of processes <strong>in</strong> orderto replicate (or at least appear to replicate) hum<strong>an</strong> <strong>in</strong>telligence. The <strong>an</strong>alogy to artificial <strong>in</strong>telligence extends further: oneof the fundamental goals of e-<strong>commerce</strong> is to reproduce the (presumably positive) experiences of brick-<strong>an</strong>d-mortar<strong>commerce</strong> – the helpful sales assist<strong>an</strong>t, <strong>an</strong> easy-to-browse selection, etc. Obviously, not all aspects of <strong>commerce</strong> c<strong>an</strong> bereplicated <strong>in</strong> e-<strong>commerce</strong>. For the time be<strong>in</strong>g it is impossible to touch the sweater or to smell the perfumes, <strong>an</strong>d soe-<strong>commerce</strong> must apply one of its most import<strong>an</strong>t strategies: what c<strong>an</strong>not be replicated must be replaced. Creat<strong>in</strong>gacceptable substitutes for these miss<strong>in</strong>g experiences is the key challenge for e-<strong>commerce</strong>, <strong>an</strong>d the purpose of most of its“<strong>in</strong>telligent” processes. <strong>Knowledge</strong> m<strong>an</strong>agement is the cornerstone of these processes, <strong>an</strong>d the focal po<strong>in</strong>t of our contrast.No doubt, the proper start<strong>in</strong>g po<strong>in</strong>t for a broad comparison between electronic <strong>an</strong>d brick-<strong>an</strong>d-mortar <strong>commerce</strong>would be the most obvious one – the medium itself. However, <strong>in</strong> this paper we are solely <strong>in</strong>terested <strong>in</strong> how the contrastbetween e-<strong>commerce</strong> <strong>an</strong>d traditional bus<strong>in</strong>ess affects the knowledge m<strong>an</strong>agement aspects of the e-<strong>commerce</strong> system, <strong>an</strong>dhow these differences <strong>in</strong>troduce a widen<strong>in</strong>g gap <strong>in</strong> the way knowledge is m<strong>an</strong>aged <strong>in</strong> the hum<strong>an</strong>-based enterprise versusthe electronic one. While a very large number of signific<strong>an</strong>t differences between <strong>commerce</strong> <strong>an</strong>d e-<strong>commerce</strong> c<strong>an</strong> bepo<strong>in</strong>ted to, we concentrate our attention on those that play <strong>an</strong> import<strong>an</strong>t role <strong>in</strong> the context of knowledge m<strong>an</strong>agement.We divide them <strong>in</strong>to three groups: differences of scope, differences for the customer <strong>an</strong>d differences for the system.2.1 Differences of scopeAt the core of a contrast between electronic <strong>an</strong>d traditional <strong>commerce</strong> lie some essential differences <strong>in</strong> the scope ofoperation, which have signific<strong>an</strong>t impact for knowledge m<strong>an</strong>agement: the scope of commodities, the scope of time <strong>an</strong>dthe geographical scope of the market:1. While conventional stores only carry a limited supply of items (typically the most popular), onl<strong>in</strong>e stores c<strong>an</strong> offer<strong>an</strong> almost unlimited number of goods, <strong>in</strong>clud<strong>in</strong>g specialty products targeted for a relatively small niche market.Confederations of electronic merch<strong>an</strong>ts (e.g. amazon.com, which is billed as hav<strong>in</strong>g “Earth’s Biggest Selection”)act<strong>in</strong>g as <strong>in</strong>termediaries between multiple sellers <strong>an</strong>d customers, may generate a very large commodity search space[14, 33]. Moreover, assum<strong>in</strong>g the growth of the onl<strong>in</strong>e market cont<strong>in</strong>ues, one may envision <strong>an</strong> <strong>in</strong>termediary systemthat uses the complete Internet as the source of its commodities, thus mak<strong>in</strong>g its commodity space almost as large asthe space of all commodities possibly available on Earth. It is easy to see that this has <strong>an</strong> immediate effect on theknowledge-related requirements <strong>in</strong> the system. For <strong>in</strong>st<strong>an</strong>ce, while it is possible that <strong>in</strong> a small store few salespeoplec<strong>an</strong> know most of what it is to know about the offered commodities, a large e-<strong>commerce</strong> enterprise must store <strong>an</strong>dm<strong>an</strong>age <strong>an</strong> enormous amount of data about the available commodities – even if the accumulated data about eachitem is m<strong>in</strong>imal, the shear number of items makes the total amount of data very large.International Conference on Electronic Commerce Research (ICECR-5)


<strong>Knowledge</strong> <strong>M<strong>an</strong>agement</strong> <strong>in</strong> <strong>an</strong> E-<strong>commerce</strong> <strong>System</strong> 32. Keep<strong>in</strong>g a physical store open 24 hours a day is expensive proportional to the amount of revenue it generates. An e-<strong>commerce</strong> system, on the other h<strong>an</strong>d, c<strong>an</strong> run non-stop with very little extra overhead. However, however, thisrequires the process of knowledge m<strong>an</strong>agement to be <strong>an</strong> un<strong>in</strong>terrupted one. In st<strong>an</strong>dard <strong>commerce</strong> it is possible toexecute <strong>an</strong>d <strong>in</strong>corporate the results of knowledge m<strong>an</strong>agement functions while the “system” is offl<strong>in</strong>e / the stores areclosed, but this is not the case with <strong>an</strong> e-<strong>commerce</strong> system. It is even possible to envision a situation <strong>in</strong> which thereis no slower time at night, because the store is onl<strong>in</strong>e <strong>an</strong>d customers are active all over the world.3. Physical <strong>commerce</strong> is usually concerned with sell<strong>in</strong>g goods locally; even the largest multi-national are acclimated tofit the regions <strong>in</strong> which they operate. A successful electronic enterprise will, sooner or later, reach a truly globalscale, <strong>an</strong>d be forced to accommodate local differences <strong>in</strong> order to succeed <strong>in</strong> diverse markets. Because of this theamount of knowledge that the system must m<strong>an</strong>age <strong>in</strong>creases considerably, as knowledge about all local, culturalpeculiarities must also be ma<strong>in</strong>ta<strong>in</strong>ed. In addition, the overall picture of knowledge m<strong>an</strong>agement functions becomesa bit fuzzier <strong>in</strong> certa<strong>in</strong> situations: for example, a Europe<strong>an</strong> bus<strong>in</strong>essm<strong>an</strong> work<strong>in</strong>g <strong>in</strong> US may log <strong>in</strong> late at night to <strong>an</strong>operational center <strong>in</strong> Jap<strong>an</strong> (where it is already the next day), to order goods to be delivered to Europe (where it ismorn<strong>in</strong>g) to his family. In such a situation, knowledge acquisition functions become more complicated due to thefact that the customer, the “store” <strong>an</strong>d the recipient(s) are different <strong>an</strong>d reside <strong>in</strong> different countries <strong>an</strong>d time zones(this may, for <strong>in</strong>st<strong>an</strong>ce, make the conceptualization of knowledge acquisition more difficult).2.2 Differences for the userAs is obvious to <strong>an</strong>yone who has purchased onl<strong>in</strong>e, the user’s experience of e-<strong>commerce</strong> is paradigmatically differentth<strong>an</strong> that of traditional <strong>commerce</strong>. Some of these differences are very closely related to those discussed above, whileothers are tied to the nature of computer-hum<strong>an</strong> <strong>in</strong>teraction.1. Customers of traditional stores live primarily <strong>in</strong> the vic<strong>in</strong>ity of the store, visit it dur<strong>in</strong>g normal hours of operation,purchase items from a limited selection of the most <strong>in</strong>-dem<strong>an</strong>d commodities <strong>in</strong> a given area (e.g. kosher food <strong>in</strong> NewYork, car battery heaters <strong>in</strong> Fairb<strong>an</strong>ks) <strong>an</strong>d are attached to a particular regularly visited store. As argued above, thisis not the case <strong>in</strong> e-<strong>commerce</strong> “stores”. Here customers c<strong>an</strong> come from <strong>an</strong>y place <strong>in</strong> the world. They are attached toa br<strong>an</strong>d name of “store” (such as Amazon.com), not a physical location, they are served 24/7/365 <strong>an</strong>d they expect tof<strong>in</strong>d <strong>an</strong>yth<strong>in</strong>g that they are look<strong>in</strong>g for. The knowledge m<strong>an</strong>agement related effects of these requirements have beendescribed above.2. One of the more import<strong>an</strong>t differences that affects the customer behavior is the presence or lack of a hum<strong>an</strong>salesperson [3, 28]. To be able to overcome this difference a number of knowledge based personalization techniqueshave been <strong>an</strong>d are be<strong>in</strong>g developed [12, 20, 25, 26, 34]. For further discussion of personalization <strong>an</strong>d relationshipswith the customer, see [10].3. Clients expect that their experience of <strong>an</strong> onl<strong>in</strong>e store will be <strong>in</strong> l<strong>in</strong>e with the best of <strong>an</strong> ″Internet experience,” one <strong>in</strong>which <strong>in</strong>st<strong>an</strong>t gratification is the mode – <strong>in</strong>stead of search<strong>in</strong>g across a monster size Walmart, where almosteveryth<strong>in</strong>g c<strong>an</strong> be found, but it is unclear where it is hid<strong>in</strong>g. We need to remember that <strong>in</strong> <strong>an</strong> Internet store it is easyto provide the potential customer with very detailed <strong>in</strong>formation about product(s) <strong>an</strong>d service(s), as well as referencethe large amount of additional <strong>in</strong>formation available <strong>in</strong> a multitude of other, Internet-based, repositories. This delugeof <strong>in</strong>formation, while crucial to overcom<strong>in</strong>g the deficiencies of the system (e.g. lack of physical contact with thecommodities) c<strong>an</strong> also overwhelm the customer. This <strong>in</strong>dicates clearly that a number of import<strong>an</strong>t techniques basedon knowledge m<strong>an</strong>agement should be applied <strong>in</strong> order to filter the <strong>in</strong>formation <strong>in</strong>to <strong>an</strong> easily-comprehendiblepackage. Most of these techniques are related to the general concept of personalization as <strong>in</strong>formation filter<strong>in</strong>g:selection of the right sales support strategy, right <strong>in</strong>terface, right search support etc.4. There is still the problem of content delivery: customers who do not access the Internet through a fast enoughconnection to be able to receive high quality <strong>in</strong>formation. This limitation is especially import<strong>an</strong>t when we movefrom e-<strong>commerce</strong> to mobile <strong>commerce</strong> (m-<strong>commerce</strong>) <strong>an</strong>d from high quality computer displays to PDA/cell phonedisplays that c<strong>an</strong> deliver only low quality graphics, or no graphics at all. We must be able to actively applyknowledge about the physical <strong>an</strong>d process<strong>in</strong>g capabilities of display devices <strong>in</strong> order to regulate the form as well asthe content of the <strong>in</strong>formation delivered to the contents.5. Clients expect to feel <strong>an</strong>d rema<strong>in</strong> <strong>an</strong>onymous but also to be able to browse <strong>an</strong>d purchase what they are really<strong>in</strong>terested <strong>in</strong>, two <strong>in</strong>cl<strong>in</strong>ations that are often mutually <strong>in</strong>compatible <strong>in</strong> a traditional shopp<strong>in</strong>g scenario. In <strong>an</strong>International Conference on Electronic Commerce Research (ICECR-5)


