Fundamentals of Agents and Multi-Agents

 

I would strongly recommend to everyone that is interested in agents, multi-agents, artificial life and Complexity Science in general; to learn some of the rigorous, normative and practical theories that govern agent systems.

 

This is my story.  Despite having a strong interest in this subject, and despite having done research for 6 years in the mid-80s in Nonlinear Dynamics (another branch of Complexity Science), I was unable to fully appreciate agents for almost 10 years, until I took a tutorial at the Autonomous Agents conference in St. Paul in 1998 from Tuomas Sandholm.  That made a huge difference.  I was finally able to see all the "undocumented" requirements to start to comprehend agents, multi-agents and their ongoing research.  For example, Kauffman's book, roughly presents the results that he found in the course of his research about adaptable systems, their coevolution, and his theory about the origin of life that involves self-replication, cooperative games and aggregation formation (coalitions).

 

However, to comprehend his work in detail one

needs to learn many of the fundamentals:

 

 

Multi-Agents

*  - agent taxonomies: mobile, autonomous, self-interested, multi-, software-, distributed, etc. + agents

*  - self-interest utility functions

*  - rationality full/bounded etc.

*  - computational efficiency

*  - distribution of computation

*  - social welfare

*  - Pareto efficiency

*  - individual rationality

*  - stability (in terms of utility functions)

*  - symmetry

 

Game Theory

*  - Strategies

*  - Dominant Strategy Equilibrium

*  - Nash equilibrium

*  - Existence theorems

*  - Revelation principle

 

Voting

*  - Methods of social choice

*  - Agenda paradox

*  - Pareto domination

*  - Inverted-order paradox

*  - Majority winner paradox

*  - Truthful voting

*  - Strategic (unsincere) voting

*  - Truth extraction (Clarke tax)

 

Auctions

*  - Auction settings

*  - Auctions protocols: All-pay, English, First-price sealed bid, Dutch, Vickery,

*  - Results for private and non-private auctions

*  - Vulnerabilities: shills, lying auctioneer,

 

General Equilibrium

*  - Walrasian

*  - Market setting

*  - Definition

*  - Properties

*  - Limitation

*  - Existence

*  - Uniqueness

*  etc.

 

Coalition formation

*  - Classic Game Theory

*  - Game classification

*  - Coalition structure stability

*  - Transfer schemes

*  - Core and Walrasian equilibriums for coalitions

*  - Shapley's value

*  - Cryptographic algorithm

*  - Social Welfare

*  - Optimal coalition structure

*  - Stability

*  - Non-emptiness of the BR core

*  etc.

 

 

Contract Nets:

*  - Contract net protocol

*  - TRACONET

*  - Bidding and Awarding

*  - Clustering

*  - Swaps

*  - Multiagent Contracts

*  - Results

*  - Leveled commitment contracts

 

In addition, to do practical work i.e. like simulations, one must also understand ACLs (agents communications languages), agent architectures, agent-related technologies, agent tools, and some of the practical applications of agents.  There is no doubt - it is a fascinating subject, but it is easy to miss the fact that is necessary to learn some of these mostly "undocumented" requirements,