Can agents acquire human-like behaviors in a sequential bargaining game? - Comparison of roth's and Q-learning agents

Keiki Takadama, Tetsuro Kawai, Yusuke Koyama

研究成果: Conference contribution

3 引用 (Scopus)

抄録

This paper addresses agent modeling in multiagent-based simulation (MABS) to explore agents who can reproduce human-like behaviors in the sequential bargaining game, which is more difficult to be reproduced than in the ultimate game (i.e., one time bargaining game). For this purpose, we focus on the Roth's learning agents who can reproduce human-like behaviors in several simple examples including the ultimate game, and compare simulation results of Roth's learning agents and Q-learning agents in the sequential bargaining game. Intensive simulations have revealed the following implications: (1) Roth's basic and three parameter reinforcement learning agents with any type of three action selections (i.e., e-greed, roulette, and Boltzmann distribution selections) can neither learn consistent behaviors nor acquire sequential negotiation in sequential bargaining game; and (2) Q-learning agents with any type of three action selections, on the other hand, can learn consistent behaviors and acquire sequential negotiation in the same game. However, Q-learning agents cannot reproduce the decreasing trend found in subject experiments.

元の言語English
ホスト出版物のタイトルLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
ページ156-171
ページ数16
4442 LNAI
出版物ステータスPublished - 2007
外部発表Yes
イベント7th International Workshop on Multi-Agent-Based Simulation, MABS 2006 - Hakodate, Japan
継続期間: 2007 5 82007 5 8

出版物シリーズ

名前Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
4442 LNAI
ISSN(印刷物)03029743
ISSN(電子版)16113349

Other

Other7th International Workshop on Multi-Agent-Based Simulation, MABS 2006
Japan
Hakodate
期間07/5/807/5/8

Fingerprint

Reinforcement learning

ASJC Scopus subject areas

  • Computer Science(all)
  • Biochemistry, Genetics and Molecular Biology(all)
  • Theoretical Computer Science

これを引用

Takadama, K., Kawai, T., & Koyama, Y. (2007). Can agents acquire human-like behaviors in a sequential bargaining game? - Comparison of roth's and Q-learning agents. : Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (巻 4442 LNAI, pp. 156-171). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); 巻数 4442 LNAI).

Can agents acquire human-like behaviors in a sequential bargaining game? - Comparison of roth's and Q-learning agents. / Takadama, Keiki; Kawai, Tetsuro; Koyama, Yusuke.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). 巻 4442 LNAI 2007. p. 156-171 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); 巻 4442 LNAI).

研究成果: Conference contribution

Takadama, K, Kawai, T & Koyama, Y 2007, Can agents acquire human-like behaviors in a sequential bargaining game? - Comparison of roth's and Q-learning agents. : Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). 巻. 4442 LNAI, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 巻. 4442 LNAI, pp. 156-171, 7th International Workshop on Multi-Agent-Based Simulation, MABS 2006, Hakodate, Japan, 07/5/8.
Takadama K, Kawai T, Koyama Y. Can agents acquire human-like behaviors in a sequential bargaining game? - Comparison of roth's and Q-learning agents. : Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). 巻 4442 LNAI. 2007. p. 156-171. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
Takadama, Keiki ; Kawai, Tetsuro ; Koyama, Yusuke. / Can agents acquire human-like behaviors in a sequential bargaining game? - Comparison of roth's and Q-learning agents. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). 巻 4442 LNAI 2007. pp. 156-171 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{e062a97f4c064bb59a443d11c5d3587d,
title = "Can agents acquire human-like behaviors in a sequential bargaining game? - Comparison of roth's and Q-learning agents",
abstract = "This paper addresses agent modeling in multiagent-based simulation (MABS) to explore agents who can reproduce human-like behaviors in the sequential bargaining game, which is more difficult to be reproduced than in the ultimate game (i.e., one time bargaining game). For this purpose, we focus on the Roth's learning agents who can reproduce human-like behaviors in several simple examples including the ultimate game, and compare simulation results of Roth's learning agents and Q-learning agents in the sequential bargaining game. Intensive simulations have revealed the following implications: (1) Roth's basic and three parameter reinforcement learning agents with any type of three action selections (i.e., e-greed, roulette, and Boltzmann distribution selections) can neither learn consistent behaviors nor acquire sequential negotiation in sequential bargaining game; and (2) Q-learning agents with any type of three action selections, on the other hand, can learn consistent behaviors and acquire sequential negotiation in the same game. However, Q-learning agents cannot reproduce the decreasing trend found in subject experiments.",
keywords = "Agent modeling, Agent-based simulation, Human-like behaviors, Reinforcement learning, Sequential bargaining game",
author = "Keiki Takadama and Tetsuro Kawai and Yusuke Koyama",
year = "2007",
language = "English",
isbn = "9783540765363",
volume = "4442 LNAI",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
pages = "156--171",
booktitle = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

}

TY - GEN

T1 - Can agents acquire human-like behaviors in a sequential bargaining game? - Comparison of roth's and Q-learning agents

AU - Takadama, Keiki

AU - Kawai, Tetsuro

AU - Koyama, Yusuke

PY - 2007

Y1 - 2007

N2 - This paper addresses agent modeling in multiagent-based simulation (MABS) to explore agents who can reproduce human-like behaviors in the sequential bargaining game, which is more difficult to be reproduced than in the ultimate game (i.e., one time bargaining game). For this purpose, we focus on the Roth's learning agents who can reproduce human-like behaviors in several simple examples including the ultimate game, and compare simulation results of Roth's learning agents and Q-learning agents in the sequential bargaining game. Intensive simulations have revealed the following implications: (1) Roth's basic and three parameter reinforcement learning agents with any type of three action selections (i.e., e-greed, roulette, and Boltzmann distribution selections) can neither learn consistent behaviors nor acquire sequential negotiation in sequential bargaining game; and (2) Q-learning agents with any type of three action selections, on the other hand, can learn consistent behaviors and acquire sequential negotiation in the same game. However, Q-learning agents cannot reproduce the decreasing trend found in subject experiments.

AB - This paper addresses agent modeling in multiagent-based simulation (MABS) to explore agents who can reproduce human-like behaviors in the sequential bargaining game, which is more difficult to be reproduced than in the ultimate game (i.e., one time bargaining game). For this purpose, we focus on the Roth's learning agents who can reproduce human-like behaviors in several simple examples including the ultimate game, and compare simulation results of Roth's learning agents and Q-learning agents in the sequential bargaining game. Intensive simulations have revealed the following implications: (1) Roth's basic and three parameter reinforcement learning agents with any type of three action selections (i.e., e-greed, roulette, and Boltzmann distribution selections) can neither learn consistent behaviors nor acquire sequential negotiation in sequential bargaining game; and (2) Q-learning agents with any type of three action selections, on the other hand, can learn consistent behaviors and acquire sequential negotiation in the same game. However, Q-learning agents cannot reproduce the decreasing trend found in subject experiments.

KW - Agent modeling

KW - Agent-based simulation

KW - Human-like behaviors

KW - Reinforcement learning

KW - Sequential bargaining game

UR - http://www.scopus.com/inward/record.url?scp=37349063167&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=37349063167&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:37349063167

SN - 9783540765363

VL - 4442 LNAI

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 156

EP - 171

BT - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

ER -