<strong>Knowledge</strong> <strong>M<strong>an</strong>agement</strong> <strong>in</strong> <strong>an</strong> E-<strong>commerce</strong> <strong>System</strong> 4e-<strong>commerce</strong> system this is more feasible, but <strong>in</strong> order to do it the knowledge m<strong>an</strong>agement processes must be madeas tr<strong>an</strong>sparent as possible, so that we may offer the improved service that accomp<strong>an</strong>ies <strong>in</strong>creased knowledge of thetarget market without scar<strong>in</strong>g the customer by act<strong>in</strong>g like a “big brother” (also see discussion of personalization <strong>in</strong>[10]).6. In customer-oriented bus<strong>in</strong>ess we are <strong>in</strong>terested not only <strong>in</strong> what the customer has purchased, but also why shepurchased it, as this knowledge is <strong>in</strong>credibly useful <strong>in</strong> “personaliz<strong>in</strong>g” the user’s onl<strong>in</strong>e shopp<strong>in</strong>g experience <strong>an</strong>dsuggest<strong>in</strong>g related products. This situation is usually only possible <strong>in</strong> very small regional stores, <strong>in</strong> which the ownerknows most of the customers <strong>an</strong>d c<strong>an</strong> readily learn or deduce the reasons for a particular purchase (if a couple isbuy<strong>in</strong>g a new washer because they are expect<strong>in</strong>g a baby, the store owner c<strong>an</strong> suggest a number of additionalproducts that should be purchased as well). This capability is not apparent <strong>in</strong> large cha<strong>in</strong> stores. While it is possiblethat a skilled salesperson c<strong>an</strong> learn someth<strong>in</strong>g <strong>in</strong> conversations with the customer, these contacts are usually notenough. On the large scale it is only through extensive knowledge that we may attempt to make educated guesses asto why a purchase decision has been made. Also, because the user is identified to the system (e.g. by a log<strong>in</strong> name)all of his actions may be related, versus the disconnected <strong>an</strong>d mostly <strong>an</strong>onymous nature of real-life tr<strong>an</strong>sactions.While this is a very positive development, it poses <strong>an</strong> import<strong>an</strong>t challenge to knowledge m<strong>an</strong>agement, as the correctknowledge about each customer has to be acquired to provide her with the small store feel<strong>in</strong>g (see [21]).2.3 Differences for the systemF<strong>in</strong>ally, we have to address the differences between <strong>commerce</strong> <strong>an</strong>d e-<strong>commerce</strong> that occur <strong>in</strong>side of the system (here thesystem is understood very broadly <strong>an</strong>d me<strong>an</strong>s <strong>an</strong>y regular commercial entity or <strong>an</strong>y e-<strong>commerce</strong> system).1. There exist a number of techniques <strong>an</strong>d tools that were developed to help m<strong>an</strong>age knowledge <strong>in</strong> traditionalenterprises, <strong>in</strong> which knowledge embodied <strong>in</strong> workers is the primary focus [32]. In the hum<strong>an</strong> enterprise we areparticularly <strong>in</strong>terested <strong>in</strong> extract<strong>in</strong>g this knowledge, <strong>an</strong>d the process of exch<strong>an</strong>ge of <strong>in</strong>formation between knowledgebearers. In e-<strong>commerce</strong> we have a much different situation, because most knowledge m<strong>an</strong>agement occurs “<strong>in</strong>side”the IT <strong>in</strong>frastructure <strong>an</strong>d therefore the hum<strong>an</strong> factor is m<strong>in</strong>imal. This makes knowledge m<strong>an</strong>agement difficult, as allof it has to happen <strong>in</strong> <strong>an</strong> almost-automatic fashion. At the same time, due to the lack of hum<strong>an</strong> <strong>in</strong>volvement, one ofthe more import<strong>an</strong>t problems of knowledge m<strong>an</strong>agement systems (hum<strong>an</strong>- <strong>an</strong>d paper-based as well as electronic),the problem of knowledge shar<strong>in</strong>g, does not exist (computers do not m<strong>in</strong>d shar<strong>in</strong>g knowledge with other computers).2. Automatic record<strong>in</strong>g of data related to the function<strong>in</strong>g of the e-<strong>commerce</strong> system is easily available; no traditionalsystem has such stores of data readily available. It should be obvious that this collected data will become the crucial<strong>in</strong>put to the knowledge m<strong>an</strong>agement process, once it is tr<strong>an</strong>sformed <strong>in</strong>to <strong>in</strong>formation <strong>an</strong>d then <strong>in</strong>to knowledge.3. The need for <strong>in</strong>ventory m<strong>an</strong>agement is different. While new approaches are start<strong>in</strong>g to take effect among large cha<strong>in</strong>retailers, typically what is “visible” to the customer is what available <strong>an</strong>d it is “impossible” to oversell. In electronic<strong>commerce</strong>, if a large number of tr<strong>an</strong>sactions occur simult<strong>an</strong>eously, special care must be taken to only sell goods thatare available. On the other h<strong>an</strong>d, s<strong>in</strong>ce <strong>in</strong>formation about the <strong>in</strong>ventory levels is const<strong>an</strong>tly available sales strategiesmay be dynamically adjusted. This leads us back to knowledge m<strong>an</strong>agement. <strong>Knowledge</strong> employed <strong>in</strong> <strong>in</strong>ventorym<strong>an</strong>agement is directly related to knowledge-driven sale strategies, knowledge used <strong>in</strong> commodity searches,negotiations with suppliers of a given commodity etc.4. F<strong>in</strong>ally, e-<strong>commerce</strong> systems are characterized by <strong>an</strong> almost unlimited <strong>an</strong>d immediately accessible data aboutproducts <strong>an</strong>d services. Therefore <strong>an</strong> import<strong>an</strong>t task of knowledge m<strong>an</strong>agement is the ability to cope with large <strong>an</strong>dcont<strong>in</strong>uous flow of <strong>in</strong>formation.3. KNOWLEDGE CREATION: INFORMATION INTO KNOWLEDGE<strong>Knowledge</strong> m<strong>an</strong>agement experts describe knowledge creation as the cont<strong>in</strong>uous process of tr<strong>an</strong>sform<strong>in</strong>g tacit knowledge<strong>in</strong>to explicit knowledge, that which c<strong>an</strong> be expressed with a formal representation [5, 31]. This def<strong>in</strong>ition usually refers toknowledge acquisition from hum<strong>an</strong> experts. On a large scale, the difficulty of gather<strong>in</strong>g knowledge from experts (oftenreferred to as the bottleneck of the knowledge system) has necessitated the use of alternative, less troublesome sources ofInternational Conference on Electronic Commerce Research (ICECR-5)


<strong>Knowledge</strong> <strong>M<strong>an</strong>agement</strong> <strong>in</strong> <strong>an</strong> E-<strong>commerce</strong> <strong>System</strong> 5knowledge. As the amount of data stored <strong>in</strong> local databases <strong>an</strong>d other electronic repositories cont<strong>in</strong>ues to <strong>in</strong>crease,techniques for tr<strong>an</strong>sform<strong>in</strong>g this data <strong>in</strong>to <strong>in</strong>formation <strong>an</strong>d then knowledge become more import<strong>an</strong>t. The use of thesetechniques requires much less hum<strong>an</strong> <strong>in</strong>put th<strong>an</strong> traditional expert knowledge extraction. In the ideal situation, the expertsare given only the task of verify<strong>in</strong>g the generated knowledge <strong>an</strong>d controll<strong>in</strong>g the acquisition process. In our approach wewill follow the general structure of knowledge m<strong>an</strong>agement function, which consists of knowledge creation, tr<strong>an</strong>sfer <strong>an</strong>dstorage, <strong>an</strong>d application. We will therefore start from discuss<strong>in</strong>g the sources of data that is tr<strong>an</strong>sformed <strong>in</strong>to <strong>in</strong>formation<strong>an</strong>d then <strong>in</strong>to knowledge (while this is a mental shortcut, we will name these data sources: sources of knowledge).Overall, <strong>in</strong> the context of <strong>an</strong> e-<strong>commerce</strong> system we consider three primary sources of knowledge: users of the system,environment, <strong>an</strong>d experts. We will now address each of these sources <strong>in</strong> some detail <strong>an</strong>d follow with a brief overview ofknowledge acquisition techniques <strong>an</strong>d mech<strong>an</strong>isms.3.1 Users of the systemBecause the process of gather<strong>in</strong>g data from sources with<strong>in</strong> the e-<strong>commerce</strong> system c<strong>an</strong> be automated <strong>an</strong>d operatecont<strong>in</strong>uously, various types of electronic