4. How to loose 1,000,000,000 US$ in half a day
Failures:
Low quality of data collecting process, hence
Via @Bauckhage currency of information was not considered
Leiden, July 11, 2012 Steffen Staab 4
6. Definition of Trust
http://en.wikipedia.org/wiki/Trust_%28social_sciences%29
Definitions of trust[2][3] typically refer to a situation
characterised by the following aspects:
One party (trustor) is willing to rely on the actions of another
party (trustee);
the situation is directed to the future;
the trustor (voluntarily or forcedly) abandons control over the
actions performed by the trustee;
as a consequence, the trustor is uncertain about the
outcome of the other's actions; he can only develop and
evaluate expectations. The uncertainty involves the risk of
failure or harm to the trustor if the trustee will not behave as
desired.
Leiden, July 11, 2012 Steffen Staab 6
7. Core Question
What does it mean
if
I say that I trust person P performing action A
or a
System S trusts person or system P performing action A
?
Leiden, July 11, 2012 Steffen Staab 7
9. Information &
Communication Perspective
on Trust
Leiden, July 11, 2012 Steffen Staab 9
10. Communication: Should we trust that people understand what we mean?
Failure: Unintended recipient of information
Leiden, July 11, 2012 Steffen Staab 10
13. Information and Communication - Credibility
About the truth of information in which we trust, believability [Metzger 07]:
Expertise
Trustworthiness
Sender Receiver
Message
Leiden, July 11, 2012 Steffen Staab 13
15. Online Investment Scams
Trust
Identity of
Risk debitors
(greed)
BTW: I am
not sure you
can trust this
website
Failure: Identity scam & users missassess trustworthiness
Leiden, July 11, 2012 Steffen Staab 15
16. The education perspective
What do I need to teach someone such that
If he trusts person P performing action A he does not suffer?
[Metzger, 2007]
A system S trusts person P performing action A
if P satisfies the assessment criteria for trust
Leiden, July 11, 2012 Steffen Staab 16
17. Credibility: What people do and what they should do
(Teach to) Assess credibility by checklist:
accuracy,
Can be
authority, reasoned
• Identity, qualifications with in the
– Whois, Traceroute, NSlookup/Dig semantic
objectivity, web if
currency, described!
(eg. [Schenk])
and coverage or scope
Internet users may be easily
Assessing what people do on the Web: Interestingly, what focus-group participants
said they looked for in assessing credibility wasdesign. the researchers found they
duped by slick Web not what
⇒meta strategies
actually looked at during the observational portion of the study.
Leiden, July 11, 2012 Steffen Staab 17
19. A system S trusts person P performing action A
if P belongs to trusted group
Leiden, July 11, 2012 Steffen Staab 19
20. Social Science – 1
Luhmann:
Levels of increasing freedom to act
Familarity
Based on what we know
No deviation from the known
Confidence
Founded on laws, fallback positions,...
Trust
Acting under risk
Trust is the reduction of complexity
Leiden, July 11, 2012 Steffen Staab 20
21. Social Science – 2: Social Theory of Balance
Friend
Foe
Leiden, July 11, 2012 Steffen Staab 21
23. Structural Balance for Groups of 3
Definition:
A triangle is balanced if all 3 relations between the
nodes are positive or if there is exactly one positive
relationship
Leiden, July 11, 2012 Steffen Staab 23
24. Structural Balance for a Network
Definition:
A network is called structurally balanced if all groups of
triangles are structurally balanced.
Balance Theorem:
If a labeled complete graph is balanced,
then either all pairs of nodes are friend,
or else the nodes can be divided into two groups, X and Y,
such that each pair of people in X likes each other,
each pair of people in Y likes each other,
and everyone in X is the enemy of everyone in Y.
