A Multifaceted, Personalisable, Specialisable Model of Trust

1
A Multi-faceted, Personalisable, Specialisable
Model of Trust

• Karl Quinn,
• Knowledge Data Engineering Group,
• Trinity College Dublin,
• Ireland.
• 4th International Conference on Trust Management
  (iTrust 06), Pisa, Italy, May 2006

2
Overview

• Trust Hypothesis.
• Multi-faceted, Personalisable, and Specialisable
  Model of Trust.
• Experimentation Evaluation.
• Generating Personalised Models of Trust.
• Application Use-case.
• Summary.

3
Definitions of Trust

• Many definitions use synonyms or trust inspiring
  terms...
• Belief McKnight 1
• Credibility or Reliability Golbeck 2
• Confidence or Faith -Shadbolt 3
• Reputation Golbeck 4
• Competence and Honesty Grandison 5
• Competence and Reliability -McAllister 6
• Trust is a fashionable but overloaded term with
  lots of intertwined meanings 7
• - Why is Trust Bad for Security, Dieter
  Gollmann, Keynote, Policy 2005.
• Editor in chief International Journal of
  Information Security

4
iTrust 2006 Live
5
Multi-faceted Model of Trust
Upper Model
OWL
Meta Model
6
Multi-faceted, Personalised, Model of Trust
Personalised Model
Upper Model
Jena
OWL
OWL
Meta Model
Protégé
Domain Specific Model
OWL
7
Experiment Goals

• Four goals in which we wanted to evaluate
• Upper ontology
• Are the trust concepts useful to people in
  determining trust?
• Meta-model
• (a) Does the hypothesis that certain concepts are
  abstract or concrete prove correct?
• Need for personalisation
• Is personalisation required when modelling trust,
  and to what extent?
• Examine whether individual model of trust change
  as context alters

8
Experiment Overview

• Computer Science Psychology TCD
• Three scenarios 10, 100, 1000
• Rate each trust concepts on a Likert scale,
• Rank the concepts 1, 2, and 3,
• Finally, choose a concept that most influenced
  each of the concepts ranked 1, 2, and 3.
• Target audience
• Advertised via email to TCD, Ericsson, forums w/
  U2 tickets to entice,
• 282 fully completed questionnaires.

9
Experiment Results 1

• Upper Ontology
• Q How useful is concept X to you when
  determining trust in the seller in scenario Y?

78
91
10
Experiment Results 2

• Upper Ontology
• An SPSS analysis of the data found a positive
  correlation, 0.318 at the 0.01 level of
  significance, between increasing risk and the
  increasing Likert scales the subjects provided
  for the trust concepts.
• OR
• As risk increases so too does the subjects
  reliance on the trust concepts also rise.

11
Experiment Results 3

• Meta-Model
• Concepts with the lowest frequencies across all
  three ranking scores faith (3.7), confidence
  (7), and belief (8.8).
• Concepts with the highest amount of very low and
  low Likert scores across all three scenarios are
  belief (19) and faith (16.5).
• Least influential concepts at every risk level
  are faith (3.8), belief (7.2), and confidence
  (7.9).

12
Experiment Results 4

• Personalisation (argument for)
• Differences in weights that individual subjects
  assign to certain trust concepts
• Differences
• Reputation is most highly ranked concept.
• Average of 82 number one votes across all
  scenarios,
• Yet this only reflects 29 of overall vote,
• 71 think differently.
• Similarities
• Faith has the least amount of number one votes.
• Average of 7 number one votes across all
  scenarios,
• 97.5 do not rank faith number one.

13
Experiment Results 5

• Personalisation (further argument for)
• Differences in weights that individual subjects
  assign to certain trust concepts as show in 2006
  experiment via personalised model generated by
  HITS algorithm

Aggregate Rankings for Reputation and Faith over
282 Subjects (Experiment 2006)
14
Experiment Results 6

• Personalisation
• Range and diversity of individual models of
  trust found within a population

Aggregate Rankings over 282 Subjects (Experiment
2006) - Via HITS algorithm -
Aggregate Rankings over 279 Subjects (Experiment
2004) - Asked Directly -
15
Personalisation 1

• Based on Kleinbergs Hypertext Induced Topic
  Search (HITS) algorithm.

