Nettime mailing list archives

<nettime> Reputation Web
Reto Bachmann-GmÃr on Mon, 23 Jul 2007 19:15:43 +0200 (CEST)

[Date Prev] [Date Next] [Thread Prev] [Thread Next] [Date Index] [Thread Index]

<nettime> Reputation Web

This is an OpenPGP/MIME signed message (RFC 2440 and 3156)
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: quoted-printable

The growth of recommendation systems during the last years has lead to
the advent of reputation systems, they can be considered as
meta-recommendation systems in that recommender and recommendations are
them self recommended, this way the concept of "reputation" is brought
to the digital space. Original plain recommendation systems where on the
"one (claimed) identity - one vote" basis quickly revealed themselves as
hardly resistant against attacks, as a remedy reputation system (or more
specifically trust metric systems) were introduced. Despite the growth
of such systems the (perceived) ability to select information based on
well-founded trust estimations as well as the ability to build a
reputation in current online systems is still low and offline channels
such as printed book are generally considered of having more reliable
and verifiable trust criteria (it should be noted that the degree of
this difference varies between domains, while in some areas and genres
publishing online is well respected in others it is regarded as
disqualifying proof of not to being good enough to find a publisher).

These systems can be distinguished along the axis of centralised -
decentralised reputation value and the axis implicit - explicit trust
assertions. The axis of decentralisation can itself be split into the
dimension of social and the dimension of technological/architectural
decentralisation: socially centralised systems treat quality as an
intersubjective global or at least as a system-wide property while
technological/architectural centralised systems rely on central hubs and
databases. These two dimensions seem to have a high degree of mutual
independence: systems like freenet implement socially centralised
information filtering while aiming complete technological
decentralisation while systems like movielens are technologically
centralised but adhere to the notion of subjectivity of quality and
reputation. The two dimensions however collide in the extremes, on one
side social decentralisation on technological centralised system can be
compromised by various forms of censorship, on the other side
technologically decentralised system with system-wide relevance metrics
may lead to a lack of incentive to collaborative behaviour when the user
community isn't sufficiently homogeneous both in terms of relevance
judgements and technological resources.

Cryptographic key infrastructures form a basis for the robustness for
reputation systems as a reputation system clashes if it is possible to
impersonate someone else an thus stealing her reputation
capital/property. However cryptographic systems, both the centralised
public key infrastructure and the more decentralised "web of trust" (s.
pgp) focus on identity not on reputation, i.e. this infrastructure
allows assertions on the probability of someone being the person she
claims to be (typically relying on the recognition of the identity by
governmental institution, virtually always for pki but also in pgp key
signing policies [2]) and not about the probability of a person's
judgements and ideas to be (considered) accurate and of relevance to the
recipient. In fact for the reputation web matching a reputation owner to
a physical person may be of little or of no importance at all, such a
mapping and its credibility is only needed when someone wants to
transfer existing reputation capital (from outer-cyberspace) to the
reputation web and vice-versa.

Currently the majority of applications on the reputation web focus on
information selection (as with collaborative filtering) and/or the
establishment of social relations (as with social software) recently
application for decentralised payment[3] have been proposed, in it's
consequence this could lead to the advent of decentralised
(pseudo-)currencies [8].

As technological foundation the semantic web and RDF technologies seem
to gain acceptance for RW applications. While the graph approach of RDF
seems to be ideal to describe the relations of trust in the RW, the
applications have shown weaknesses in the current RDF machinery in the
ability to express probabilistic statements as well as in the ability to
quote and to express provenance of assertions which are both crucial for
the RW (note: reification is unsuitable for quoting).
Proposed approaches include the concept of named graphs [5] especially
with the extending concept of RDF-molecules[9] as well as the
introduction of quadruples [6]. Despite these flaws the ability of
computer systems to grasp the semantic content of documents greatly
improves their ability to reliably estimate reputation, for instance the
PageRank [7] algorithm which does not use semantic web technologies is
unable to distinguish a link in the "hall of fame" from on in the "hall
of shame", both links equally [!] increase the ranking of the linked

References (incomplete)
1. http://www.levien.com/free/tmetric-HOWTO.html
2. http://www.debian.org/events/keysigning
3. Fugger, Ryan, "Decentralized Payment",
4. http://www.mindswap.org/papers/Email04.pdf
5. Carroll, Stickler
6. Andreas Harth and Stefan Decker: : Yet Another RDF Store: Perfect
Index Structures for Storing Semantic Web Data With Context, DERI
Technical Report, 2004.
7. google pagerank:....
8. Komarov, Alexander, Geekcredit, http://home.gna.org/geekcredit/
9. L. Ding, T. Finin, Y. Peng, P. Pinheiro da Silva, D. L. McGuinness,
2005, http://www.ksl.stanford.edu/people/pp/papers/Ding_ISWC_2005.pdf

Content-Type: application/pgp-signature; name="signature.asc"
Content-Description: OpenPGP digital signature
Content-Disposition: attachment; filename="signature.asc"

Version: GnuPG v1.4.7 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org



#  distributed via <nettime>: no commercial use without permission
#  <nettime> is a moderated mailing list for net criticism,
#  collaborative text filtering and cultural politics of the nets
#  more info: majordomo {AT} kein.org and "info nettime-l" in the msg body
#  archive: http://www.nettime.org contact: nettime {AT} kein.org