Peer-to-peer Knowledge Management
Propose the idea of P2P knowledge management
What does P2P mean?
How can peer-to-peer help KM?
Discuss problems with P2P KM
Free-ride Issue and incentive design
Impact of technology on organizational exchange network
De Liu, CREC
What does Peer-to-Peer mean?
Participating computers are both clients and servers.
Three kinds of P2P applications
P2P information sharing (Napster; FreeNet)
distributed resources and/or peer-to-peer routing
P2P Collaboration/Communication (groove; ICQ)
Distributed computing (SETI@home)
Coordinate multiple computers for complex tasks
De Liu, CREC
More Features of P2P information sharing
Information available at certain time is the sum of the information content each node shares.
Nodes can easily join or quit.
Make use of under-utilized resources
It usually create some redundancy to reduce the impact of single node failure.
Load-balancing mechanism.
De Liu, CREC
Drawbacks of P2P information sharing
Anonymous . Copyright
Redundancy . Up-to-date
Quality of Service
Search Efficiency
Free-Ride!
Nearly 70% of Gnutella users share no files, nearly 50% of all responses are returned by the top 1% of sharing hosts. … We argue that free riding leads to degradation of the system performance and adds vulnerability to the system. If this trend continues copyright issues might become moot compared to the possible collapse of such systems. (Adar and Huberman 2000, Free Riding on Guttela)
De Liu, CREC
Why we propose P2P KM
P2P technology fit well with KM objectives
Unlock “hidden” organization knowledge
Enhance interactions through “push” & “pull”
Foster knowledge sharing communities
Contribute to propagation of tacit knowledge
Tolerate fuzzy knowledge and different points of view
We expect free-riding issue can be better handled within organization context.
Other desirable features: flexible, robust, secure, cost efficiency (?)
De Liu, CREC
So what is P2P KM?
two components:
P2P information sharing component explicit knowledge transfer
Notification of update and new release (“push”)
Access control
Content-based addressing
P2P collaboration component implicit knowledge transfer
Multiple channels (text, multi-media)
Real-time collaboration and non-real time collaboration
P2P KM and centralized effort of knowledge management
De Liu, CREC
Potential problems of P2P KM
Quality of knowledge
Lack of consistency
Free-rider problem
Knowledge is often treated as public good.
It’s rational for self-interest agents to consume public good without contributing (free ride).
Consequently some socially desirable good is not provided by agents due to inadequate compensation.
This under-provision of public good problem may eventually defeat the P2P knowledge management system.
De Liu, CREC
Related IS/Mgmt Literature
Most research in IS/Mgmt try to ask a question: “what are reasons for knowledge sharing”
Altruism (constant et al 94, Palfrey and Prisbrey 1997, Davenport and Prusak 1998) & Group commitment (O’ Reilly an Chatman 1986; Kollock 1999)
Reciprocity (Davenport and Prusak 1998; Ekeh 1974)
Reputation (Davenport and Prusak 1998, Rheigngold 1993)
self-efficacy (Kollock 1999, Bandura 1995)
Factors favor these motivations
future interaction; Record of past interactions; Identity persistence; Well defined and defended group boundary;Visibility and recognition of contributions; Group size; etc (Kellock 1999)
De Liu, CREC
Related Economics Literature
Knowledge and Knowledge exchange
Value of knowledge is difficult to measure or verify;
Value of knowledge may be different for different agents
Value of knowledge may involves some uncertainty.
Knowledge exchange is often embedded and affected by social networks.
So, unlike other more tangible goods, there is lack of a price system (market) or complete contract form for knowledge.
De Liu, CREC
Related Economics Literature
Wage schedules that encourage team production collaboration. (Multi-Agent: Che & Yoo 2001; Grooves 1973;)
Should competition or collaboration be encouraged?
What is the optimal wage schedule for teams?
Incentives and Task structure (Multi-task: Holmstrom, Milgrom 1982; Itoh 1991)
Either the task itself can be rewarded or the marginal opportunity cost for the task can be lowered by removing or reducing the incentives on competing tasks.
De Liu, CREC
Network (Jackson & Wolinsky 1995; Watts 1999; Kranton 1996, 1999, 2000)
What kind of network structure is optimal? (P2P? Star?)
Can agents form efficient network?
Questions and Suggestions?
@
De Liu, CREC
(Resources are essentially distributed; server doesn’t have the power)
(cost efficiency? Secure? Performance? etc…)
Why we pursue p2p
no way to do it in a centralized way
P2p represents a better solution
Last time: P2P provides an alternative way of managing information resources.
How well these motivations can work to achieve knowledge sharing efficiency?
Can organizations achieve efficiency through appropriate incentive alignment mechanisms?