data are the most relied-upon orig<strong>in</strong>al sources of knowledge for the system.These <strong>in</strong>clude the more tr<strong>an</strong>sparent data sources related to customers such orig<strong>in</strong>at<strong>in</strong>g IP logs, cookies stored acrosssessions <strong>an</strong>d verbose <strong>in</strong>formation logs generated by the system dur<strong>in</strong>g <strong>in</strong>teraction with users (which <strong>in</strong>corporate<strong>in</strong>formation about actions, as well as non-actions of the client; the fact that the user did not select a given pop-up-b<strong>an</strong>neradmay be as import<strong>an</strong>t as the fact that she has selected a given photo camera). In addition, there are several directmethods of extract<strong>in</strong>g <strong>in</strong>formation from the user, such as questionnaires [9, 21]. From this <strong>in</strong>formation about the customerknowledge c<strong>an</strong> also be extracted about <strong>in</strong>dividual <strong>an</strong>d group behavior, about customs, <strong>in</strong>cl<strong>in</strong>ation, etc. We also need totake <strong>in</strong>to account that the system is <strong>an</strong> <strong>in</strong>termediary between suppliers <strong>an</strong>d customers. In this way one c<strong>an</strong> argue thatwhile to relationship is not completely symmetric, suppliers constitute <strong>an</strong> additional group of users of the system. Mak<strong>in</strong>gthis, slightly controversial assumption, we c<strong>an</strong> conceptualize data collected about suppliers <strong>in</strong> a similar way to that of thedata collected from the customers. Here the data concerns the negotiation processes (auctions), commodity characteristics(prices, amounts, time), data about the quality of delivered commodities (obta<strong>in</strong>ed from customers <strong>an</strong>d matched withsuppliers), number of product returns, reaction time to our requests etc. From this data we gather the knowledge how toconduct the negotiation, with whom, on what conditions, which products to elim<strong>in</strong>ate or <strong>in</strong>corporate.3.2 The environmentWhile the customers <strong>an</strong>d the suppliers are def<strong>in</strong>itively part of the environment that the system operates <strong>in</strong> (see [9] formore details), we have separated them due to the fact that they provide a very specific <strong>an</strong>d immediately applicable type ofknowledge about the suppliers to the system <strong>an</strong>d the customers (<strong>in</strong>clud<strong>in</strong>g potential customers) of the system itself. Theenvironment is the world <strong>in</strong> which the customers <strong>an</strong>d the suppliers <strong>an</strong>d the system operate <strong>an</strong>d that <strong>in</strong>fluences theiractions. This <strong>in</strong>fluence goes directly to the suppliers <strong>an</strong>d customers <strong>an</strong>d is mediated through them as it <strong>in</strong>fluences thesystem. In the ideal situation the system should be able to consider all factors that <strong>in</strong>fluence the customers <strong>an</strong>d their<strong>in</strong>teractions with the system. The typical examples of such factors are: economic policy, new technology, fashion trends,cultural events, weather, etc. While easy to see that these factors c<strong>an</strong> have <strong>an</strong> import<strong>an</strong>t <strong>in</strong>fluence on the e-<strong>commerce</strong>system, due to their generality they are the most difficult to p<strong>in</strong>po<strong>in</strong>t <strong>an</strong>d to build knowledge about them throughautomatic knowledge acquisition processes.3.3 ExpertsThough knowledge acquisition from hum<strong>an</strong> experts may be a very difficult task, some knowledge may be ga<strong>in</strong>ed <strong>in</strong> noother fashion. Expert knowledge is first extracted (us<strong>in</strong>g st<strong>an</strong>dard techniques) as the basis for the system’s knowledge – afoundation of hum<strong>an</strong> experience-derived assumptions, which the system’s knowledge c<strong>an</strong> grow <strong>an</strong>d evolve upon. As thesystem uses this knowledge to learn <strong>an</strong>d exp<strong>an</strong>d, the experts are still needed to deal with atypical situations, statisticaloutliers, which the automated knowledge techniques have difficulty <strong>in</strong>tegrat<strong>in</strong>g <strong>in</strong>to the system. The task of the system <strong>in</strong>these situations is to provide the hum<strong>an</strong> expert with enough <strong>in</strong>formation to help reach<strong>in</strong>g a correct decision <strong>in</strong> as easy aspossible way. In the next stage this, more <strong>in</strong>tuitive, knowledge is <strong>in</strong>corporated <strong>in</strong>to the system.International Conference on Electronic Commerce Research (ICECR-5)


<strong>Knowledge</strong> <strong>M<strong>an</strong>agement</strong> <strong>in</strong> <strong>an</strong> E-<strong>commerce</strong> <strong>System</strong> 63.4 <strong>Knowledge</strong> acquisition techniques <strong>an</strong>d methods<strong>Knowledge</strong> acquisition from the data stored <strong>in</strong> the system applies primarily to the data collected from/about the customers<strong>an</strong>d the environment (as seen above, knowledge acquisition from experts is achieved <strong>in</strong> a different, more direct, way).Techniques for knowledge acquisition from electronic data/<strong>in</strong>formation sources has been studied for some time <strong>an</strong>d somesignific<strong>an</strong>t adv<strong>an</strong>ces have been already made. The general name for the processes that extract knowledge from electronicsources is knowledge discovery from databases (see also [7, 13, 15, 17, 27]).As it was stressed <strong>in</strong> the Introduction, knowledge m<strong>an</strong>agement is a cont<strong>in</strong>uous process of knowledge acquisition,tr<strong>an</strong>sfer <strong>an</strong>d application, where by apply<strong>in</strong>g knowledge new knowledge c<strong>an</strong> be acquired aga<strong>in</strong>. Therefore, it is thelearn<strong>in</strong>g process that results from <strong>in</strong>teractions between the system <strong>an</strong>d the three data sources described above that is oneof the more import<strong>an</strong>t sources of knowledge <strong>in</strong> the system. This process mimics that of real org<strong>an</strong>izations <strong>in</strong> the worldeconomy, <strong>in</strong> which knowledge is acquired, not through a s<strong>in</strong>gle act, but through a cont<strong>in</strong>uous process of improvement onknowledge ga<strong>in</strong>ed thus far. In order to meet the requirements of a dynamically <strong>an</strong>d often unpredictably ch<strong>an</strong>g<strong>in</strong>g world,knowledge acquisition must <strong>in</strong>volve both knowledge adaptation <strong>an</strong>d knowledge evolution.<strong>Knowledge</strong> adaptation <strong>in</strong>volves addition of new knowledge to that already exist<strong>in</strong>g, the creation of entirely newknowledge, the deprecation of irrelev<strong>an</strong>t knowledge <strong>an</strong>d the deletion of dead knowledge. <strong>Knowledge</strong> acquired from theknowledge may be relatively const<strong>an</strong>t or especially liquid, because it often concerns trends <strong>in</strong> hum<strong>an</strong> behavior, whichr<strong>an</strong>ge from basic social tendencies to current fads. All of this knowledge must be correlated <strong>an</strong>d evaluated accord<strong>in</strong>g tohow much it adds to the system’s exist<strong>in</strong>g body of knowledge as well as other factors such as time dependency. Similarly,knowledge, which has ceased to be useful to the system, especially because it has become outdated, must be deprecated<strong>an</strong>d removed from the system. This <strong>in</strong>cludes time-dependent events as well as models <strong>an</strong>d techniques the system employs<strong>in</strong> bus<strong>in</strong>ess that have been superseded by better ones. In e-<strong>commerce</strong> systems adaptation is not a homogenous process.There are at least two types of adaptation: on-l<strong>in</strong>e <strong>an</strong>d offl<strong>in</strong>e. The first occurs while the customer is onl<strong>in</strong>e <strong>an</strong>d thesystem is assist<strong>in</strong>g him, while the latter is a complex process that <strong>in</strong>tegrates <strong>an</strong>d <strong>an</strong>alyzes all of the <strong>in</strong>formation collectedby the system that may <strong>in</strong>fluence system knowledge. In this case the process of knowledge adaptation of <strong>in</strong>separable fromthat of knowledge evolution. This refers to the process of apply<strong>in</strong>g trend <strong>an</strong>alysis <strong>in</strong> <strong>an</strong> attempt to <strong>an</strong>swer the questions:what is ch<strong>an</strong>g<strong>in</strong>g? how frequent does a ch<strong>an</strong>ge take place? This time dependent <strong>an</strong>alysis gives us <strong>in</strong>sight <strong>in</strong>to knowledgeevolution <strong>in</strong> the environment <strong>an</strong>d <strong>in</strong> the e-<strong>commerce</strong> system. To be able to conduct such <strong>an</strong>alysis it is necessary to collecthistorical data about the e-<strong>commerce</strong> system.4. HIGHER-LEVEL-KNOWLEDGE: GENERAL CONSIDERATIONSAt this po<strong>in</strong>t we assume that there exists a layer of low-level knowledge acquired from the sources described above <strong>an</strong>dresid<strong>in</strong>g <strong>in</strong> the knowledge base, which is divided <strong>in</strong>to source areas (user behavior, hum<strong>an</strong> experts, news, etc.). While thislow-level knowledge may be <strong>in</strong>terest<strong>in</strong>g <strong>in</strong> itself, it is not especially useful for e-<strong>commerce</strong>. In order for the system totake adv<strong>an</strong>tage of knowledge from disparate sources it must group these knowledge build<strong>in</strong>g blocks <strong>in</strong>to a functionalframework. In fact, there are m<strong>an</strong>y such frameworks exist<strong>in</strong>g <strong>in</strong> the system at the same time, each a different view on asubset of the system’s knowledge. The existence of a variety of knowledge frameworks closely parallels knowledge roles<strong>in</strong> hum<strong>an</strong> org<strong>an</strong>ization, <strong>in</strong> which no one knowledge bearer is responsible for all of the functions of the org<strong>an</strong>ization, <strong>an</strong>dthus each only requires a subset of the org<strong>an</strong>ization’s knowledge. In fact, beyond the small scale it is impossible for <strong>an</strong>yentity to possess more th<strong>an</strong> a subset of the total knowledge. It is the union of these subsets, embodied <strong>in</strong> the form of“components” (hum<strong>an</strong>, electronic or other knowledge-bear<strong>in</strong>g mediums) possess<strong>in</strong>g knowledge that def<strong>in</strong>es theorg<strong>an</strong>ization. Of course there is <strong>an</strong> <strong>in</strong>herent overlap between knowledge blocks used by various components. Differentuses of the same knowledge lead to the further ref<strong>in</strong>ement of that knowledge, as general knowledge splits <strong>in</strong>to morespecialized parts. This pattern of generalization to specialization creates a highly relev<strong>an</strong>t web-like structure of low-levelknowledge with<strong>in</strong> the system. More precisely, there may be m<strong>an</strong>y such structures built out of low-level knowledge, e.g.various departments require different knowledge to fulfill their goals. Observe also, that knowledge that is very import<strong>an</strong>tto one group may be less import<strong>an</strong>t to the other; this does not ch<strong>an</strong>ge the knowledge <strong>in</strong> itself, but only the role it plays <strong>in</strong>the knowledge framework build to support actions of a given group.International Conference on Electronic Commerce Research (ICECR-5)


