265b976267
- Source: inbox/archive/2024-04-00-albarracin-shared-protentions-multi-agent-active-inference.md - Domain: collective-intelligence - Extracted by: headless extraction cron (worker 3) Pentagon-Agent: Leo <HEADLESS>
46 lines
3.3 KiB
Markdown
46 lines
3.3 KiB
Markdown
---
|
|
type: claim
|
|
domain: collective-intelligence
|
|
description: "We-intentions arise when agents share the anticipatory components of their world models, not from aggregating individual intentions"
|
|
confidence: experimental
|
|
source: "Albarracin et al., 'Shared Protentions in Multi-Agent Active Inference', Entropy 2024"
|
|
created: 2026-03-11
|
|
secondary_domains: [ai-alignment]
|
|
depends_on:
|
|
- "collective intelligence is a measurable property of group interaction structure not aggregated individual ability"
|
|
---
|
|
|
|
# Group intentionality — the 'we intend to X' that exceeds individual intentions — emerges from shared anticipatory structures within agents' generative models rather than from aggregating individual goals
|
|
|
|
Albarracin et al. (2024) provide a formal account of group intentionality using active inference and category theory. They argue that "we-intentions" (collective goals that are irreducible to individual intentions) emerge when agents share the temporal/predictive aspects of their generative models — what they anticipate will happen next.
|
|
|
|
This contrasts with traditional accounts that treat group intentions as either:
|
|
1. Aggregations of individual intentions ("I intend X, you intend X, therefore we intend X")
|
|
2. Emergent properties requiring special explanation
|
|
|
|
Instead, shared protentions (shared anticipations) are a structural property of multi-agent systems. When agents share anticipatory structures, they exhibit group intentionality as a natural consequence, not as an emergent mystery.
|
|
|
|
## Evidence
|
|
|
|
The paper uses category theory to formalize how shared generative models create group-level intentional states. Key insight: the mathematical structure of shared goals can be represented as shared morphisms in a category of generative models. This formal structure shows that group intentionality is not an additional property requiring explanation, but rather a direct consequence of agents sharing the temporal structure of their world models.
|
|
|
|
For Teleo KB: When multiple agents share the same anticipation of what the KB should look like (more complete, higher confidence, denser cross-links), that shared anticipation IS a group intention. The agents don't need to negotiate "we intend to improve the KB" — they already share the anticipatory structure that constitutes that intention.
|
|
|
|
## Phenomenological Grounding
|
|
|
|
The paper grounds this in Husserl's phenomenology: protention is the pre-reflective anticipation of the immediate future that structures conscious experience. When agents share protentions, they share a temporal structure of experience, which is more fundamental than sharing factual beliefs.
|
|
|
|
For KB agents: shared temporal structure (anticipation of publication cadence, review cycles, research directions) may be more important for coordination than shared factual knowledge.
|
|
|
|
## Limitations
|
|
|
|
The formalization is theoretical. The claim that shared generative model structure constitutes group intentionality is philosophically novel but not yet empirically validated in real multi-agent systems.
|
|
|
|
---
|
|
|
|
Relevant Notes:
|
|
- [[collective intelligence is a measurable property of group interaction structure not aggregated individual ability]]
|
|
- [[shared-anticipatory-structures-enable-decentralized-multi-agent-coordination]]
|
|
|
|
Topics:
|
|
- [[collective-intelligence/_map]]
|