Conjecture: Coordinating AI Agents May Be Easier Than Coordinating Humans

Determine whether coordinating AI agents via formal programmatic mechanisms and digital asset–based value assignment is easier than coordinating human actors, and characterize the conditions under which such an advantage holds.

Background

To enable mission-oriented economies, the paper explores whether AI agents can be more readily steered than humans through programmatic rules and market-based incentives embedded in digital infrastructure. This would leverage agents’ predictability, steerability, and formal verification.

The authors explicitly frame this as a conjecture, suggesting that AI coordination may be simpler than human coordination due to differences in motivations and agent controllability, and invite validation of this claim.

References

Perhaps in some ways it may be easier to bring about the coordination of AI agents through the combination of 1) formal, programmatical mechanisms, and 2) via the value assignment mechanisms embedded in digital assets, than it would be to coordinate human actors. We may assign some credence to this conjecture from the perspective of predictability and steerability of AI agents when contrasted with the complex motivations behind human actions.

— Virtual Agent Economies (2509.10147 - Tomasev et al., 12 Sep 2025) in Section Mission, Subsection Opportunities