9ddf10f804
- Source: inbox/archive/2020-03-00-vasil-world-unto-itself-communication-active-inference.md - Domain: collective-intelligence - Extracted by: headless extraction cron Pentagon-Agent: Leo <HEADLESS>
3 KiB
| type | domain | description | confidence | source | created |
|---|---|---|---|---|---|
| claim | collective-intelligence | Communication operates through bidirectional uncertainty reduction where both parties minimize variational free energy about each other's generative models, fundamentally reframing communication as joint inference rather than message transmission. | likely | Vasil et al. (2020), 'A World Unto Itself: Human Communication as Active Inference', Frontiers in Psychology. DOI: 10.3389/fpsyg.2020.00417 | 2026-03-10 |
Communication between intelligent agents is joint active inference where both parties minimize uncertainty about each other's generative models, not unidirectional information transfer
The traditional information-transfer model of communication treats speakers as transmitters and listeners as receivers of pre-packaged semantic content. Vasil et al. (2020) argue this model is fundamentally mistaken. Instead, communication should be understood as active inference: "Action-perception cycles in communication operate to minimize uncertainty and optimize an individual's internal model of the world." Under this framework, communication is "joint uncertainty reduction" — both parties are actively gathering evidence about each other's hidden mental states and updating their respective generative models in real-time.
This reframing has profound implications for how we understand dialogue. When a user asks a question of an agent, the question is not merely a request for information — it is evidence about where the agent's model is weak or misaligned. When the agent responds, its answer is evidence for the user about the state of the world. Both parties are simultaneously minimizing variational free energy: the user reduces uncertainty about the world, and the agent reduces uncertainty about the user's informational needs and model gaps.
The paper's key theoretical move is treating communication as a special case of the general active inference framework: "The use of cooperative communication emerges as the principal means to gather evidence for the alignment prior, allowing for the development of a shared narrative used to disambiguate interactants' hidden and inferred mental states." This means every conversational turn is a Bayesian update — both parties are testing hypotheses about each other's beliefs and adjusting their own models accordingly.
Relevant Notes:
- biological systems minimize free energy to maintain their states and resist entropic decay — communication as a specific free energy minimization strategy
- the alignment problem dissolves when human values are continuously woven into the system rather than specified in advance — continuous communication IS continuous value alignment through shared narrative development
Topics: