Report on the meeting of the WG5 on Trust

Budapest, 16 May 2009

1. Objectives

The objectives of the meeting were the following:

  1. Promote the networking among the participants
  2. Start the definition of common activities
  3. Plan for the three years of the action

We made progress on the three objectives and the plan was reduced to the activities to do during the meeting in Cyprus in December.

2. Program

9:00-11:00

  • Introduction to Action and Trust WG Carles Sierra, WG chair. (Download presentation)
  • Maintenance-based Trust for Multi-Agent Systems Babak Khosravifar, Concordia University, Canada
  • Operators for Propagating Trust and their Evaluation in Social Networks Chung-Wei Hang,North Carolina State University, US
  • Pragmatic-Strategic Reputation-Based Decisions in BDI agents Isaac Pinyol, IIIA-CSIC, Spain
  • Smart Cheaters Do Prosper: Defeating Trust and Reputation Systems Reid Kerr University of Waterloo, Canada

11:30-13:30

Joint Session: Trust & Semantics

  • Advanced Policy Specification and Enforcement applied to Context-Aware Prevention in Independent Living Alessandra Mileo, Lab. NOMADIS, Italy
  • Comparing Trust Mechanisms for Monitoring Aggregator Nodes Oly Mistry, Univ. of Tulsa, US
  • Panel discussion Carles Sierra, Axel Polleres, Jordi Sabater, Andreas Herzig

3. Highlights of the presentations

Babak talked about maintenance-based trust. Introduced a model with Different roles (referee, trustworthy), direct and indirect interactions. Trust model was based on error minimisation. He showed experimental results on service provider analysis.

Chung Talked on selection of services. Concretely, on the trust on service oriented environments and on the trust on composition of services. The model includes a discount factor and is based on the preferences of the agents.

Isaac talked about a BDI architecture for the modeling of preferences where no explicit representation of trust is made.

Reid talked about attacks to trust and reputation. He argued that the effects of trust on the agent’s behaviour are important and discussed on the vulnerability of the trust values by certain types of actions.

Martin Warned about the importance of assumptions on the structure of the environment. Identity has to be guaranteed as well as certain similarity between transactions.

Àngels tried to convince everybody that Diplomacy is the best game for trust and in fact for agreement technologies.

Oly proposed a method for propagating values in sensor networks where trust is associated to the notion of reliability of sensors.

Alessandra presented her way of specifying policies using preferences. Healthcare policies that take into account the context and patient answers.

Axel puts challenges from the semantics perspective. Difficult ones. In particular, the notions of Trust on ontologies or on data.

The presentation motivated questions and debate that ended in a number of points outlined next.

4. Problems found and ideas proposed.

The connection between agent goals and trust model has to be made more clear than in current models.

It is necessary to give semantics to preferences of agents in relation to trust models.

We need to integrate the agent cognitive model with its trust model.

Emotions might be good to avoid predictability.

Vulnerability analysis is important. If there is a slim incentive to violate an agreement this will happen.

Assumptions are unclear in several models. They should be made explicit.

We should aim at processing of large quantities of data and messages to better assess trust models.

The trust on a group seems essential for complex problem solving.

Trust on data and trust on groups are unsolved challenges.

5. Actions during the meeting at EUMAS 2009

a) Keep the WG growing

b) Organise two sessions on:

1) Trust on groups

2) Trust on data (ontology)

SetPageWidth