Anastassacos, Nicolas;
(2022)
Cooperation and Social Dilemmas with Reinforcement Learning.
Doctoral thesis (Ph.D), UCL (University College London).
Preview |
Text
Cooperation_and_Social_Dilemmas_with_Reinforcement_Learning-1.pdf - Accepted Version Download (12MB) | Preview |
Abstract
Cooperation between humans has been foundational for the development of civilisation and yet there are many questions about how it emerges from social interactions. As artificial agents begin to play a more significant role in our lives and are introduced into our societies, it is apparent that understanding the mechanisms of cooperation is important also for the design of next-generation multi-agent AI systems. Indeed, this is particularly important in the case of supporting cooperation between self-interested AI agents. In this thesis, we focus on the analysis of the application of mechanisms that are at the basis of human cooperation to the training of reinforcement learning agents. Human behaviour is a product of cultural norms, emotions and intuition amongst other things: we argue it is possible to use similar mechanisms to deal with the complexities of multi-agent cooperation. We outline the problem of cooperation in mixed-motive games, also known as social dilemmas, and we focus on the mechanisms of reputation dynamics and partner selection, two mechanisms that have been strongly linked to indirect reciprocity in Evolutionary Game Theory. A key point that we want to emphasise is the fact we assume no prior knowledge and explicit definition of strategies, which instead are fully learnt by the agents during the games. In our experimental evaluation, we demonstrate the benefits of applying these mechanisms to the training process of the agents, and we compare our findings with results presented in a variety of other disciplines, including Economics and Evolutionary Biology.
| Type: | Thesis (Doctoral) |
|---|---|
| Qualification: | Ph.D |
| Title: | Cooperation and Social Dilemmas with Reinforcement Learning |
| Open access status: | An open access version is available from UCL Discovery |
| Language: | English |
| Additional information: | Copyright © The Author 2022. Original content in this thesis is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0) Licence (https://creativecommons.org/licenses/by-nc/4.0/). Any third-party copyright material present remains the property of its respective owner(s) and is licensed under its existing terms. Access may initially be restricted at the author’s request. |
| UCL classification: | UCL UCL > Provost and Vice Provost Offices > UCL BEAMS UCL > Provost and Vice Provost Offices > UCL BEAMS > Faculty of Engineering Science UCL > Provost and Vice Provost Offices > UCL BEAMS > Faculty of Engineering Science > Dept of Computer Science |
| URI: | https://discovery.ucl.ac.uk/id/eprint/10161343 |
Archive Staff Only
 |
View Item |
