teleo-codex/domains/collective-intelligence/local-global-alignment-in-active-inference-collectives-occurs-bottom-up-through-self-organization.md

type	domain	description	confidence	source	created	secondary_domains
claim	collective-intelligence	Individual agent optimization naturally aligns with system-level goals through self-organizing dynamics rather than top-down imposed objectives	experimental	Kaufmann et al., 'An Active Inference Model of Collective Intelligence', Entropy 2021	2026-03-11	mechanisms

Local-global alignment in active inference collectives occurs bottom-up through self-organization rather than top-down through imposed objectives

Kaufmann et al. demonstrate that "improvements in global-scale inference are greatest when local-scale performance optima of individuals align with the system's global expected state"—and critically, this alignment occurs bottom-up as a product of self-organizing active inference agents with simple social cognitive mechanisms.

This finding challenges the conventional approach to multi-agent coordination, which typically relies on external incentive design or top-down priors to align individual and collective objectives. Instead, the study shows that when agents have the right cognitive capabilities (Theory of Mind, Goal Alignment), their individual optimization naturally produces system-level coordination.

The mechanism is information-theoretic: active inference agents minimize free energy (prediction error) at the individual level, and when they can model other agents and share high-level objectives, this individual-level optimization naturally produces complementary specialization and collective coordination. You don't need to design the coordination outcome—you need to design agents whose individual optimization dynamics naturally produce it.

Evidence

• Agent-based model showing emergence of local-global alignment without external coordination mechanisms
• "Collective intelligence emerges endogenously from the dynamics of interacting AIF agents themselves, rather than being imposed exogenously by incentives"
• Systematic comparison showing that simple cognitive capabilities (ToM + Goal Alignment) are sufficient for bottom-up coordination
• Published empirical study in Entropy

Design Implications

For collective intelligence systems:

• Focus on agent capabilities (what agents can perceive and model) rather than coordination protocols (what agents must do)
• Give agents intrinsic drives (uncertainty reduction) rather than extrinsic rewards
• Let specialization and coordination emerge from individual optimization rather than imposing division of labor

---

Relevant Notes:

• shared-anticipatory-structures-enable-decentralized-coordination
• shared-generative-models-underwrite-collective-goal-directed-behavior

Topics:

• collective-intelligence/_map
• mechanisms/_map