A Trust based Recommendation System on a Social Network

1
A Trust based Recommendation System on a Social
Network

• Debdoot Mukherjee
• 2006MCS2261

2
What is it ?

• An architecture in a social network wherein an
  agent in need of a service can make use of the
  information in the knowledge base of his contacts
  to find a reliable service provider offering that
  service.
• Use of trust to filter information

3
Trust

• Trust (or, symmetrically, distrust) is a
  particular level of the subjective probability
  with which an agent assesses that another agent
  or a group will perform a particular action, both
  before he can monitor such action (or
  independently of his capacity of ever to be able
  to monitor it) and in a context in which it
  affects his own action

4
Areas of study

• Social Psychology
• Philosophy
• Economics
• Marketing
• Multi Agent Systems
• Distributed AI
• E-Commerce
• Anti-phishing
• Semantic Web
• Social Networks
• Intrusion Detection in Networks
• Wikipedia
• PKI
• Recommendation Systems

5
Types of trust

• Basic Trust
• General Trust
• Situational Trust

6
Properties

• Context-dependent
• Personal
• Asymmetry
• Subjective
• Positive or Negative
• Transitivity
• Dynamic
• Composability
• Interesting when risk involved

7
Scale
8
Computation of Trust
9
Recommendation systems

• Suggest items that users may find interesting
• Information overload problem
• Search Engines lack personalization
• Content based
• Collaborative filtering
• Build Social Network

10
Challenges

• Network Density
• Preference Heterogeneity
• Sparseness of Knowledge

11
Motivation

• Dont know where to shop?
• Ask a friend
• Who is the best provider of the service X in the
  locality Y?
• Its a small world

12
Model Description

• Agents
• Items
• Service providers
• Knowledge Base
• Graph of Social Network

13
Trust Issues

• Which scale to use?
• Which model to use?
• How to rank the responses?

14
The Search begins

• Flooding
• Complete Search
• Optimum Result
• Too expensive

15
The way out.

• Constrain search space
• Query only on trusted paths
• How far from optimum??

16
Thats not all

• Number of items too large.
• Classify items
• Create an ontology graph

17
A fringe problem

• If (itemi, service-providerk) and (itemj,
  service-providerk) are found to have high trust
  values ? itemi and itemj are related
• For many values of k ? Be Sure

18
How to search then??

• Query returns all items that belong to the
  category
• Trust for (itemi, service-providerk) computed
  from known trust value for (itemj,
  service-providerk) and the edge label for (itemj,
  itemi) in the ontology graph

19
Here comes Trudy

• Malicious agents
• Selfish agents
• Random agents
• How do they affect performance?
• Threshold ? Control Variable

20
Swiss Guys are at it

• A probabilistic model
• Concentration on RS benchmarks
• Flooding algorithm not scalable
• Not robust
• Too simplistic

21
Their Claims..

• A trust based approach works better than a
  frequency based one Self organizes to an optimum
• Self organizes to an optimum
• Long trust paths develop allowing an agent to
  rely on recommendations of another agent of
  similar preferences but far away.

22
References

• Walter Frank E., Battiston Stefano, Schweitzer
  Frank, A Model of a Trust-based Recommendation
  System on a Social Network, Nov. 2006.
• Marsh, S. Formalising Trust as a Computational
  Concept. PhD thesis, University of Stirling,
  1994.
• Guha, R., R. Kumar, P. Raghavan, And A. Tomkins.
  2003. Propagation of Trust and Distrust.
  Proceedings of the 13th Annual International
  World Wide WebConference, New York, NY.
• Abdul-Rahman Alfarez, Hailes Stephen, Supporting
  trust in virtual communities, Proceedings of the
  33rd Hawaii International Conference on System
  Sciences 2000
• Gambetta D. Can We Trust Trust?. In, Trust
  Making and Breaking Cooperative Relations, Basil
  Blackwell. Oxford, 1990
• Massa Paolo and Bhattacharjee Bobby, Using Trust
  in Recommender Systems An Experimental Analysis,
  In Proceedings of the Second International
  Conference on Trust Management (ITRUST 2004)
• Y.-H. Tan and W. Thoen, Towards a Generic Model
  of Trust for Electronic Commerce. In Proceedings,
  12th International Bled Electronic Commerce
  Conference, Bled Slovenia, 1999.

23
Thank You !!