Dynamics Underneath Symbols

A Case Study in Autonomous Agents

p. 187-200

Abstract

Our cognition is structuring the informational layer, consisting of perception, anticipation, and action, and it should also be sustained on a physical basis. In this paper, we aim to explore the relationship between the informational layer and the physical layer from a dynamical systems point of view. As an example, the fluctuation of choice is investigated by using a simulated agent. By setting a T-maze, the agent should choose one arm of the maze if a corresponding token is presented. We prepared two types of tokens, corresponding to the left and right arm of the maze. After training the network of the agent to successfully choose the corresponding arm, we presented two tokens simultaneously to the agent and observed its behavior. As a result, we found several behaviours, which are difficult to speculate on from a case in which only a single token is presented to the agent. Detailed analyses and the implications of the model are discussed.

Text

Download Facsimile [PDF, 6.5M]

References

Bibliographical reference

Kohei Nakajima, « Dynamics Underneath Symbols », CASYS, 28 | 2014, 187-200.

Electronic reference

Kohei Nakajima, « Dynamics Underneath Symbols », CASYS [Online], 28 | 2014, Online since 10 October 2024, connection on 14 November 2024. URL : http://popups.lib.uliege.be/1373-5411/index.php?id=4410

Author

Kohei Nakajima

Artificial Intelligence Laboratory, Department of Informatics, University of Zurich

By this author

Copyright

CC BY-SA 4.0 Deed