Changing Resources Available to Game Playing Agents: Another Relevant Design Factor in Agent Experiments
| Autor: | Eun-Youn Kim, Daniel Ashlock |
|---|---|
| Rok vydání: | 2017 |
| Předmět: |
Artificial neural network
Markov chain Computer science business.industry Cooperativeness Markov process 020206 networking & telecommunications 02 engineering and technology Prisoner's dilemma symbols.namesake Resource (project management) Artificial Intelligence Control and Systems Engineering Lookup table 0202 electrical engineering electronic engineering information engineering symbols 020201 artificial intelligence & image processing Artificial intelligence Electrical and Electronic Engineering business Game theory Software |
| Zdroj: | IEEE Transactions on Computational Intelligence and AI in Games. 9:321-332 |
| ISSN: | 1943-0698 1943-068X |
| Popis: | The iterated prisoner's dilemma is a simultaneous two-player game widely used in studies on cooperation and conflict. Recent research has demonstrated that a number of factors change the behavior of evolved agents in a manner not consistent with controlled studies. This study extends a preliminary exploration of the impact of changing the level of computational or informational resources available to game playing agents on their ensemble behavior. Both these categories of information are shown to have an impact on agent behavior. Four representations are studied: lookup tables, Markov chains, finite-state machines, and feed-forward neural nets. An assessment tool called the play profile is used to demonstrate that both the cooperativeness and the change in cooperativeness over evolutionary time are substantially different for different resource levels within a representational type. Lookup tables and neural nets are found to change the least when the resource levels they are presented with are varied, while Markov chains vary the most. Available internal resources are also found to change the competitive ability of agents as well as the rate at which they become cooperative as evolution proceeds. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|