Information Systems Frontiers 2:3/4, 299±315, 2000# 2000 Kluwer Academic Publishers. Manufactured in The Netherlands.Oracles, Bards, and Village Gossips, or Social Roles and MetaKnowledge ManagementSimon Masterton* and Stuart WattKnowledge Media Institute Open University, Walton Hall, MiltonKeynes, UK. MK7 6AAE-mail: [email protected]. Knowledge management systems are used widely inmany different organizations, yet there are few models andtheories which can be used to help introduce and apply themsuccessfully. In this paper, we analyze some of the more commonproblems for knowledge management systems. Using thisbackground, we adapt models and theories from social andorganizational psychology and computer supported collabora-tive work, and discuss a variety of different knowledgemanagement systems in these contexts. We argue that knowl-edge management systems routinely adopt different social roleswithin an organization, and that these social roles can have amajor in¯uence on a system's acceptability. With theseprinciples in mind, we draw out some general practical lessons,and a ``character space'' framework, which can help to informthe design of future knowledge management systems, so as tominimize the problems of acceptability within a givenorganization.Key Words. knowledge management tools, intelligent agents,groupware, social roles``Where is the wisdom we have lost in knowledge?Where is the knowledge we have lost in information?''T. S. Eliot, Choruses from the rock``And `T. S. Eliot' is an anagram of `toilets', so heshould know''Alan Plater, Oliver's Travels1. IntroductionLike many things in life, knowledge management isimportant and useful, but hard to do well. For thisreason, a large number of tools are often used to help,some of which have been specially developed for thepurpose. More often than not, these tools are neverreally assessed to see how good they are for thepurpose, but even when they are (e.g., Ackerman,1994; Masterton, 1997, 1998), the lessons rarely makeit into other systems.In other words, the ®eld of knowledge managementhas a knowledge management problemÐan issue wewill come back to later in this paper. There is a lotexpertise out there, spread out through people'sexperiences with many different systems. It is timeto take stock, to look at some of the problems thathappen in people's actual use of knowledge manage-ment systems, and to draw some lessons from theseproblemsÐlessons which can be used to inform thefuture design of this kind of system.Knowledge management is an issue for organiza-tions, so it inevitably involves many different peopledoing many different jobs. Some may be mostlyproviders of knowledge, for example, while othersmay be organizers of it, or users of it. Because thereare people in different roles involved in the knowl-edge management processes, knowledge managementis an example of ``multiple role groupware'' (Watt,1993), and has a tendency to ``role con¯ict'' when theneeds of the different participants are not commensur-able. This often shows up as a disparity between whodoes the work and who gets the bene®t. A commonreason for the failure of multiple role groupwaresystems is when the system itself is biased in such a*Now at the Knowledge Mapping and Structuring Unit, Unilever,Port Sunlight, Bebington, Wirral, UK. CH63 3JW. E-mail:[email protected] that it reinforces the role con¯ict (Watt, 1993;Grudin, 1994)Ðand knowledge management systemsare easy prey to the same problem.To put it simply, intelligent agents are an elegantsolution to this problem. Without going into them intoo much detail, intelligent agents are computerprograms that are constantly running around actingon the behalf of one or many people (Maes, 1994;Watt, 1997). The advantage of intelligent agentsystems is that the normal quid pro quo rules ofhuman social interaction need no longer apply. Agentscan be constantly on the lookout for relevantinformation, or for links between different bits ofinformation, and although they never do it as well as aperson would, they don't expect to be paid for theirtrouble. They can make admirable assistants, inprinciple (Maes, 1994), but they can easily performother, very different, but useful, roles in knowledgemanagement, for example, as matchmakers (Fonerand Crabtree, 1996), newspaper editors (Domingueand Scott, 1988), librarians (Watt, 1998), bards(Masterton, 1997, 1998), and village gossips(Krulwich and Burkey, 1996, 1997). In effect, theycan take the roles where there isn't any bene®t,leaving people in a position where they always get thebest from the system.This justi®es agents as a technology for knowledgemanagement, but in another sense, there is a kind ofinevitability about it. Not only is the human-humanrelationship fundamentally social, so is the human-computer relationship (Nass et al., 1994).On the whole, then, there is a place for agentsystemsÐalthough very understated, not personi®ed,and rather role-centered agent systemsÐwithinknowledge management. But there are dangers.Overblown, very anthropomorphic agent systems arerarely successful outside the ®elds of entertainmentand education. So there are risks. But careful designcan, both in theory and in practice, offer agent systemsthat work effectively in knowledge management.Indeed, we will come back to a few of these later, inSection 4.Independently of whether or not they take anagent-based approach, knowledge management sys-tems are very diverse, playing many different roleswithin an organization. It is hard to take lessons froma system in one role and apply those lessons in otherroles, because these different roles shape verydifferent patterns of interaction between the peopleand the systems involved. For this reason, knowledgemanagement systems pose some rather special andrather hard problems for those who wish to assess theusability of these systems, and learn from it.In this paper we are going to look at the usability ofknowledge management systems a bit more system-atically. We are going to look at some of the mostcommon problems with knowledge managementsystems, and how some existing systems have triedto alleviate those problems. We will show this inpractice by discussing some example knowledgemanagement systems as case studies. With thisfoundation, we will draw out some of the lessons tobe learned, and argue that an awareness of opportu-nities offered by new agent technology, and anawareness of the importance of usability,