ACM Student Chapter

1
ACM Student Chapter Union University
Haifei Li, Ph.D. Department of Mathematics and
Computer Science Union University Jackson, TN
2
Ties with ACM

• ACM student member when I was a graduate student
  at the University of Florida.
• ACM (regular) member when I am a professor.
• Faculty advisor for Nyack College student chapter
  of ACM.

3
Algorithms for Automated Negotiations and Their
Applications in Information Privacy
Haifei Li and David Ahn Nyack College Patrick C.
K. Hung Hong Kong University of Science and
Technology
4
Overview

• Introduction
• Related Work
• Algorithm to find Pareto Optimal Solutions
• Algorithm to Conduct Bilateral Negotiations
• Credit Card as Private Information
• Conclusion and Discussion

5
Introduction

• Negotiation is an active research topic. Hard to
  automate. Humans resist to automation effort.
• Negotiation is often viewed as more of an art,
  instead of a science.
• Automated algorithms are needed in order to
  implement negotiation systems. Negotiation for
  Information privacy is a new area.
• Strong assumptions may make the algorithms too
  idealistic to be useful.

6
Related Work

• Negotiation Support System
• Focus on support, not automate.
• Negotiation Agent
• No consensus on how agents negotiate, what to be
  negotiated.
• Game Theory
• rational assumption is too strong
• Machine Learning
• genetic algorithm
• bayesian probability
• Information Privacy
• Focus on enforcing companies Privacy Policy.
  Consumer is powerless, at the mercy of Big
  Brother, Big Buck.

7
Pareto Optimal Solution
Points on the Pareto Frontiers are Pareto optimal
solutions
Pareto Frontier
utilities for Agent Y
C
D
A
E
B
utilities for Agent X
8
Assumption for the Algorithm

• Negotiation attributes are pre-defined, and they
  are not dynamically added. The main task of
  bilateral negotiation is to assign values to
  negotiation attributes.
• Utilities (preferences) for some negotiation
  attribute values of both sides are not
  monotonically increasing or decreasing.
  Otherwise, the algorithm is trivial.
• Delivery schedule as an example
• It is possible that the enterprises preference
  over delivery is not as early as possible (for
  buyer), or as late as possible (for supplier).
• In JIT (Just-In-Time) manufacturing, the raw
  materials may need to arrive at the specified
  time.
• The result preference over the delivery schedule
  is NOT a monotonically increasing or decreasing
  function over the time.

9
Delivery Schedule as the Example
for buyer
Utility
The preference from the buyers side is 2, 6, 1,
5, 4, and 3. The preference from the sellers
side is 3, 4, 2, 1, 5, and 6.
for seller
Weeks (or days)
1
2
3
4
5
6
10
Algorithm to Conduct Bilateral Negotiation

• Only Pareto Optimal Solutions are left after the
  first algorithm has been applied.
• For each negotiator, each solution is assigned a
  counting number.
• Negotiators exchange proposals/counterproposals
  by consulting the pre-assigned counting number.

11
Proof of Guaranteed Termination
No-Increasing Curve for N2
P21
P12
Y Ordered Preference List of Alternatives
ya
Pa
No-Decreasing Curve for N1
X Number of Proposal Exchange
P11
P22
xa
12
Credit Card as Private Information
Service Requestor
Service Provider
Visa (1)
DC (1)

1. Visa 3
2. MC 2
3. AMEX 5
4. DC4

1. DC 2
2. MC 3
3. AMEX 4
4. Visa5

Visa (2)
DC (2)
Visa (3)
MC (1)
Accept MC
13
Conclusion

• Two algorithms have been presented.
• Application of these two algorithms in
  information privacy.
• Credit card example may not be very convincing.
• Algorithms are general enough to be used in other
  domains.