The Complexity of Cooperation: Agent-Based Models of Competition and Collaboration

The Complexity of Cooperation: Agent-Based Models of Competition and Collaboration

The Complexity of Cooperation: Agent-Based Models of Competition and Collaboration

The Complexity of Cooperation: Agent-Based Models of Competition and Collaboration

Synopsis

Robert Axelrod is widely known for his groundbreaking work in game theory and complexity theory. He is a leader in applying computer modeling to social science problems. His book The Evolution of Cooperation has been hailed as a seminal contribution and has been translated into eight languages since its initial publication. The Complexity of Cooperation is a sequel to that landmark book. It collects seven essays, originally published in a broad range of journals, and adds an extensive new introduction to the collection, along with new prefaces to each essay and a useful new appendix of additional resources. Written in Axelrod's acclaimed, accessible style, this collection serves as an introductory text on complexity theory and computer modeling in the social sciences and as an overview of the current state of the art in the field.

The articles move beyond the basic paradigm of the Prisoner's Dilemma to study a rich set of issues, including how to cope with errors in perception or implementation, how norms emerge, and how new political actors and regions of shared culture can develop. They use the shared methodology of agent-based mo

Excerpt

This chapter began with a hammer and a nail. the nail was a problem I wanted to solve. the hammer was a tool I wanted to try out that looked well suited to driving my nail. the problematic nail was the question of whether the success of the tit for tat strategy in my computer tournaments depended in large part on the prior beliefs of the people who submitted strategies about what the other submissions would be like. in other words, would the tournament results be influenced by what people believed others would be doing, or would something like the reciprocity of tit for tat succeed in a tournament setting without any preconceptions about the tendencies or even the responsiveness of others?

Answering this question would require a method of generating new strategies that would not involve human preconceptions. a tool for doing precisely this was developed by John Holland, a computer scientist at Michigan. I knew John well from a small research group we were both part of for years. This was the bach group, named after its original members: Arthur Burks, myself, Michael Cohen, and John Holland. John's genetic algorithm technique (Holland 1975) was inspired by the ability of evolution to discover adaptive solutions to hard problems. By the mid-1980's, the genetic algorithm had proven to be an effective search technique for discovering effective solutions in highly complex computer problems (Goldberg 1989; Mitchell 1996).

My own interest in evolutionary simulation dates back to 1960, when I did a high school science project on computer simulation of hypothetical life forms and environments. At the University of Chicago I was a math major, but also spent a summer with the Committee on Mathematical Biology reading about evolutionary biology. Despite these interests in computers, mathematics, and evolution, I chose to go to graduate school in political science. the problems I most wanted to work on dealt with the prevention of conflict between nations, especially nuclear war. After getting my Ph.D. from Yale, I taught international politics at Berkeley and then moved to the University of Michigan, where I took a joint appointment in the Department of Political Science and what is now the School of Public Policy.

At Michigan, Michael Cohen became my closest colleague. He kept suggesting that I meet John Holland and learn about his work. Eventually I succumbed—to my great joy and benefit. Among other things, I

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.