16
Personalisation 2
17
Application Use-case
18
Conclusions

• Conclusions
• Multi-faceted, Personalisable model of trust.
• Domain specific.
• Evaluation.
• Application use-case scenario.
• Future Work
• Evaluate accuracy of the model.

19
Publications

• PUBLICATIONS 2004
• Quinn, K., O'Sullivan, D., Lewis, D., Wade,
  V.P., 'Composition of Trustworthy Web Services',
  Information Technology and Telecommunications
  Annual Conference, Limerick, Ireland, October
  20-21, 2004.
• Brennan, R., Quinn, K., O'Sullivan, D., Lewis,
  D., Wade V.P., 'On the Application of Paired
  Comparison to Trust', 2nd International Workshop
  on Managing Ubiquitous Communications and
  Services (MUCS), Dublin, Ireland, December 13-14,
  2004.
• PUBLICATIONS 2005
• Quinn, K., O'Sullivan, D., Lewis, D., Wade,
  V.P., 'deepTrust Management Application for
  Discovery, Selection, and Composition of
  Trustworthy Services', 9th IFIP/IEEE
  International Symposium on Integrated Network
  Management (IM 2005), Nice,France, 15-19 May,
  2005 . Feeney, K., Quinn, K., O'Sullivan, D.,
  Lewis, D., Wade, V.P., 'Relationship-Driven
  Policy Engineering for Autonomic Organisations',
  IEEE 6th International Workshop on Policies for
  Distributed Systems and Networks (POLICY 2005),
  Stockholm, Sweden, 6-8 June, 2005 .
• Quinn, K., O'Sullivan, D., Lewis, D., Wade,
  V.P., 'Policy Ontologies Panelist', 4th
  International Semantic Web Conference, Semantic
  Web Policy Workshop (SWPW 2005), Galway, Ireland,
  7th November, 2005 . t
• PUBLICATIONS 2006Quinn, K., Kenny, A., Feeney,
  K., O'Sullivan, D., Lewis, D., Wade, V.P., 'A
  Framework for the Decentralisation and Management
  of Collaborative Applications in Ubiquitous
  Computing Environments ', 10th IEEE/IFIP Network
  Operations and Management Symposium (NOMS 2006),
  Vancouver, Canada, 3-7 April, 2006. Quinn, K.,
  O'Sullivan, D., Lewis, D., Wade, V.P., 'The
  Design, Generation, and Utilisation of a
  Semantically Rich Personalised Model of Trust',
  4th International Conference on Trust Management
  (iTrust 2006), Pisa, Italy, 15-19 May, 2006.
  Quinn, K., O'Sullivan, D., Lewis, D., Wade,
  V.P., 'Trust Meta-Policies for Flexible and
  Dynamic Policy Based Trust Management', 7th
  International Workshop on Policy (POLICY 2006),
  London, Ontario, Canada, 5-7 June, 2006.

20
References

• 1 McKnight, H.D., Chervany, N.L., The Meanings
  of Trust Technical Report 94-04, Carlson School
  of Management, University of Minnesota, 1996.
• 2 Golbeck, J., Hendler, J., Parsia, B. Trust
  Networks on the Semantic Web, 12th International
  Web Conference (WWW03), Budapest, Hungary, May
  2003.
• 3 Shadbolt, N., A Matter of Trust, IEEE
  Intelligent Systems, pp. 2-3 January/February
  2002.
• 4 Golbeck, J., Hendler, J., Inferring
  Reputation on the Semantic Web, 13th
  International Web Conference (WWW2004), New York,
  NY, USA, May 2004.
• 5 Grandison, T., Sloman, M., A Survey of Trust
  in Internet Applications, IEEE Communications
  Surveys, 3, pp. 2-16, Fourth Quarter 2000.
• 6 D. McAllister, "Affect- and cognition-based
  trust as foundations for interpersonal
  cooperation in organizations," Academy of
  Management Journal, vol. 381, pp. 24-59, 1995.
• 7 Gollmann, D., Why Trust is Bad for
  Security, Keynote Speech, IEEE 6th International
  Workshop on Policies for Distributed Systems and
  Networks (POLICY 2005), Stockholm, Sweden, 6-8
  June, 2005.