Automated Negotiation Lecture 1: Introduction and Background Knowledge.

Automated Negotiation

Lecture 1: Introduction and Background Knowledge

Introduction: Motivation

Agents search & make contracts

-- Through peer-to-peer negotiation or a mediated marketplace.-- Agents can be real-world parties or software agents that work on behalf of real-world parties.

Increasingly important from a practical perspective

-- Developing communication infrastructure (Internet, WWW, NII, EDI, KQML, FIPA, Concordia, Voyager, Odyssey, Aglets, AgentTCL, Java Applets, ...)-- Electronic commerce on the Internet: Goods, services, information, bandwidth, computation, storage...-- Industrial trend toward virtual enterprises & outsourcing-- Automated negotiation allows dynamically formed alliances on a per order basis in order to capitalize on economies of scale, and allow the parties to stay separate when there are diseconomies of scale

Introduction: Motivation

Fertile, timely research area

-- Deep theories from game-theory & CS merge. Started together in the 1940’s [Morgenstern & von Neumann]. There were a few decades of little interplay. Upswing of interplay in the last few years.-- The intersection is a very fruitful, relatively open research area.

It is in this setting that the prescriptive power of game theory really comes into play.

-- Market rules need to be explicitly specified-- Software agents designed so as to act optimally-- Computational capabilities can be quantitatively characterized, and prescriptions can be made about how the agents should use their computation optimally

System with Self-Interested Agents

Includes computational or human agents Mechanism (e.g., rules of an auction) specifies

legal actions for each agent & how the outcome is determined as a function of the agents’ strategies

Strategy (e.g., bidding strategy), Agent’s mapping from known history to action

Rational self-interested agent chooses its strategy to maximize its own expected utility given the mechanism-- strategic analysis required for robustness -- noncooperative game theory

System with Self-Interested Agents

Computational Complexity

In executing the mechanism In determining the optimal strategyIn executing the optimal strategy

Has significant impact on prescriptions.

Has received little attention in game theory.

Ecommerce Process

1. Interest generation

2. Finding

3. Negotiating

4. Contract execution

5. After sales

MAS in Different EC Stages

1. Interest generation-- Funded adlets that coordinate-- Avatars for choosing which ads to read-- Customer models for choosing who to send ads and how much $ to offer

2. Finding-- Simple current systems: BargainFinder, Jango-- Meta-data, XML-- Standardized feature lists on goods to allow comparison-- How do these get (re)negotiated Different vendors prefer different feature lists Shopper agents need to understand the new lists How do algorithms cope with new features?-- Want to get a bundle: need to find many vendors

MAS in Different EC Stages

3. Negotiating-- Advantages of dynamic pricing: Right things sold to (and bought from) right parties at right time. So, world becomes a better place (social welfare increases)-- Further advantages from discriminatory pricing: Can increase social welfare.-- Fixed-menu take-it-or-leave-it offers -> negotiation Cost of generating & disseminating catalogs? Other customers see the price? Negotiation overhead? Personalized menus (check customer’s web page, links to & from it, what other similar customers did, customer profiles) Generating/printing the menu may be intractable, Negotiation will focus the generation, but vendor may bias prices & offerings based on path-- Preferences over bundles-- Coalition formation

MAS in Different EC Stages

4. Contract execution

-- Digital payment schemes-- Safe exchange

5. After sales

Outline

Utility----Quantification of Decision Result

Game Theory----Modelling of Decision

Quantification

Reason:Some concepts, like ‘Good’, ‘Bad’ is hard to comprehend by computer.

Method:Use real numbers (utility) to instead.

Decision Making

S: a set on environment statesD: a set of possible decisionsR: a set of achievable results

Result is influenced by both decision and environment state.

Decision Making

M: S x D ----> R

R = M (s, d), s∈S, d∈D

Decision Making

∵Environment state is usually uncertain.

∴For each s∈S there is a probability of occurrence of s.

∴With the mapping M this distribution for each d ∈ D induces a distribution on R.

∴So making the best decision mean choosing the "best" distribution on R among those available.

Decision Making

Example: A Picnic Decision

D={I, O}I: Picnic IndoorO: Picnic Outdoor

S={T, C}T: Thunderstorm Weather Forecast: P(T)=0.3C: Clear P(C)=0.7

R={A, B, G, E}A: AwfulB: BadG: GoodE: Excellent

Decision Making

Example: A Picnic Decision

Definition of M:M(T, O) = AM(T, I) = BM(C, I) = GM(C, O) = E

Decision Making

Example: A Picnic Decision

For Indoor: 30% Bad 70% Good

For Outdoor: 30% Awful 70% Excellent

Decision Making

Utility Function & Expected Utility

Utility Function: U(ri), ri∈Re.g.:U(A)=0, U(B)=2, U(G)=5, U(E)=10

Expected Utility: u(d), d∈Du(d) = ∑ P(ri) * U(ri)

Decision Making

Example:

u(I) = 0.3 * 2 + 0.7 * 5 = 4.1

u(O) = 0.7 * 10 = 7

Game Theory

Problem: In a game, players will get different outcomes by using different strategies. What strategies should they choose for improving their outcomes?

Game Theory

Matrix Form:

3, 3 5, 0

0, 5 1, 1

A B

A

B

Play1

Pla

y2

Play1’s Income Play2’s Income

Game Theory

Extensive Form:

Play 1

Play 2

A B

A B A B

3, 3 0, 5 5, 0 1, 1

Game Theory

Dominant Strategy:

In some games, a player can choose a strategy that "dominates" all other strategies in his strategy set: Regardless of what he expects his opponents to do, this strategy always yields a better payoff than any other of his strategies.

Game Theory

Dominant Strategy Equilibrium:

It is a strategy profile where each agent has picked its dominant strategy.

Game Theory

Nash Equilibrium:

No players can increase their utility by changing their strategies.

3, 3 5, 0

0, 5 1, 1

A B

A

B

Play1

Pla

y2

Play1’s Income Play2’s Income

Nash Equilibrium Point

Game Theory

Criticisms of Nash Equilibrium

-Not unique in all games.

-Does not exist in all games.

-May be hard to compute.

Game Theory

Existence of Nash Equilibrium

Any finite game where each action node is alone in its information set, i.e. at every point in the game, the agent whose turn it is to move knows what moves have been played so far. And the game is dominance solvable by backward induction.