Leiden, July 11, 2012 Steffen Staab 24
26. Weakly Balanced Networks
Definition of Weak Structural Balance Property:
There is no set of three nodes
such that the edges among them consist of
exactly two positive edges and one negative edge
Leiden, July 11, 2012 Steffen Staab 26
29. Risk Management/Economic perspective
Risk is a pair
Value/Cost of an event arising
Probability that the event will arise
Trust means willingness to bear a risk
A system S trusts person P performing action A
if the expected overall value/utility is positive
In particular trust issues arise in markets with information
asymmetry – e.g. E-Bay
Leiden, July 11, 2012 Steffen Staab 29
30. A Web Market Example
Trust issues arising in markets with information asymmetry
– e.g. E-Bay
Assume 50 good cars, 50 bad cars could be for sale
200 buyers willing to buy
Assume buyers are willing to pay up to 12 for good cars
and up to 6 for bad cars
Assume sellers are willing to sell from 10 upwards and 5
upwards for good and bad cars respectively
Information Asymmetry:
Sellers judge good/bad accurately
Buyers cannot judge good/bad at all, but know about
willingness of sellers to sell
Leiden, July 11, 2012 Steffen Staab 30
31. Economic perspective: Expected Value
Information Asymmetry:
Expected value of a car for a buyer at most (12+6)/2=9
At 9 sellers of good cars do not sell, therefore rational buyers
cannot expect any good cars to be on the market!
• No good cars are sold, because of a lack of trust!
Self-fulfilling expectations!
Market Failure!
Solution: Reputation reduces information asymmetry
If ¾ of cars sold as good cars are good, then expected
value is ¾*12+1/4*6=10.5 – i.e. good cars can be sold!
Leiden, July 11, 2012 Steffen Staab 31
32. Asymmetric information and Trust signals
Used car markets
Partial remedy
• Guarantees by traders
– Reduces subsequent costs for buyers of lemons
– Strong signal that the car has decent quality
Labor market
Partial remedy:
• Education certificates
– Education leads to knowledge
– Certificate is signal for intellectual and work capacity
Insurance
Buyer of insurance knows more
• Very partial remedy: incentives system to take sports
courses
Leiden, July 11, 2012 Steffen Staab 32
34. +++ „Los Angeles (dpa) – In der
kalifornischen Kleinstadt Bluewater
soll es nach einem Bericht des
örtlichen Senders vpk-tv zu einem
Selbstmordanschlag gekommen
sein. Es habe in einem Restaurant
zwei Explosionen gegeben...“ +++
German Press Agency DPA, 10 Sep 2009
Leiden, July 11, 2012 Steffen Staab 34
35. Guerilla Marketing
Failure: Information sources had no reputation from
Leiden, July 11, 2012
third parties!
Steffen Staab 35
36. Reputation perspective
Belief in benevolence vs believe in competence
A system S trusts person P performing action A
if sufficient reputation could be aggregated
Leiden, July 11, 2012 Steffen Staab 36
37. Reputation scoring as link prediction
me
Predict which unknown link would also be good
to have
Standard algorithm: find friends-of-friends
Leiden, July 11, 2012 Steffen Staab 37
39. Friend of a friend
Leiden, July 11, 2012 Steffen Staab 39
40. Reputation Scoring in Social Networks
• Some variation of link prediction (here is just one – big - family of methods)
• Counting and weighting paths
Leiden, July 11, 2012 Steffen Staab 40
42. Distrust computation
Prediction of negative links
Few networks with negative links (Slashdot zoo)
Several methods for handling negative links available
Social factors
Facebook unlinking prediction [Quercia et al]
Age gap
Low number of common friends (embeddedness)
No common female friend
One neurotic or introvert
Results seem to be comparable to „unlinking“ in real life
Leiden, July 11, 2012 Steffen Staab 42
44. Hacked Web Sites: Did government post this?
Failure: IT security failed
Leiden, July 11, 2012 Steffen Staab 44
45. Security perspective
Authorization
Specific person P is allowed to do action A
Authentication
Proof to be a specific person
Sometimes: Tokens that lend authority and/or
authentication via centralized or decentralized trust center
A system S trusts Person P to perform A if
authentication and authorization can be proven
Leiden, July 11, 2012 Steffen Staab 45
48. Trust AND Web Data
WWW /
How does Trust deviate for Web Data? E-Mail
TCP/IP
People are coupled more loosely
• Fewer possibilities for
– Reputation building
– Personal ties
Increased chance of encountering misbehavior
• Decentralization on the Web
Web data does not focus trust – it only extends the issue
WWW /
E-Mail
TCP/IP
Leiden, July 11, 2012 Steffen Staab 48
49. Conclusion
Survey of trust issues
Incomplete
Interdisciplinary
Interwoven
• With each other
– E.g. trust/reputation as computed from social network analysis
• With further Web topics
We need
Experiments So far:
Models strengths in one of these
Analytic techniques areas, but not in all!
