Society of mind
Explanation
Marvin Minsky, one of the founders of artificial intelligence, published in 1986 a book titled The Society of Mind with a provocative thesis: the mind is not a single unified entity, but a swarm of specialised agents that cooperate and compete. Each agent is simple and dumb on its own; intelligence emerges from the organised interaction of thousands of them.
Minsky took the metaphor from human societies. Just as a city operates through the coordination of millions of people each performing limited tasks, the mind would operate through the coordination of thousands of mental processes each specialised (recognising faces, detecting motion, planning routes, producing language, feeling fear). No agent knows what the mind as a whole does; the mind as a whole is what they do together.
This image has computational roots: Minsky had been working on symbolic AI for decades and could see clearly that the hard problems of intelligence were not solved with a single brilliant algorithm but with modular architectures. Subsequent cognitive architectures (SOAR, ACT-R, LIDA, CLARION) are direct descendants of this intuition.
For consciousness, the society of mind has profound implications. If there is no unified I but a federation of agents, the sense of unity of consciousness would be a downstream effect, a narrative generated for communicative and self-reflective use, not a primary datum. Dennett would elaborate this idea in his Theory of the Narrative Centre of Gravity.
These ideas chimed with neuroscientific findings: split-brain patients show that the two hemispheres can act as semi-independent agents; localised brain lesions affect specific functions without disorganising the whole mind; the experiments of Libet and others show unconscious processes preceding conscious will.
The society of mind remains one of the most fertile metaphors for understanding the mind. Modern cognitive architectures, multi-agent systems in AI, research on cognitive modularity and the neuroscience of brain networks have vindicated it from various angles. The open question is how and why a unified sense of being a subject emerges from that federation of agents.
Strengths
- Computationally articulable model.
- Coherent with findings on cerebral modularity.
- Decisive influence on AI and robotics.
- Allows naturalising consciousness without homunculi.
Main critiques
- Does not explain qualia or the hard problem.
- The 'agent' is a black box: what makes a collection of non-minds produce a mind?
- Risk of regress: within each agent there can be sub-agents.
- Insufficient attention to phenomenological unity.