<strong>Knowledge</strong> <strong>M<strong>an</strong>agement</strong> <strong>in</strong> <strong>an</strong> E-<strong>commerce</strong> <strong>System</strong> 7As the different people / departments <strong>in</strong> a traditional <strong>commerce</strong> org<strong>an</strong>ization have different roles <strong>an</strong>d employ theorg<strong>an</strong>ization’s knowledge <strong>in</strong> different fashions / from different perspectives, so too does the e-<strong>commerce</strong> system conta<strong>in</strong> avariety of components, each of which requires access to a subset of the system’s knowledge <strong>an</strong>d has a “perspective” onthat subset. For example, the components driv<strong>in</strong>g the user <strong>in</strong>terface to the system are primarily concerned with knowledgethat affects how users <strong>in</strong>teract with the system. This <strong>in</strong>cludes knowledge of hum<strong>an</strong> sensory psychology <strong>an</strong>d <strong>in</strong>terface“usability” as well as the limits of the technologies used for display<strong>in</strong>g the <strong>in</strong>terface (such as the web). All of thisknowledge is low-level, acquired from hum<strong>an</strong> experts <strong>an</strong>d system learn<strong>in</strong>g as described above. It is the perspective of theuser <strong>in</strong>terface components, which groups the otherwise disparate low-level knowledge <strong>in</strong>to a framework useful fordriv<strong>in</strong>g the user <strong>in</strong>terface.This perspective-framework is referred to as <strong>an</strong> ontology. While this term has a traditional philosophical me<strong>an</strong><strong>in</strong>g ofthe “study of be<strong>in</strong>g”, <strong>in</strong> the world of the Internet it refers to a formalized medium of represent<strong>in</strong>g knowledge. The basicstructure employed here has its orig<strong>in</strong>s <strong>in</strong> the frame-based representation described by Marv<strong>in</strong> M<strong>in</strong>sky [18]. In essence, aframe is a subset of knowledge, which is used by <strong>an</strong> “actor” to conceptualize the situation. A frame may conta<strong>in</strong> subframes,which are other perspectives subsumed by the greater picture.Figure 1. Perspectives on low-level knowledge <strong>in</strong> the systemThis idea extends naturally <strong>in</strong>to the doma<strong>in</strong> of e-<strong>commerce</strong>, where ontology is a way of structur<strong>in</strong>g the electronicreality that <strong>an</strong> e-<strong>commerce</strong> system has to deal with. S<strong>in</strong>ce various functional units have to achieve <strong>in</strong>dividual goals, eachwill need its own ontology. As <strong>an</strong> example we may consider the user <strong>in</strong>terface components that require ontology of<strong>in</strong>terface-related knowledge. Each component with<strong>in</strong> this set of user <strong>in</strong>terface components may have a smaller ontologythat refers to the specific doma<strong>in</strong> of action of that component. Thus the component ontologies are subsumed by the largerontology needed by user <strong>in</strong>terfaces, which ontology is subsumed by that pert<strong>in</strong>ent to e-<strong>commerce</strong> customer support,which <strong>in</strong> turn is encompassed by the ontology for the system itself, which refers to all knowledge available to the system<strong>an</strong>d how it is <strong>in</strong>terrelated to each other. Realistically, such a total ontology may not exist, even for such a limited doma<strong>in</strong>as e-<strong>commerce</strong> [8]. Practical considerations aside, m<strong>an</strong>y <strong>in</strong> the field of ontologies <strong>an</strong>d sem<strong>an</strong>tic process<strong>in</strong>g are currentlyattempt<strong>in</strong>g to def<strong>in</strong>e even larger ontologies, for describ<strong>in</strong>g the entirety of hum<strong>an</strong>-recognized reality [4]. In Figure 1 weillustrate the situation where two low level ontologies are utilized by two higher-level ontologies. Nodes <strong>in</strong> the ontologyInternational Conference on Electronic Commerce Research (ICECR-5)


<strong>Knowledge</strong> <strong>M<strong>an</strong>agement</strong> <strong>in</strong> <strong>an</strong> E-<strong>commerce</strong> <strong>System</strong> 8with small circles denote po<strong>in</strong>ters to low-level knowledge <strong>in</strong> the system, while the rema<strong>in</strong><strong>in</strong>g nodes signify abstractstructures with<strong>in</strong> the ontology which clarify the relations between the po<strong>in</strong>ted-to knowledge.It must be emphasized that the variety of ontologies for structur<strong>in</strong>g knowledge <strong>in</strong> the system does not ch<strong>an</strong>ge theunderly<strong>in</strong>g knowledge itself. The characteristics of a chair do not ch<strong>an</strong>ge just because we call it someth<strong>in</strong>g other th<strong>an</strong> achair, or view it from the side or h<strong>an</strong>g<strong>in</strong>g from the ceil<strong>in</strong>g. Similarly, the knowledge that a rare commodity <strong>in</strong> greatdem<strong>an</strong>d will cost more does not ch<strong>an</strong>ge, regardless of whether it is the advertis<strong>in</strong>g eng<strong>in</strong>e or the <strong>in</strong>ventory m<strong>an</strong>agermak<strong>in</strong>g use of this knowledge. However, the feedback loop of the knowledge m<strong>an</strong>agement system does allow for thispossibility, regardless of how static the knowledge may appear to be (see discussion below).There are a huge number of possible ontologies with<strong>in</strong> the e-<strong>commerce</strong> system, some of them designed forfunctional roles (such as driv<strong>in</strong>g the user <strong>in</strong>terface) <strong>an</strong>d others def<strong>in</strong>ed to provide a high-level overview of the system.These latter ontologies/perspectives are of particular <strong>in</strong>terest to us as researchers, as they allow us to observe <strong>an</strong>dm<strong>an</strong>ipulate the knowledge <strong>in</strong> the system, keep<strong>in</strong>g <strong>in</strong> m<strong>in</strong>d that even at the greatest level of abstraction these perspectivesstill refer to the low-level knowledge ga<strong>in</strong>ed from experts, user <strong>in</strong>teractions, etc., <strong>an</strong>d c<strong>an</strong>, if necessary, be concretized toknowledge actually represented <strong>in</strong> the knowledge base.F<strong>in</strong>ally, let us address the question of meta-knowledge. Its st<strong>an</strong>dard def<strong>in</strong>ition is “knowledge about knowledge.”Observe first, that ontologies c<strong>an</strong> already be considered examples of meta-knowledge, because they org<strong>an</strong>ize knowledge<strong>an</strong>d thus represent knowledge about knowledge. S<strong>in</strong>ce ontologies c<strong>an</strong> conta<strong>in</strong> ontologies, we are deal<strong>in</strong>g with ahierarchical structure, where each higher level represents a reflection about knowledge occurr<strong>in</strong>g on lover levels. In thisway our system does not support a st<strong>an</strong>dard division of knowledge <strong>in</strong>to knowledge <strong>an</strong>d meta-knowledge, but rather arelational web of knowledge that represents the system’s view of the reality it is immersed <strong>in</strong>.5. KNOWLEDGE STRUCTURE: PERSPECTIVESSo far we have <strong>in</strong>troduced the sources <strong>an</strong>d techniques of acquisition of low-level knowledge. We have also discussed theidea of apply<strong>in</strong>g ontologies to impose useful relations on the build<strong>in</strong>g blocks of low-level knowledge <strong>an</strong>d argued thatthere are multiple perspectives/ontologies on knowledge that will co-exist <strong>in</strong> the system. Let us now look at some of themore import<strong>an</strong>t of these perspectives, divided <strong>in</strong>to two categories: five that are most relev<strong>an</strong>t to hum<strong>an</strong> <strong>an</strong>alysts, <strong>an</strong>d threethat are particularly useful for implementation of the system.5.1 Empirical-model-situation knowledge structureThis ontology considers three types of knowledge with<strong>in</strong> the system: empirical knowledge, model knowledge <strong>an</strong>dsituational knowledge. It is based on how knowledge is applied by hum<strong>an</strong>s <strong>in</strong> ord<strong>in</strong>ary situations. As hum<strong>an</strong>s we possessempirical knowledge (such as the natural laws, <strong>an</strong>d also what is referred to as “common sense” but is not all thatcommon), which is the foundation of our models for view<strong>in</strong>g the world. These models, <strong>in</strong> order to be applicable toeveryday situations, must be concretized by the details of the situations. The same approach to underst<strong>an</strong>d<strong>in</strong>g theutilization of knowledge with<strong>in</strong> the e-<strong>commerce</strong> system, while heavily abstracted from the actual implementation