Abstract
In file sharing networks, users can either act for personal gain by
downloading files, or help the network by uploading files. Similar
scenarios are important in many diverse situations where it does
not pay to be nice to others. As a result, self-interested agents
shirk collective behaviour, leading to poor outcomes for everyone.
In response, in this paper we introduce a new metric for social
dilemmas that quantifies the discrepancy between what is rational
for individual agents and what is rational for the group, which we
call the exchange threshold. It is the smallest fraction of individual
rewards that need to be shared to make the unique Nash equilibrium
a social welfare optimum. The exchange threshold differs to notions
of altruism or prosocial behaviours because agents are transferring
their rewards in order to induce others to care about their welfare.
We investigate how the exchange threshold of a strategic game
representing a tragedy of the commons varies with the abundance
of resources to provide a deeper understanding of the underlying
incentives.
downloading files, or help the network by uploading files. Similar
scenarios are important in many diverse situations where it does
not pay to be nice to others. As a result, self-interested agents
shirk collective behaviour, leading to poor outcomes for everyone.
In response, in this paper we introduce a new metric for social
dilemmas that quantifies the discrepancy between what is rational
for individual agents and what is rational for the group, which we
call the exchange threshold. It is the smallest fraction of individual
rewards that need to be shared to make the unique Nash equilibrium
a social welfare optimum. The exchange threshold differs to notions
of altruism or prosocial behaviours because agents are transferring
their rewards in order to induce others to care about their welfare.
We investigate how the exchange threshold of a strategic game
representing a tragedy of the commons varies with the abundance
of resources to provide a deeper understanding of the underlying
incentives.
| Original language | English |
|---|---|
| Number of pages | 8 |
| Publication status | Accepted/In press - 28 Mar 2023 |
| Event | Adaptive and Learning Agents Workshop - London, United Kingdom Duration: 29 May 2023 → 30 May 2023 https://alaworkshop2023.github.io/ |
Workshop
| Workshop | Adaptive and Learning Agents Workshop |
|---|---|
| Country/Territory | United Kingdom |
| City | London |
| Period | 29/05/2023 → 30/05/2023 |
| Internet address |
Keywords
- Game Theory
- Social Dilemma
- Multiagent Reinforcement Learning