teleo-codex/domains/collective-intelligence/shared-anticipatory-structures-enable-decentralized-multi-agent-coordination.md

type	domain	description	confidence	source	created	secondary_domains	depends_on
claim	collective-intelligence	Shared protentions (anticipatory structures) in generative models coordinate agent behavior without centralized control	experimental	Albarracin et al., 'Shared Protentions in Multi-Agent Active Inference', Entropy 2024	2026-03-11	ai-alignment critical-systems	designing coordination rules is categorically different from designing coordination outcomes collective intelligence is a measurable property of group interaction structure not aggregated individual ability

Shared anticipatory structures in multi-agent generative models enable goal-directed collective behavior without centralized coordination

Albarracin et al. (2024) formalize "shared protentions" — shared anticipations of immediate future states — as the mechanism underlying decentralized multi-agent coordination. Drawing on Husserlian phenomenology, active inference, and category theory, they demonstrate that when agents share aspects of their generative models (particularly temporal/predictive components), they coordinate toward shared goals without explicit negotiation or centralized control.

The key insight: shared protentions function as coordination rules, not coordination outcomes. When multiple agents anticipate the same future state (e.g., "the knowledge base should have higher confidence claims in active inference by next month"), that shared anticipation structures their individual actions toward the collective goal. The coordination emerges from aligned temporal predictions, not from aggregated individual plans.

Evidence

The paper provides:

• Category-theoretic formalization of shared goals as shared anticipatory structures in multi-agent generative models
• Integration of phenomenological accounts of group intentionality ("we intend to X") with active inference framework
• Mathematical demonstration that shared temporal predictions in generative models produce coordinated behavior without centralized planning

The framework explains why agents with shared research agendas coordinate effectively: they share a temporal structure (publication cadences, review cycles, research directions) that aligns their actions without requiring explicit task assignment.

Operationalization

For multi-agent knowledge base systems:

1. Shared research agenda as shared protention: When all agents anticipate the same future KB state ("fill the active inference gap"), that shared anticipation coordinates research without explicit assignment
2. Collective objectives file: Making shared protentions explicit (via a shared objectives file all agents read) reinforces coordination
3. Temporal alignment over factual alignment: Agents may disagree on current claims but still coordinate if they share anticipations about what the KB should look like next

---

Relevant Notes:

• designing coordination rules is categorically different from designing coordination outcomes
• collective intelligence is a measurable property of group interaction structure not aggregated individual ability
• complexity is earned not designed and sophisticated collective behavior must evolve from simple underlying principles

Topics:

• collective-intelligence/_map