1.3.1. Collective Abilities and Informal Networks
From the discussion of the knowledge-creating process in Section 1.2.2, it is clear that knowledge management depends heavily on the collec- tive abilities of many persons who share knowledge and expand and enrich it in the process. Probst et al. (2003; p. 20) state that “collective knowledge, which comprises more than the sum of the knowledge of a number of individuals, is of particular importance for the long-term survival of an organisation” (translated by the author). This collective
knowledge often turns out to develop outside the boundaries of corpo- rate hierarchies and processes.
Davenport and Prusak (1998) consider the daily practice of know- ledge workers as opposed to pre-determined organisational structures and processes. They use a metaphor of ”knowledge markets” (Daven- port and Prusak, 1998; p. 25f), in which buyers, sellers, and brokers interact to distribute knowledge. Factors such as altruism, trust, reci- procity, and repute are presented as the pricing system within these markets. The knowledge markets often disregard formal organizational structures and are driven by pragmatic factors: ”Knowledge markets cluster around formal and informal networks, so providing informa- tion about these networks is a good way to make knowledge visible”
and further (Davenport and Prusak, 1998; p. 38): ”What sounds like workplace gossip is often a knowledge network updating itself.” Krack- hardt and Hanson (1993) support the observation that the actual work- ing practice in organisations often circumvents the official hierarchies and reporting relationships and develops its own network structure.
1.3.2. The Cathedral and the Bazaar: The Knowledge Acquisition Bottleneck
Aside from the above-mentioned “unmanageability” of knowledge as claimed by Nonaka et al., there are further problems with the kind of knowledge artifacts that are produced in heavily structured knowledge processes. Wagner (2004), for example, points out that formalized know- ledge representations as the outcome of knowledge processes such as described in (Sure, 2003) suffer from several difficulties:
“Narrow bandwidth. The channels that exist to convert orga- nizational knowledge from its source (either experts or documents, or transactions) are relatively narrow.
Acquisition latency. The slow speed of acquisition is frequent- ly accompanied by a delay between the time when know- ledge (or the underlying data) is created and when the acquired knowledge becomes available to be shared.
Knowledge inaccuracy. Experts make mistakes and so do data mining technologies (finding spurious relationships). Fur- thermore, maintenance can introduce inaccuracies or in- consistencies into previously correct knowledge bases.
Maintenance Trap. As the knowledge in the knowledge base grows, so does the requirement for maintenance. Fur- thermore, previous updates that were made with insuf- ficient care and foresight (‘hacks’) will accumulate and will render future maintenance increasingly more diffi- cult [. . . ].”
Problems such as these occurring when knowledge is to be elicited for use in a KM system have been called the knowledge acquisition (KA) bottleneck (Hayes-Roth et al., 1983). Similar arguments are made for the special case of ontology development by Hepp (2007), namely, that on- tology construction is often too time-consuming, too costly, and does not reflect the end users’ understanding of the domain; furthermore, he stresses that there are conflicts for knowledge workers splitting their efforts between contributing to ontology maintenance and doing their actual work.
Wagner draws a parallel between the situation in knowledge man- agement and a dichotomy present in open-source software engineering.
As pointed out by Raymond (1998), there are two basic models of or- ganization for free software projects: in the Cathedral model, software is built by “carefully crafted by individual wizards or small bands of mages”, who control the development process and release versions of the software to the public. On the other hand, in the Bazaar model, ev- erybody is invited to contribute according to his own possibilities. Un- finished versions of the software systems are available to the public, so that debugging and testing can be supported by a large number of early users. Transferred to the KM setting, a bazaar-style development of KM tools and processes would mean that a participatory style of interaction would be encouraged in order to get as much user input as possible, even if it meant that users would have to deal with less-then-perfect responses from the system at times.
Another parallel which we see in software engineering is the shift from top-down, heavily structured methodologies such as the Rational Uniform Process (RUP) (Jacobson et al., 1999) to agile software devel- opment methodologies such as Extreme Programming (XP). The latter, for example, proposes simplicity, rapid feedback and embracing change as its fundamental principles (Beck, 2000). These principles dictate that functionality be added in incremental steps such that new code can be integrated and tested immediately and fed back into the development lifecycle. The development cycles are thus shortened to their absolute minimum, while heavyweight processes such as the RUP assume that
after thorough analysis and design phases, a more or less finished soft- ware system will be implemented at the end.
The emphasis in heavyweight processes on completing analysis and design prior to actually implementing a software system may lead to what has been called “Analysis Paralysis” and “Death by Planning”
(Brown et al., 1998)—software systems that never get out of the analysis or design phases, because too much emphasis is put on getting every analysis and design decision right the first time before implementation commences. Thus, by the time a system would finally be implemented, it might have become obsolete already. In this thesis, we will thus focus on two KM paradigms that encourage the participation of large num- bers of users in a KM system by lowering the effort needed to contribute.
This way, even if the quality of each individual contribution is lower than in a more structured KM endeavour, the chance that a “right” an- swer for a user’s knowledge need will be present in the system can be increased.