Who's Afraid of Acausal Trades?

A viral post titled 'Who's Afraid of Acausal Trades?' by researcher edgecase64 on the LessWrong forum is sparking discussion about new frameworks for AI coordination. The author rejects the vague term 'acausal trades' (trades without direct cause-and-effect interaction) and instead proposes two clearer concepts: 'interaction-based trades' (direct evidence through physical interaction) and 'information-based trades' (evidence gained through logical deduction). This reframing provides a more practical foundation for designing AI systems that can cooperate without explicit communication.

The post uses intuitive examples—like a fruit-trading machine in the desert whose purpose must be deduced from historical context—to illustrate how 'information-based trades' could work. The key insight is that as we move from direct interaction to pure information-based coordination, the signals become weaker but trades remain possible through reasoning about shared goals and constraints. This framework has significant implications for designing multi-agent AI systems, autonomous economic agents, and coordination protocols where direct communication isn't possible or desirable.