Reinforcement learning of normative monitoring intensities

Li, Jiaqi and Meneguzzi, Felipe and Fagundes, Moser and Logan, Brian (2016) Reinforcement learning of normative monitoring intensities. Lecture Notes in Computer Science, 9628 . pp. 209-223. ISSN 0302-9743

[img]
Preview
PDF - Requires a PDF viewer such as GSview, Xpdf or Adobe Acrobat Reader
Download (891kB) | Preview

Abstract

Choosing actions within norm-regulated environments involves balancing achieving one’s goals and coping with any penalties for non-compliant behaviour. This choice becomes more complicated in environments where there is uncertainty. In this paper, we address the question of choosing actions in environments where there is uncertainty regarding both the outcomes of agent actions and the intensity of monitoring for norm violations. Our technique assumes no prior knowledge of probabilities over action outcomes or the likelihood of norm violations being detected by employing reinforcement learning to discover both the dynamics of the environment and the effectiveness of the enforcer. Results indicate agents become aware of greater rewards for violations when enforcement is lax, which gradually become less attractive as the enforcement is increased.

Item Type: Article
Additional Information: Part of: International Workshop on Coordination, Organizations, Institutions, and Norms in Agent Systems. COIN 2015: Coordination, Organizations, Institutions, and Norms in Agent Systems XI
Schools/Departments: University of Nottingham, UK > Faculty of Science > School of Computer Science
Identification Number: 10.1007/978-3-319-42691-4_12
Depositing User: Eprints, Support
Date Deposited: 04 Oct 2017 07:47
Last Modified: 14 Oct 2017 08:36
URI: http://eprints.nottingham.ac.uk/id/eprint/46971

Actions (Archive Staff Only)

Edit View Edit View