of thesystem, allows us to underst<strong>an</strong>d the knowledge m<strong>an</strong>agement processes from a hum<strong>an</strong> perspective, <strong>an</strong>d <strong>in</strong> m<strong>an</strong>y situationscompare <strong>an</strong>d adapt the actions <strong>an</strong>d reactions of hum<strong>an</strong> commercial org<strong>an</strong>izations to that of e-<strong>commerce</strong> systems. Let usconsider each level of knowledge approached from this perspective <strong>in</strong> some detail.5.1.1 Empirical knowledgeEmpirical knowledge of the world is the most general <strong>an</strong>d most fundamental type of knowledge, <strong>an</strong>d it rema<strong>in</strong>s relativelyconst<strong>an</strong>t throughout ch<strong>an</strong>g<strong>in</strong>g models <strong>an</strong>d situations. At the core of this knowledge are the statements, pr<strong>in</strong>ciples <strong>an</strong>drules describ<strong>in</strong>g economic phenomena, which are central to the operation of <strong>an</strong>y market player. Of course, this knowledgeis available to all players, though it is rarely stated explicitly but is mostly present <strong>in</strong> the common cultural knowledge ofhum<strong>an</strong> org<strong>an</strong>izations. Examples of empirical knowledge <strong>in</strong>clude the laws of supply <strong>an</strong>d dem<strong>an</strong>d, price curves, basicmarket models <strong>an</strong>d strategies, etc. These rules are much too general to build competitive sales strategies, negotiate priceswith suppliers, or otherwise ref<strong>in</strong>e the system, but merely allow the system to operate <strong>in</strong> the marketplace.International Conference on Electronic Commerce Research (ICECR-5)


<strong>Knowledge</strong> <strong>M<strong>an</strong>agement</strong> <strong>in</strong> <strong>an</strong> E-<strong>commerce</strong> <strong>System</strong> 95.1.2 Model knowledgeThis knowledge concerns the models that govern our approach to <strong>commerce</strong>. These models evolve over time, albeit <strong>in</strong> asstable a progression as possible given the dynamics of e-<strong>commerce</strong>. As the system operates <strong>an</strong>d <strong>in</strong>teracts with users,suppliers <strong>an</strong>d the environment the general e-<strong>commerce</strong> models we have formulated are ref<strong>in</strong>ed until they are specificenough to br<strong>an</strong>ch <strong>in</strong> to different models, which may be applied <strong>in</strong> different situations accord<strong>in</strong>g to different cues.<strong>Knowledge</strong> of these cues <strong>an</strong>d when/how to apply the various customer, supply, etc. models to various situations is afurther abstraction of this model knowledge. Multiplicity of models results from generally accepted pr<strong>in</strong>ciple that there isnot s<strong>in</strong>gle model for <strong>an</strong>y s<strong>in</strong>gle complex problem (see for example [19]). We c<strong>an</strong> consider for example model of salestrategy that c<strong>an</strong> be enriched by add<strong>in</strong>g or remov<strong>in</strong>g some of its properties. <strong>Knowledge</strong> from this level, when allows thesystem to achieve its determ<strong>in</strong>ed goals may be sometimes used to modify the empirical knowledge (become one of thepr<strong>in</strong>ciples guid<strong>in</strong>g the system). Roughly speak<strong>in</strong>g knowledge on this level refers, for example, to different customerbehavior groups accord<strong>in</strong>g to their reaction to showed commodities advertisement.5.1.3 Situational knowledgeEmpirical <strong>an</strong>d model knowledge are not enough for the system to act <strong>an</strong>d react <strong>in</strong> a real situation, with real customers,real suppliers <strong>an</strong>d a real environment <strong>in</strong> which the system is mak<strong>in</strong>g (or los<strong>in</strong>g!) money. For this we require specificsituational knowledge. This knowledge is the lowest level of abstraction over the low-level knowledge ga<strong>in</strong>ed from thevarious sources, though it is high enough that we c<strong>an</strong> <strong>an</strong>alyze its play <strong>in</strong> the abstracted mech<strong>an</strong>isms of the system withoutreduc<strong>in</strong>g the situational knowledge to its constituent parts. This fast-ch<strong>an</strong>g<strong>in</strong>g, situational knowledge is often valid only <strong>in</strong>a given time frame <strong>an</strong>d sometimes only once (e.g. dur<strong>in</strong>g a session with a customer). Because the e-<strong>commerce</strong> system iscustomer-driven, most of these situations concern a s<strong>in</strong>gle customer, though the knowledge is not always limited to thatcustomer’s time-limited session but may sp<strong>an</strong> across sessions; here there is a problem <strong>in</strong> dist<strong>in</strong>guish<strong>in</strong>g which situationalknowledge is applicable only once <strong>an</strong>d which c<strong>an</strong> be applied multiple times.In fact, this knowledge is merely the “fill<strong>in</strong>g” for the model knowledge, which describes our general approach to theclient. If our bus<strong>in</strong>ess/customer model proscribes a certa<strong>in</strong> approach (e.g. especially obtrusive advertis<strong>in</strong>g), then thesituational knowledge for accomplish<strong>in</strong>g this model-def<strong>in</strong>ed objective will never come <strong>in</strong>to play. At a certa<strong>in</strong> dist<strong>an</strong>cefrom both the situational <strong>an</strong>d model knowledge effects on the <strong>in</strong>teraction with the customer is empirical knowledge,which casts a more subtle but nevertheless signific<strong>an</strong>t <strong>in</strong>fluence on the delivery of the system to the customer. We willnot advertise <strong>an</strong> item which we do not have access to a stock of, or <strong>an</strong> item that doesn’t exist. World knowledge placesthese constra<strong>in</strong>ts on our model knowledge, so that these possibilities never enter the picture of how the system functions.Here we must recognize that though the world knowledge constra<strong>in</strong>s the model knowledge which itself constra<strong>in</strong>ts thesituational knowledge, the latter are not subsets of the former, <strong>an</strong>d <strong>in</strong> fact, the relationship between them (or <strong>an</strong>yknowledge) c<strong>an</strong>not be described by sets or even object hierarchies. The relationships are cascad<strong>in</strong>g cause <strong>an</strong>d effect. Inaddition, there is a feedback loop between all three, which allows situational or model knowledge to affect worldknowledge, situational to affect model, etc. though this loop is highly restra<strong>in</strong>ed to prevent time- or other contextdependentvariables from affect<strong>in</strong>g more const<strong>an</strong>t factors.5.2 EconomicFrom this perspective, the e-<strong>commerce</strong> system may be considered solely as a market entity. The ontology of this viewdescribes the r<strong>an</strong>ge of economic laws <strong>an</strong>d theories as well as practical models of enterprise org<strong>an</strong>ization, specificallyapplied to the e-<strong>commerce</strong> doma<strong>in</strong>. Like <strong>an</strong>y other economic entity, the e-<strong>commerce</strong> system has <strong>in</strong>come, cost, profits aswell as economic strategies. This is the perspective of the economist, <strong>an</strong>d may be used to <strong>an</strong>alyze <strong>in</strong> relation to othermarket players, traditional or electronic.5.3 PersonalizationThe knowledge m<strong>an</strong>agement processes of <strong>an</strong> e-<strong>commerce</strong> system <strong>an</strong>d the knowledge that they operate on may be seen as<strong>an</strong> eng<strong>in</strong>e for personalization of user content. Some of the motivations <strong>an</strong>d mech<strong>an</strong>isms for reach<strong>in</strong>g this end through theuse of knowledge have been described previously; <strong>in</strong> earlier works [10, 23, 24] we describe some specific me<strong>an</strong>s towardInternational Conference on Electronic Commerce Research (ICECR-5)


<strong>Knowledge</strong> <strong>M<strong>an</strong>agement</strong> <strong>in</strong> <strong>an</strong> E-<strong>commerce</strong> <strong>System</strong> 10this. In general, the ma<strong>in</strong> goal of content personalization, as viewed from the e-<strong>commerce</strong> perspective, is to support theclient-system <strong>in</strong>teractions <strong>an</strong>d provide a quality substitute for the miss<strong>in</strong>g hum<strong>an</strong> salesperson. In particular, throughpersonalization the system aims at: provid<strong>in</strong>g a better quality of buy<strong>in</strong>g experience through underst<strong>an</strong>d<strong>in</strong>g <strong>an</strong>d predict<strong>in</strong>gneeds of <strong>in</strong>dividual customers, improvement of satisfaction of the customer, improved effectiveness of the system (asmeasured through the total number of items sold to each client), build<strong>in</strong>g of a br<strong>an</strong>d recognition <strong>an</strong>d a relationshipbetween clients <strong>an</strong>d e-stores, that will result <strong>in</strong> customer return<strong>in</strong>g <strong>an</strong>d mak<strong>in</strong>g additional purchases of commodities.Be<strong>in</strong>g able to directly <strong>in</strong>teract with the client <strong>an</strong>d track all of the <strong>in</strong>teractions (see Section 2), allows the system to get toknow the client <strong>an</strong>d to adjust the offer to her expectations. In addition it allows the system to <strong>in</strong>fluence her choices as wellas promote new products. An ontology designed for personalization comb<strong>in</strong>es knowledge about products <strong>an</strong>d associatedsale strategies comb<strong>in</strong>ed with knowledge of the customer <strong>in</strong> order to make relev<strong>an</strong>t suggestions. In addition, generalknowledge about products comb<strong>in</strong>ed with the <strong>an</strong>alysis of market trends <strong>an</strong>d client cluster<strong>in</strong>g allows the system to achievesuccess <strong>in</strong> deal<strong>in</strong>g with new clients. Summariz<strong>in</strong>g, the personalization eng<strong>in</strong>e perspective <strong>in</strong>corporates knowledge ofusers (described terms of populations, clusters <strong>an</strong>d <strong>in</strong>dividuals), sale strategies as well as supply dynamics.5.4 