Leiden, July 11, 2012 Steffen Staab 49
51. Survey type articles
Luhmann: Vertrauen - ein Mechanismus der Reduktion sozialer Komplexität (1968)
N. Luhmann, Trust and Power. John Wiley & Sons, 1979.
Jin-Hee Cho, Ananthram Swami, Ing-Ray Chen, A Survey on Trust Management
for Mobile Ad Hoc Networks, IEEE COMMUNICATIONS SURVEYS &
TUTORIALS, VOL. 13, NO. 4, FOURTH QUARTER 2011
S. Staab et al., “The Pudding of Trust,” IEEE Intelligent Systems, vol. 19, no. 5, pp.
74-88, 2004.
Donovan Artz, Yolanda Gil: A survey of trust in computer science and the
Semantic Web. J. Web Sem. 5(2): 58-71 (2007)
Jennifer Golbeck (2008) "Trust on the World Wide Web: A Survey", Foundations
and Trends in Web Science: Vol. 1: No 2, pp 131-197.
http:/dx.doi.org/10.1561/1800000006
Piotr Cofta (2011) "The Trustworthy and Trusted Web",
Foundations and Trends in Web Science: Vol. 2: No 4, pp 243-381.
http://dx.doi.org/10.1561/1800000016
Miriam J. Metzger: Making sense of credibility on the Web: Models for evaluating
online information and recommendations for future research. JASIST 58(13):
2078-2091 (2007)
Leiden, July 11, 2012 Steffen Staab 51
52. Specific articles/books:
Jennifer Golbeck PhD Thesis U Maryland
Jerome Kunegis PhD Thesis U Koblenz
Sepandar D. Kamvar, Mario T. Schlosser, Hector Garcia-Molina: The Eigentrust
algorithm for reputation management in P2P networks. WWW 2003: 640-651
Simon Schenk, Renata Queiroz Dividino, Steffen Staab: Using provenance to
debug changing ontologies. J. Web Sem. 9(3): 284-298 (2011)
Xian Li, Timothy Lebo, Deborah L. McGuinness: Provenance-Based Strategies to
Develop Trust in Semantic Web Applications. IPAW 2010: 182-197
Luca de Alfaro, Ashutosh Kulshreshtha, Ian Pye, B. Thomas Adler: Reputation
systems for open collaboration. Commun. ACM 54(8): 81-87 (2011)
R. Guha, R. Kumar, P. Raghavan, and A. Tomkins, “Propagation of trust and
distrust,” in Proceedings of the 13th international conference on World Wide Web.
ACM, 2004, pp. 403–412.
Daniele Quercia, Mansoureh Bodaghi, Jon Crowcroft. Loosing “Friends” on
Facebook. In: Proc. WebSci 2012, Evanston, June 2012. ACM.
Leiden, July 11, 2012 Steffen Staab 52
Notes de l'éditeur
The assessment of the believability of a message depends on source, receive as well as on the message itself; Incuding source‘s expertise, trustworthiness and ! Attractiveness (Halo effect; most expensive consulting companies have the best looking Consultants; turning this around: if all your consultants look great, then you probably pay too much) Interesting: Metzger creates a dichotomy between authoritative publishers in previous times and the low authority of publishing on the web – whereas at any time, it was not truth that was most published, but the stuff the sold best (which might be news about inexisting UFOs, Hitler‘s book, etc.)
Accuracy refers to the degree to which a Web site is free from errors, whether the information can be verified offline, and the reliability of the information on the site. The author- ity of a Web site may be assessed by noting who authored the site and whether contact information is provided for that person or organization, what the author’s credentials, quali- fications, and affiliations are, and whether the Web site is recommended by a trusted source. Objectivity involves iden- tifying the purpose of the site and whether the information provided is fact or opinion, which also includes understand- ing whether there might be commercial intent or a conflict of interest on the part of the source, as well as the nature of relationships between linked information sources (e.g., the meaning of “ sponsored links ” on a Google search output page). Currency refers to whether the information is up to date. Coverage refers to the comprehensiveness or depth of the information provided on the site. These recommenda- tions require a range of activities on the part of users, from simple visual inspection of a Web site to more laborious information verification and triangulation efforts. CURRENCY: TELL STORY ABOUT UNITED AIRLINES
- restrict to unweighted, undirected, unipartite graphs.