| Abstract |
The influence maximisation problem has been extensively studied in theoretical models, where macroscopic behaviour in complex networks are approximated using mathematical models. However, the unpredictability and irrationality of human behaviour and decision-making in many cases renders theoretically derived optimal strategies unreliable and often computationally expensive in practical settings. To address this, we design an experiment in which we simulate a real-world influence maximisation setting (such as the Belt and Road initiative adopted by China), where players compete to maximise and maintain influence in networks to achieve personal goals. In this game, human players compete against an AI or other human players to maximise their influence spread within arbitrary synthetic networks, under budget and time constraints. Data collected through the experiment will be used to determine any discernible patterns in strategies typically adopted by participants that can aid the design of predictive models, to be used to anticipate adversarial strategies in real-time, mitigating risks that arise from uncertainty and incomplete information of adversarial tactics and finally present effective counter-strategies that outperform them. |