Org<strong>an</strong>izationalIn our earlier work <strong>an</strong>alyz<strong>in</strong>g abstractions of a real e-<strong>commerce</strong> system, we divided the functions of the system <strong>in</strong>tosupply support (SS) <strong>an</strong>d customer support (CS) subsystems, which <strong>in</strong>teract via a communication ch<strong>an</strong>nel [9]. Theknowledge m<strong>an</strong>agement of the system may be viewed similarly: the knowledge utilized by the supply subsystem islargely system knowledge, required for the support of tr<strong>an</strong>saction m<strong>an</strong>agement (buy<strong>in</strong>g <strong>an</strong>d sell<strong>in</strong>g) <strong>an</strong>d commoditym<strong>an</strong>agement (selection, delivery <strong>an</strong>d <strong>in</strong>ventory). The customer subsystem, on the other h<strong>an</strong>d, mimics the role of a hum<strong>an</strong>seller, support<strong>in</strong>g the customer <strong>in</strong> search<strong>in</strong>g, select<strong>in</strong>g <strong>an</strong>d purchas<strong>in</strong>g merch<strong>an</strong>dise [9]. These subsystems are largely<strong>an</strong>alogous to those found <strong>in</strong> well-established hum<strong>an</strong> org<strong>an</strong>izations: production <strong>an</strong>d development on one side <strong>an</strong>dmarket<strong>in</strong>g <strong>an</strong>d customer relations on the other.While knowledge about suppliers is mostly localized <strong>in</strong> the SS subsystem <strong>an</strong>d knowledge about customers is <strong>in</strong> theCS subsystem, the two spheres are <strong>in</strong>terdependent. The success of <strong>an</strong>y form of <strong>commerce</strong> lies <strong>in</strong> the reconciliation ofcustomer <strong>in</strong>terests with bus<strong>in</strong>ess objectives, <strong>an</strong>d e-<strong>commerce</strong> is no different <strong>in</strong> this sense. Consider that one of the goalsof the supply system is to sell the commodity it has the most of (<strong>in</strong> order to reduce <strong>in</strong>ventory) as well as the products thatare most profitable. This knowledge of products must be tr<strong>an</strong>sferred to the CS subsystem so that these products c<strong>an</strong> bepromoted, particularly to a target base of customers who are more likely to purchase them. On the other h<strong>an</strong>d, the supplysystem must const<strong>an</strong>tly accept feedback from the customer support system, <strong>in</strong> order to match the supply of a product toits dem<strong>an</strong>d. Similarly, a customer may have <strong>in</strong>dividual desires, which are not satisfied by a general selection ofcommodities, <strong>an</strong>d these must also be taken <strong>in</strong>to account by the supply system. <strong>Knowledge</strong> of this customer <strong>an</strong>d hisexpectations are crucial to both the CS <strong>an</strong>d SS subsystems – “the customer is always right” still applies to e-<strong>commerce</strong>,<strong>an</strong>d <strong>in</strong> fact it is even more import<strong>an</strong>t because of the customer’s additional expectations. F<strong>in</strong>ally, knowledge of the outsideworld is pert<strong>in</strong>ent to both subsystems: the system must function realistically (as described above) as well as follow trends<strong>an</strong>d other <strong>in</strong>formation that is not mediated through the customer <strong>an</strong>d supply sides. An ontology for the CS-SS modelsencompasses knowledge from hum<strong>an</strong> m<strong>an</strong>agement experts, product pl<strong>an</strong>ners <strong>an</strong>d supply-side coord<strong>in</strong>ators as well as thatfrom customers relations specialists. It also <strong>in</strong>cludes knowledge describ<strong>in</strong>g the <strong>in</strong>ternal processes of successfulorg<strong>an</strong>izations, from <strong>an</strong>d upon which the system may adapt <strong>an</strong>d grow through its cont<strong>in</strong>ual operation <strong>in</strong> the market.5.5 Customer support strategiesThis ontology subsumes that of personalization, <strong>an</strong>d references the entirety of knowledge necessary for serv<strong>in</strong>gcustomers. Thus there is some overlap with the economic, org<strong>an</strong>izational <strong>an</strong>d multiple other ontologies that relate toknowledge of customer support. Here the e-<strong>commerce</strong> system is perceived as a me<strong>an</strong>s to the end of customer satisfaction– “the customer is always right” – <strong>an</strong>d we consider m<strong>an</strong>y qualifications for this, for <strong>in</strong>st<strong>an</strong>ce:- <strong>in</strong>teraction with the system (easy to use, aesthetically pleas<strong>in</strong>g),- r<strong>an</strong>ge of commodities offered,- quality of service – timel<strong>in</strong>ess, returns, responsiveness to feedback, etc.,International Conference on Electronic Commerce Research (ICECR-5)


<strong>Knowledge</strong> <strong>M<strong>an</strong>agement</strong> <strong>in</strong> <strong>an</strong> E-<strong>commerce</strong> <strong>System</strong> 11- guar<strong>an</strong>tees of customer privacy,- convenient forms of payment,- the system’s behavior when a product is out of stock,as well as other service-oriented aspects, while still keep<strong>in</strong>g <strong>in</strong> m<strong>in</strong>d the profit goals of the enterprise. Thus the strategy ofthis perspective is “sell what you c<strong>an</strong> to <strong>an</strong>yone you c<strong>an</strong>.” It is impossible to only address customer needs <strong>an</strong>dexpectations; the bus<strong>in</strong>ess perspective has to be taken <strong>in</strong>to account as well. Customer support strategies have to strike abal<strong>an</strong>ce between the goals of the system <strong>an</strong>d the customer.5.6 <strong>Knowledge</strong> spacesThis implementation-level ontology categorizes knowledge created from <strong>in</strong>formation sources <strong>an</strong>d hum<strong>an</strong> experts <strong>in</strong>toknowledge about users, knowledge about the system <strong>an</strong>d knowledge about the environment <strong>in</strong> which the system resides.These are spheres of knowledge, knowledge spaces, <strong>an</strong>d are closely related to <strong>an</strong>d structured accord<strong>in</strong>g to the orig<strong>in</strong>alsources, which the referred-to knowledge is acquired from. <strong>Knowledge</strong> created from <strong>in</strong>teractions with users falls <strong>in</strong> to theuser sphere, while knowledge concern<strong>in</strong>g the system’s operation as <strong>an</strong> enterprise (<strong>in</strong>clud<strong>in</strong>g supply support) resides <strong>in</strong> thesystem sphere. <strong>Knowledge</strong> derived from the outside world is <strong>in</strong> the environment sphere. <strong>Knowledge</strong> about users is fairlyeasy to def<strong>in</strong>e (though not so easy to gather): who the users are, what they like <strong>an</strong>d dislike, how they employ the system,etc. <strong>Knowledge</strong> about the system is much broader, <strong>an</strong>d <strong>in</strong>cludes knowledge of: current <strong>an</strong>d potential products offered bythe system, suppliers of those products, the system’s bus<strong>in</strong>ess models <strong>an</strong>d operat<strong>in</strong>g parameters <strong>an</strong>d other categoriesrelated to the e-<strong>commerce</strong> system’s role <strong>in</strong> the environment. <strong>Knowledge</strong> <strong>in</strong> the environment is everyth<strong>in</strong>g that is externalto the system, but must still be known to it; the system does not exist <strong>in</strong> a void, <strong>an</strong>d if it is to succeed it must be “aware”of what of the world <strong>in</strong> which it operates. This requires knowledge of the theoretical (such as the basic laws of economy)to the practical (network outages).The knowledge space ontology divides the functions of the e-<strong>commerce</strong> system <strong>an</strong>d associated knowledge at a verylow level, <strong>an</strong>d thus it is especially useful <strong>in</strong> implementation. Components that operate <strong>in</strong> the user sphere <strong>in</strong>teract witheach other <strong>in</strong> a common doma<strong>in</strong>, while components deal<strong>in</strong>g with the environment sphere may be configured as a “wall”around the system. In addition, there are l<strong>in</strong>ks between the spheres <strong>in</strong> the form of components whose work<strong>in</strong>g ontologiesrefer to portions of multiple spheres.5.8 Dynamics of ch<strong>an</strong>geOne of the most import<strong>an</strong>t considerations <strong>in</strong> m<strong>an</strong>ag<strong>in</strong>g knowledge <strong>in</strong> a near real-time environment such as <strong>an</strong> e-<strong>commerce</strong>system is time dependence. From this perspective <strong>an</strong> ontology may be def<strong>in</strong>ed which divides knowledge <strong>in</strong>to const<strong>an</strong>t,slowly-ch<strong>an</strong>g<strong>in</strong>g <strong>an</strong>d fast-ch<strong>an</strong>g<strong>in</strong>g knowledge. This ontology may be correlated with the empirical-model-situationalontology, <strong>in</strong> identify<strong>in</strong>g const<strong>an</strong>t knowledge as empirical knowledge, slow-ch<strong>an</strong>g<strong>in</strong>g knowledge as model knowledge <strong>an</strong>dsituational knowledge as fast-ch<strong>an</strong>g<strong>in</strong>g knowledge. The correlation is not exact, of course. Fast-ch<strong>an</strong>g<strong>in</strong>g knowledge<strong>in</strong>cludes most knowledge derived from user <strong>in</strong>teractions, but may describe knowledge which was, from the hum<strong>an</strong>perspective, identified as empirical. The time-dependency categorizations are very fuzzy, because situations may arise <strong>in</strong>which const<strong>an</strong>t knowledge may require modification on the basis of fast-ch<strong>an</strong>g<strong>in</strong>g or slow-ch<strong>an</strong>g<strong>in</strong>g, model-typeknowledge. Such a shift would describe the ch<strong>an</strong>ge <strong>in</strong> th<strong>in</strong>k<strong>in</strong>g before <strong>an</strong>d after the large-scale development of theInternet as a commercial doma<strong>in</strong>. The time-m<strong>an</strong>agement functions of the system may be localized <strong>in</strong> <strong>an</strong> <strong>in</strong>dividualcomponent that “watches” for ch<strong>an</strong>ges <strong>an</strong>d <strong>in</strong>terruptions of this nature, or it may be <strong>in</strong>tegrated <strong>in</strong>to each of the exist<strong>in</strong>groles. The purpose of this ontology is to make the system “aware” of time <strong>an</strong>d the dynamics of ch<strong>an</strong>ge <strong>in</strong> the knowledgefeedback loop, so that the system does not get “stuck” or become outmoded.5.9 <strong>Knowledge</strong> tr<strong>an</strong>sfer (flow)In addition to the established, role-attached perspectives on knowledge with<strong>in</strong> the system, there is the view of knowledgeas a tr<strong>an</strong>sferable qu<strong>an</strong>tity. Here low-level knowledge is the modicum of exch<strong>an</strong>ge between knowledge entities.International Conference on Electronic Commerce Research (ICECR-5)


