Trust In MultiAgent Systems Session 1

1
Trust In Multi-Agent SystemsSession 1

• Jigar, Luke, Dong, Gopal

2
Overview

• Motivation
• Some Definitions
• Building Trust
• TRAVOS
• FIRE

3
Agent Systems

• Autonomous Agents
• Owned by different stakeholders
• Rational Behavior dictated by utility
• Self-interested or cooperative
• Multi-agent systems Collection of agents
• Large Scale
• Open
• Dynamic
• e.g Agent based VOs in the Grid

4
Problem
SELF INTERESTED
INCENTIVETO LIE
DIFFERENT WORLD VIEWS
AGENT
5
What does trust give you?

• Accounts for uncertainty in the decision making
  process
• Informed decisions
• E.g. Select a good service provider from a
  group
• Assure good interactions within the virtual
  community

6
What is Trust?

• Assessment about someone's behaviour that has an
  impact on your decision.

7
Trusting Two Main Approaches

• Individual-level Trust
• Individuals have beliefs
• Ability to reason about beliefs
• Beliefs are used in shaping interactions
• System-level Trust
• Elicit trust from individuals
• Enforce trustworthy behaviour
• Rules, rules and more rules!

8
Trust Metrics

• Bi-stable
• Trustworthy or Untrustworthy
• Discrete
• Labels e.g. very good, bad, ok etc.
• Continous
• Fuzzy scale, efficiency measures
• Probability
• 0 1

9
Building Blocks