A Little Bit Goes a Long Way: Modeling Universal Basic Income for Noncooperative Artificial Agents

With the rise of autonomous, artificially intelligent computer agents, it is broadly accepted that these agents must behave ethically. This requirement is often expressed as a contrast between beneficial, cooperative agent behaviors and detrimental, greedy ones. Many disciplinary approaches have shown that some level of cooperation is beneficial for survival and flourishing. However, it is unrealistic to enforce mass cooperation; some agents (be they real or artificial) will invariably act selfishly. Since mass cooperation is fundamentally impractical to enforce, we look for alternative means to achieve similar societal outcomes. We show that a moderate distribution of universal basic income (UBI) is a sufficient surrogate for global cooperation among artificial societies. In particular, we compare five different UBI schemes in the agent-based Sugarscape model and find that artificial societies with the best performing UBI schemes are 5× larger, engage in 6× as much trade, and have agents which live almost 2× as long as in societies without UBI. These results match or exceed those of societies that engage in utilitarian, cooperative behavior patterns.