<strong>Knowledge</strong> <strong>M<strong>an</strong>agement</strong> <strong>in</strong> <strong>an</strong> E-<strong>commerce</strong> <strong>System</strong> 12“<strong>Knowledge</strong> tr<strong>an</strong>sfer” is a view upon a view, <strong>an</strong> ontology of ontologies, which not only describes the existence of specificknowledge but also the knowledge that governs the <strong>in</strong>teraction between users of that knowledge. In this way knowledgetr<strong>an</strong>sfer overlooks the customer-supply ontology, the org<strong>an</strong>izational ontology <strong>an</strong>d others, as well as factors which aregenerally not considered when def<strong>in</strong><strong>in</strong>g knowledge entities / “stationary” perspectives on the system:- ch<strong>an</strong>nels <strong>an</strong>d mediums of knowledge tr<strong>an</strong>sfer (electronic),- possible barriers to knowledge exch<strong>an</strong>ge,- the guar<strong>an</strong>teed delivery of knowledge to the people <strong>an</strong>d / or components for whom it is <strong>in</strong>tended,- the dem<strong>an</strong>ds <strong>an</strong>d requests of knowledge entities with<strong>in</strong> the system, <strong>an</strong>d the acceptable <strong>an</strong>d appropriate parameters forknowledge tr<strong>an</strong>sfer,- maps of knowledge <strong>an</strong>d knowledge ch<strong>an</strong>nels as well as schedules for the flow of knowledge with<strong>in</strong> the system.Though it is one of the primary functions of knowledge m<strong>an</strong>agement, knowledge tr<strong>an</strong>sfer plays a role <strong>in</strong> the system unlikeknowledge creation or application, <strong>in</strong> that it is <strong>in</strong>tegrated with the system itself. In fact, the design of knowledge tr<strong>an</strong>sferis the design of the entire knowledge system, <strong>in</strong> which knowledge is created <strong>an</strong>d applied. Here we consider knowledgetr<strong>an</strong>sfer as a very high-level perspective on the knowledge <strong>in</strong> the system, a necessary set of relations for apply<strong>in</strong>g thelow-level knowledge <strong>in</strong> a coherent way but one that is not especially concerned with the content of that knowledge.6. KNOWLEDGE APPLICATIONF<strong>in</strong>ally there is the process of knowledge application with<strong>in</strong> the e-<strong>commerce</strong> environment. This process is the only onehere described that directly supports the goals of the system. The creation of knowledge <strong>an</strong>d the design of views upon itare the foundations for knowledge application. In order for this application to be successful (<strong>an</strong>d thus for the e-<strong>commerce</strong>system to succeed <strong>in</strong> the market), the system must have the right knowledge <strong>in</strong> the right form to apply it. As has beenpreviously established, <strong>an</strong>y given ontology with<strong>in</strong> the system may be reduced to its constituent knowledge; at its lowestlevel, this knowledge is a set of production rules which ultimate direct the actions of the system <strong>in</strong> respond<strong>in</strong>g to them<strong>an</strong>y different situations it will encounter <strong>in</strong> the market.Here we return aga<strong>in</strong> to the example of the user <strong>in</strong>terface. For purposes of <strong>an</strong>alysis, we start from the end product ofthe knowledge m<strong>an</strong>agement processes: the display of items to the user. Let us suppose the user prefers items shown <strong>in</strong> listform, with no thumbnails or other images but one-l<strong>in</strong>e descriptions; this preference only applies to one type of item, suchas co<strong>in</strong>s, a type which the user is not particularly <strong>in</strong>terested <strong>in</strong> the visual description above but prefers the <strong>in</strong>formation<strong>in</strong>tensiveform of a text list. How does the system know to do this, when the default is to use thumbnails to display theco<strong>in</strong>s, a few to a page?Let us consider the variety of knowledge sources necessary for this application <strong>an</strong>d thus a possible ontology forrelat<strong>in</strong>g them. To start, the user <strong>in</strong>terface component must have a basic knowledge of item types, <strong>in</strong>clud<strong>in</strong>g co<strong>in</strong>s. Theseitem types have different ways they may be displayed – some items may only be displayed <strong>in</strong> text, others with text <strong>an</strong>doptional images, images with optional text <strong>an</strong>d so on. In addition, the ontology must <strong>in</strong>clude/reference knowledge of themethods of display, which depend on the user’s me<strong>an</strong>s of brows<strong>in</strong>g the items (this is knowledge on the level ofimplementation). This leads to knowledge of the user’s situation – not only how they are brows<strong>in</strong>g the co<strong>in</strong>s, but alsowhat they w<strong>an</strong>t to see <strong>in</strong> brows<strong>in</strong>g them. This knowledge c<strong>an</strong> only be ga<strong>in</strong>ed from <strong>an</strong>alysis of user’s past behavior,particular <strong>in</strong> relation to co<strong>in</strong>s but also <strong>in</strong> other brows<strong>in</strong>g situations. If the user has never before looked at co<strong>in</strong>s, but hasalways chosen the text-only list format for display, then he will probably w<strong>an</strong>t to view list<strong>in</strong>gs of co<strong>in</strong> items <strong>in</strong> the samefashion (If… then = a knowledge statement, this one regard<strong>in</strong>g a correlation between previous brows<strong>in</strong>g habits <strong>an</strong>d achoice the system should make). To further complicate the situation, the order <strong>an</strong>d characteristics of the list should alsobe considered variable, subject to knowledge-produced system choices.It is obvious that the user <strong>in</strong>terface component must access a large subset of the system’s knowledge <strong>in</strong> order toarrive at the display of co<strong>in</strong>s <strong>in</strong> list format for this user. What is less obvious is how this actually occurs. The <strong>in</strong>terfacecomponent is <strong>in</strong> fact divided <strong>in</strong>to m<strong>an</strong>y different sub-components, each of which makes a particular sub-choice (such asthe order of the list) <strong>an</strong>d uses a sub-ontology to do this. The entire component (to identify it as more th<strong>an</strong> its low-levelInternational Conference on Electronic Commerce Research (ICECR-5)


<strong>Knowledge</strong> <strong>M<strong>an</strong>agement</strong> <strong>in</strong> <strong>an</strong> E-<strong>commerce</strong> <strong>System</strong> 13parts) need only consider the m<strong>an</strong>ipulation of these sub-components <strong>an</strong>d the ontology, which <strong>in</strong>cludes the generalconcepts of user <strong>in</strong>terfaces.This idea extends to the entire doma<strong>in</strong> of the e-<strong>commerce</strong> system. Very high-level ontologies, from the hum<strong>an</strong> or thecomponent/subsystem/functional role perspective, should safely be able to m<strong>an</strong>ipulate high-level ontologies withoutviolat<strong>in</strong>g the logical consistency of the constituent ontologies. This is very powerful <strong>in</strong> hum<strong>an</strong> h<strong>an</strong>ds, because it allows usto m<strong>an</strong>ipulate the knowledge of the system without remov<strong>in</strong>g ourselves from the low-level function<strong>in</strong>g of it.7. KNOWLEDGE MANAGEMENT BASED ADAPTIVITYOne of the more import<strong>an</strong>t themes that permeate this paper is the fact that knowledge m<strong>an</strong>agement is synonymous withch<strong>an</strong>ge: adaptation <strong>an</strong>d evolution. It is assumed that at the beg<strong>in</strong>n<strong>in</strong>g the system operates on pr<strong>in</strong>ciples extracted fromhum<strong>an</strong> experts that represent their view of the population of the potential customers. It also <strong>in</strong>corporates a number oftheoretical laws that have been proposed as high-level abstractions of the economic reality. One of these theories claimsthat it is almost impossible to build a correct model of <strong>commerce</strong> (<strong>an</strong>d therefore also e-<strong>commerce</strong>) reality based only ontheory <strong>an</strong>d knowledge extracted from hum<strong>an</strong> experts. The system has to be adjusted to the deal with real-life customers,who may behave differently th<strong>an</strong> the theory predicted. In addition, as the time goes by, clients’ <strong>in</strong>terests <strong>an</strong>d needs ch<strong>an</strong>gedue to their ag<strong>in</strong>g as well as due to the ch<strong>an</strong>ges <strong>in</strong> the environment (e.g. bell bottom je<strong>an</strong>s were popular once). To be ableto successfully work <strong>in</strong> the const<strong>an</strong>tly ch<strong>an</strong>g<strong>in</strong>g world, the system has to be adaptive. S<strong>in</strong>ce knowledge m<strong>an</strong>agement is aprocess of const<strong>an</strong>tly adjust<strong>in</strong>g knowledge through its application <strong>an</strong>d through collection of additional data <strong>an</strong>d extract<strong>in</strong>gknowledge from it <strong>an</strong>d <strong>in</strong>corporat<strong>in</strong>g it <strong>in</strong>to the system, with the goal to const<strong>an</strong>tly improve the exist<strong>in</strong>g model(s) ofreality <strong>an</strong>d effectiveness of the operation of the system, it is knowledge m<strong>an</strong>agement that is the basis for systemadaptivity.8. CONCLUDING REMARKSIn this paper we have <strong>an</strong>alyzed the knowledge m<strong>an</strong>agement aspects of <strong>an</strong> e-<strong>commerce</strong> system, through the <strong>in</strong>tegration ofknowledge functions with the realities <strong>an</strong>d potential capabilities of e-<strong>commerce</strong>. To achieve this goal we gave discussedthe three fundamental functions of knowledge m<strong>an</strong>agement: knowledge acquisition, tr<strong>an</strong>sfer <strong>an</strong>d applications <strong>an</strong>d showedhow these functions become the basis of e-<strong>commerce</strong> system adaptivity. We have offered a number of perspectives onthe knowledge <strong>in</strong> the system, as well as some possible positions from which this knowledge may be applied to bothsystem-side <strong>an</strong>d customer scenarios. Furthermore, these perspectives allow us to reconcile the m<strong>an</strong>y different views of e-<strong>commerce</strong> <strong>an</strong>d knowledge <strong>in</strong> e-<strong>commerce</strong>, on the level of hum<strong>an</strong> <strong>an</strong>alysis as well as that of implementation. Some ofthese views may be explicitly codified <strong>in</strong>to ontologies <strong>an</strong>d utilized <strong>in</strong> m<strong>an</strong>ipulat<strong>in</strong>g the knowledge <strong>an</strong>d knowledgefunctions <strong>in</strong> the system. These ontology-perspectives allow the same low-level knowledge, ga<strong>in</strong>ed from a variety ofsources (such as hum<strong>an</strong> experts, data m<strong>in</strong><strong>in</strong>g <strong>an</strong>d automatic learn<strong>in</strong>g) to be approached <strong>in</strong> such a way that the necessaryelements are related <strong>in</strong> a m<strong>an</strong>ner most useful for the task at h<strong>an</strong>d.The results of these theoretical <strong>in</strong>vestigations are currently be<strong>in</strong>g employed <strong>in</strong> the process of implement<strong>in</strong>g ademonstrator system for e-travel support. The proposed system will be based on software agents <strong>an</strong>d will follow thegeneral decomposition proposed <strong>in</strong> [1]. The knowledge m<strong>an</strong>agement processes described here <strong>an</strong>d <strong>in</strong> [24] will becomethe keystone of the system. We will report on the progress of this implementation <strong>in</strong> the near future.References[1] R. Angryk, V. Gal<strong>an</strong>t, M. Paprzycki <strong>an</strong>d M. Gordon. Travel Support <strong>System</strong> – <strong>an</strong> Agent-Based Framework. InH. R. Arabnia <strong>an</strong>d Y. Mun (eds.), Proceed<strong>in</strong>gs of the International Conference on Internet Comput<strong>in</strong>g (IC'02),CSREA Press, Las Vegas, NV, 2002, 719-725.International Conference on Electronic Commerce Research (ICECR-5)


<strong>Knowledge</strong> <strong>M<strong>an</strong>agement</strong> <strong>in</strong> <strong>an</strong> E-<strong>commerce</strong> <strong>System</strong> 14[2] A. Baborski <strong>an</strong>d R. Bonner. M<strong>an</strong>ag<strong>in</strong>g Corporate <strong>Knowledge</strong> – Two Approaches. In <strong>Knowledge</strong> Acquisition <strong>an</strong>dDistributed Learn<strong>in</strong>g <strong>in</strong> Resolv<strong>in</strong>g M<strong>an</strong>agerial Issues, Mälardalen University Press, 2001.[3] Buss<strong>in</strong>es week 07.08.1999 http://www.bus<strong>in</strong>essweek.com/cgib<strong>in</strong>/ebiz/ebiz_frame.pl?url=/ebiz/9908/ec0803.htm[4] Project CYC, www.cyc.com, www.opencyc.org[5] T. Davenport <strong>an</strong>d L. Prusak. Work<strong>in</strong>g <strong>Knowledge</strong>: How Org<strong>an</strong>izations M<strong>an</strong>age What They Know. Harvard Bus<strong>in</strong>essSchool Press, Boston, 1998.[6] R. Davies. The creation of new knowledge by <strong>in</strong>formation retrieval <strong>an</strong>d classification. In The Journal ofDocumentation, 1989, 45-55.[7] U. M. Fayyad, G. Piatetsky-Shapiro, P. Smyth <strong>an</strong>d R. Uthurusamy. From data m<strong>in</strong><strong>in</strong>g to knowledge discovery: <strong>an</strong>overview. In U.M. Fayyad at all. (ed.) Adv<strong>an</strong>ces <strong>in</strong> <strong>Knowledge</strong> Discovery <strong>an</strong>d Data M<strong>in</strong><strong>in</strong>g, AAAI Press/The MITPress, Menlo Park, CA, 1996, 1-34.[8] D. Fensel. Sem<strong>an</strong>tic Web Enabled Web Services, presentation dur<strong>in</strong>g BIS 2002 Conference, Poznań, Pol<strong>an</strong>d, April,2002[9] V. Gal<strong>an</strong>t, J. Jakubczyc <strong>an</strong>d M. Paprzycki. Infrastructure for E-Commerce. In Nycz M., Owoc M. L. (eds.),Proceed<strong>in</strong>gs of the 10 th Conference on <strong>Knowledge</strong> Extraction from Databases, Wrocław University of EconomicsPress, 2002, 32-47.[10] V. Gal<strong>an</strong>t <strong>an</strong>d M. Paprzycki. Information Personalization <strong>in</strong> <strong>an</strong> Internet Travel Support <strong>System</strong>. In Abramowicz W.(ed.), Proceed<strong>in</strong>gs of the BIS’2002 Conference, Poznań University of Economics Press, Poznań, Pol<strong>an</strong>d, 2002,pp. 191-202.[11] J. A. Jakubczyc <strong>an</strong>d M. L. Owoc. <strong>Knowledge</strong> <strong>M<strong>an</strong>agement</strong> <strong>an</strong>d Artificial Intelligence. Argumenta Oeconomica. 1(6),1998.[12] A. Klosow. Aspekty adaptacyjne mech<strong>an</strong>izmów personalizacji w portalach <strong>in</strong>ternetowych. Prace Naukowe AE Nr931, Wydawnictwo AE, 2002[13] N. A. Lorentzos, C. P. Yialouris <strong>an</strong>d A. B Sideridis. Time-evolv<strong>in</strong>g rule-based knowledge bases. Data & <strong>Knowledge</strong>Eng<strong>in</strong>eer<strong>in</strong>g, 29(3), March 1999, 313-335.[14] P. Maglio <strong>an</strong>d R. Barrett. Intermediaries Personalize Information Streams, Communications of the ACM, 43(8),2000.[15] M. Mach <strong>an</strong>d J. Jakubczyc. <strong>Knowledge</strong> evolution trac<strong>in</strong>g <strong>an</strong>d adaptation, In Proc. Colloquia <strong>in</strong> ArtificialIntelligence. Theory <strong>an</strong>d Applications, The Fourth Conference CAI 2002, Łódź, Pol<strong>an</strong>d, to appear.[16] A. Malavi. <strong>Knowledge</strong> <strong>M<strong>an</strong>agement</strong> <strong>an</strong>d <strong>Knowledge</strong> <strong>M<strong>an</strong>agement</strong> <strong>System</strong>s,www.rhsmith.umd.edu/is/malavi/icis-97-KMS/sld024[17] H. M<strong>an</strong>nila. Inductive Databases <strong>an</strong>d Condensed Representations for Data M<strong>in</strong><strong>in</strong>g. In J<strong>an</strong> Maluszynski (ed.), Proc.Of the International Logic Programm<strong>in</strong>g Symposium. MIT Press, 1997.[18] M. M<strong>in</strong>sky. A framework for Represent<strong>in</strong>g <strong>Knowledge</strong>. In Psychology of Computer Vision, McGraw-Hill, 1975.[19] U. Merry. Practically Apply<strong>in</strong>g the New Science to Org<strong>an</strong>izations. New Science – New Tra<strong>in</strong><strong>in</strong>g & Development,Jun 1999. Available from http://pw2.netcom/~nmerry/pract1.htm[20] B. Mobasher, R. Cooley <strong>an</strong>d J. Srivastava. Automatic Personalization Based on Web Usage M<strong>in</strong><strong>in</strong>g.Communications of the ACM. 43(8), 2000[21] C. E. Nistor, R. Oprea, M. Paprzycki <strong>an</strong>d G. Parakh, The Role of a Psychologist <strong>in</strong> E-<strong>commerce</strong> Personalization,Proceed<strong>in</strong>gs of the E-COMM-LINE 2002 Conference, Bucharest, Rom<strong>an</strong>ia, September 2002, to appear.[22] I. Nonaka <strong>an</strong>d H. Takeuchi. The <strong>Knowledge</strong> Creat<strong>in</strong>g Comp<strong>an</strong>y. Oxford University Press, New York, 1995.International Conference on Electronic Commerce Research (ICECR-5)


<strong>Knowledge</strong> <strong>M<strong>an</strong>agement</strong> <strong>in</strong> <strong>an</strong> E-<strong>commerce</strong> <strong>System</strong> 15[23] M. Paprzycki, A. Gilbert <strong>an</strong>d M. Gordon, <strong>Knowledge</strong> Representation <strong>in</strong> the Agent-based Travel Support <strong>System</strong>,Proceed<strong>in</strong>gs of the ADVIS’02 Conference, Izmir, Turkey, 2002, to appear.[24] M. Paprzycki, M. Gordon <strong>an</strong>d V. Galas. <strong>Knowledge</strong> <strong>M<strong>an</strong>agement</strong> <strong>in</strong> <strong>an</strong> Agent-based E-Commerce <strong>System</strong>. InProceed<strong>in</strong>gs of the ECOM-02 Conference. Gdańsk, Pol<strong>an</strong>d, November, 2002, to appear.[25] E. Pednault. Representation is Everyth<strong>in</strong>g. Communications of the ACM. 43(8), 2000[26] D. Riecken. Personalized Views of Personalization. Communications of the ACM. 43(8), 2000[27] J. F. Roddick <strong>an</strong>d S. Rice. Towards Induction <strong>in</strong> Databases. In Proc. 9th Australasi<strong>an</strong> Information <strong>System</strong>sConference, 1998, 534-542.[28] A. Rosenbaum. Consumers Will<strong>in</strong>g to provide Personal Information <strong>in</strong> Exch<strong>an</strong>ge for Improved Service <strong>an</strong>d Benefits.Wakefield, MA. May 9, 2001. http://www.perso<strong>an</strong>lization.org/[29] J. M. Saussois <strong>an</strong>d K. Larsen. Zarządz<strong>an</strong>ie wiedzą w społeczeństwie uczącym się. In OECD org<strong>an</strong>izacja współpracygospodarczej i rozwoju. Radom, 2000.[30] J. C. Spender. Org<strong>an</strong>izational <strong>Knowledge</strong>, Learn<strong>in</strong>g <strong>an</strong>d Memory: Three Concepts <strong>in</strong> Search of Theory. Journal ofOrg<strong>an</strong>izational Ch<strong>an</strong>ge <strong>M<strong>an</strong>agement</strong>, 9(1), 1996.[31] K. Sveiby. The new org<strong>an</strong>izational wealth: m<strong>an</strong>ag<strong>in</strong>g <strong>an</strong>d measur<strong>in</strong>g knowledge-based assets. Berrett-KoehlerPublishers, S<strong>an</strong> Fr<strong>an</strong>cisco, 1997.[32] A. Tiw<strong>an</strong>a. The <strong>Knowledge</strong> <strong>M<strong>an</strong>agement</strong> Toolkit. Practical Techniques for Build<strong>in</strong>g a <strong>Knowledge</strong> <strong>M<strong>an</strong>agement</strong><strong>System</strong>, Prentice Hall, 2000.[33] F. Valera et al. Communication <strong>M<strong>an</strong>agement</strong> Experiences <strong>in</strong> E-Commerce; Communications of the ACM, 44(4),2001.[34] W. Zadrozny, M. Budzikowska, J. Chai, N. Kambhatla, S. Levesque <strong>an</strong>d N. Nicolov. Natural L<strong>an</strong>guage Dialogue forPersonalized Interaction. Communications of the ACM. 43(8), 2000.International Conference on Electronic Commerce Research (ICECR-5)

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!