Behavioral
Behavioral

Neural Networks: Connectionist Processing as Universal Principle

Behavioral Mechanics

Neural Networks: Connectionist Processing as Universal Principle

You have 20 cities. You need to find the shortest route visiting all of them exactly once. How long does this take?
stable·concept·1 source··Apr 27, 2026

Neural Networks: Connectionist Processing as Universal Principle

The Traveling Salesman Problem and Why Parallel Beats Serial

You have 20 cities. You need to find the shortest route visiting all of them exactly once. How long does this take?

Using serial processing (checking routes one by one): For 20 cities, you have roughly 20! possible routes (2.4 quintillion combinations). A computer checking one route per nanosecond would need billions of years.

Using parallel processing (connectionist network): A neural network can explore thousands of routes simultaneously. Each path strengthens or weakens based on performance feedback. Successful paths are amplified; failed paths are dampened. Within seconds, the network converges on a near-optimal solution.1

The difference is not small. It is the difference between impossible and trivial.

Bloom's insight: Individual minds and entire civilizations both work as connectionist networks, not serial processors. Your brain does not think through problems line-by-line like a computer program. Your brain explores thousands of possible thoughts simultaneously, dampens the ones that fail, strengthens the ones that work, and consciousness only perceives the winner.

A civilization does not plan its future through central authority. A civilization explores thousands of behavioral paths simultaneously (entrepreneurs trying businesses, scientists testing hypotheses, artists experimenting with forms), dampens the ones that fail (business bankruptcy, scientific refutation, artistic rejection), strengthens the ones that work (profitable businesses, confirmed theories, appreciated art), and emerges with a future no one planned.

Both systems—brain and civilization—are connectionist networks. Both are radically more powerful than serial processing would be. And both follow identical computational principles.


The Biological Feed: Why Parallel Processing Dominates

Serial processing is logically perfect but computationally impossible at scale. You can trace the reasoning step-by-step. But if the problem space is large, you will never finish.

Parallel processing is logically less certain but computationally powerful. You explore many paths at once. Most are wrong. But the right path usually emerges, even if you cannot explain why.

Evolution selected for parallel processors because they solve problems faster. An animal that can explore multiple behavioral possibilities simultaneously (fight, flee, freeze, negotiate) and pick the one that works will survive better than an animal that must deliberate serially.

Your brain is built as a parallel processor:

  • Billions of neurons firing simultaneously, not serially
  • Feedback loops that strengthen successful patterns and weaken failed ones
  • Multiple possible interpretations of sensory input, with the strongest one becoming conscious
  • Unconscious processing of millions of data points, with only the conclusion entering awareness

You experience your thought as "I decided X" but the reality is messier: your brain explored thousands of possibilities in parallel, dampened the unsuitable ones, and amplified one until it became conscious enough to report as "my decision."1

Civilizations are built the same way:

  • Millions of people pursuing different paths simultaneously
  • Market feedback (profit/loss) strengthening successful businesses, weakening failed ones
  • Scientific feedback (experimental confirmation/refutation) strengthening valid theories, weakening invalid ones
  • Cultural feedback (adoption/rejection) strengthening appealing memes, weakening ones that don't spread
  • Social feedback (status/shame) strengthening approved behaviors, weakening disapproved ones

The most successful civilization is not the one with the smartest central planner. It is the one that allows the most parallel exploration and provides the clearest feedback. This is why diverse, open societies outthink monolithic ones. Diversity = more parallel paths explored. Openness = clearer feedback about which paths work.


Analytical Case Study: Markets as Connectionist Networks

Consider price discovery in a market.

Serial approach: A central planner studies all available information about supply, demand, production costs, and consumer preferences. After months of analysis, the planner announces: "The price of wheat is $5 per bushel."

Parallel (market) approach: Thousands of farmers, traders, and consumers simultaneously signal their knowledge through offers to buy and sell. Prices continuously adjust based on these signals. Within seconds, the market discovers a price that balances supply and demand. No one planned this. No one has complete information. But the system found an answer that a serial planner could not have found faster.

The market price contains distributed information—everyone's knowledge compressed into a single number. A farmer knows her production costs and needs. A consumer knows his budget and preferences. A trader knows supply trends. None of them have the full picture. But their simultaneous actions (buying, selling) cause the price to move, and the price movement carries all their distributed knowledge in compressed form.

This is connectionist processing at scale. The market is a parallel processor. Prices are the output of millions of nodes exploring simultaneously and dampening failed approaches.1

This is why price controls fail. The central planner sets the price too low. Farmers stop producing (their costs exceed the price). Shortage. Consumer frustration. The planner responds by trying harder to force the price. This disrupts the feedback mechanism that normally coordinates supply and demand. The system becomes incoherent.

Conversely, deregulation works not because "markets are efficient" (they are not—they produce massive waste and suffering). Deregulation works because it restores the feedback mechanism. More parallel paths can be explored. More signals flow. The system becomes more coherent and powerful, even if less fair.


Implementation Workflow: Designing Better Connectionist Systems

How to recognize connectionist systems:

  1. No central plan, yet coordinated outcome. Markets organize without central authority. Languages evolve without grammar committees. Scientific progress emerges from thousands of independent researchers.

  2. Clear feedback loops connecting action to consequence. Traders get profit/loss feedback instantly. Scientists get experimental confirmation/refutation. Artists get audience response. The feedback is immediate and clear.

