- 160 views
- 272 downloads
Flow as a Metacontrol for AI Agents
-
- Author / Creator
- Mariusdottir, Thorey M
-
Flow is a psychological state that provides a person with enjoyment when their skills and the challenges of their activity match. Thus flow encourages people to persist at and return to just manageable challenges; thereby fostering the growth of skills over time. It seems natural to assume that flow is an evolutionary adaption because there may be an advantage from using each individual's survival skills to the fullest in a hostile environment. This work introduces a method for metacontrol in artificial intelligence based on the concept flow. We propose that if an artificial intelligent agent is guided to move through tasks in a way that balances their skills with the complexity of the task this may improve its performance. To this end we define a measure of flow that allows agents to seek the state closest to flow. This measure is based on the reciprocal of the absolute difference between an agent's skills and the complexity of the task. The complexity is learned through observing, at each point, the minimum skills required to complete the task. We first illustrate our approach using a simple model of a multi-level environment and then compare our approach to a scripted and random metacontrols in the video game of Angband. The results indicate that a flow-seeking metacontrol can show improvement over a random metacontrol but the scripted metacontrol performs better overall.
-
- Subjects / Keywords
-
- Graduation date
- Fall 2015
-
- Type of Item
- Thesis
-
- Degree
- Master of Science
-
- License
- This thesis is made available by the University of Alberta Libraries with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.