  3. Distributed knowledge aggregation. Price contains supply/demand information. Scientific consensus contains experimental data. Cultural taste contains aesthetic knowledge. The system compresses distributed information into simple metrics.

  4. Robust to node failure. If one economist is wrong, the market continues. If one scientist is wrong, science continues. If one artist fails, culture continues. No single node is critical.

How to improve a connectionist system:

  • Increase parallel exploration. More traders, more scientists, more artists, more entrepreneurs = more paths explored = more solutions found. Diversity increases exploration.
  • Clarify feedback loops. Better price signals, clearer experimental results, more transparent reputation systems = tighter feedback = faster learning. Transparency improves feedback.
  • Reduce barriers to entry. If good ideas from poor people are suppressed, you lose paths. If scientific ideas from outsiders are suppressed, you lose knowledge. Meritocracy improves exploration.
  • Prevent monopolistic lock-in. If one view dominates completely, parallel exploration stops. You need enough diversity that multiple views can coexist and compete.

How to damage a connectionist system:

  • Centralize decision-making. Move from market-based to command-and-control. Replace distributed feedback with central planning. The system immediately becomes less powerful.
  • Distort feedback loops. Propaganda that breaks the connection between action and honest consequence. Price controls that distort market signals. Censorship that blocks information flow.
  • Enforce conformity. Eliminate diversity of approaches. Force everyone into the same mold. Parallel exploration collapses to serial processing.
  • Protect incumbents. Regulatory capture, intellectual property lock-in, status quo bias. Prevent new paths from being explored. The system becomes rigid and brittle.

The Network as Prison: Why Connectionist Systems Are Hard to Reform

The most troubling implication: A connectionist system that is failing is nearly impossible to reform from within.

Here is the mechanism:

  1. The system has optimized for the current feedback structure.
  2. Individual nodes have adapted to the current incentives.
  3. Changing the feedback structure is perceived as threatening by nodes that benefited from the old structure.
  4. Those nodes resist change, often successfully, because they have power within the system.
  5. The system becomes locked into failure modes that cannot be escaped through incremental change.

Example: A market is clearly producing climate damage (the feedback loop does not include the cost of emissions). You want to fix it by adding a carbon tax (new feedback signal). The corporations that profit from pollution-heavy production resist with enormous political power. The system cannot reform itself because the node that benefits from the failure has power to prevent feedback-loop change.

Example: A scientific field is locked into a false paradigm (the feedback loop does not include disconfirming evidence because it is filtered out). A maverick scientist presents evidence that contradicts the paradigm. The system resists through peer review, funding denial, publication suppression. The system cannot reform itself because the nodes that benefited from the paradigm have power to prevent feedback-loop change.

The only ways to reform a connectionist system:

  • External shock (new information that cannot be ignored, new competition that cannot be resisted, crisis that forces adaptation)
  • Cascade failure (the system fails so completely that reformation becomes possible)
  • Revolutionary reset (destroy the old system entirely and build a new one from scratch)

Gradualism does not work because the system is optimized to resist incremental change.


Evidence / Tensions / Open Questions

Evidence for connectionist model:

  • Neural networks demonstrably solve optimization problems faster than serial processors1
  • Markets produce efficient price discovery without central planning
  • Scientific progress emerges from distributed research, not central planning
  • Cultural evolution produces complex patterns without top-down design
  • Biological systems (immune system, ant colonies, bird flocks) all show connectionist dynamics

Tensions in the model:

  • Central planning has succeeded. Soviet industrialization happened, China built infrastructure, NASA put humans on the moon. Command-and-control is slower than markets but not impossible.
  • Markets fail catastrophically. 1929 crash, 2008 financial crisis, environmental destruction. The feedback loop is not actually accurate—markets produce massive errors and inefficiencies.
  • Distributed systems are slow. Markets take time to discover price. Science takes years to confirm theories. Democratic decision-making is agonizingly slow. The speed advantage may be illusory.
  • Feedback loops are distorted. Markets optimize for profit, not human welfare. Science optimizes for publishability, not truth. Democracy optimizes for popularity, not justice. The feedback signal might be aligned with the wrong objective.

Open questions:

  • Can you have a connectionist system without feedback loops that reward competition? Can cooperation be the organizing principle?
  • What is the relationship between system complexity and the value of parallel vs. serial processing? When is serial actually better?
  • Can humans consciously design connectionist systems, or do they only emerge spontaneously?

Author Tensions & Convergences

Bloom draws from complexity theory (Kauffman, Holland) and cognitive science (McClelland, Rumelhart) but inverts the political implication. Complexity theorists present connectionist systems as elegant and powerful. Bloom emphasizes that they are blind and destructive when optimizing for the wrong objective.

A market is powerful at discovering prices, but indifferent to whether those prices include the cost of human suffering or environmental destruction. Science is powerful at discovering truth, but oriented toward publishable results, not wisdom. Democracy is powerful at distributing power, but mob-driven and short-sighted. The power is real; the wisdom is not.

The tension reveals: A system does not need to be wise to be powerful. A connectionist system optimized for the wrong objective will powerfully pursue the wrong goal. Fixing this requires either: (1) changing what the system optimizes for (change the feedback), or (2) accepting the system's failures as the price of its power.


Cross-Domain Handshakes

Psychology: How Individual Brains Work as Connectionist Networks

Connectionist Brain Architecture and Parallel Processing explains the neural substrate of connectionist thinking. Your brain is literally a parallel processor—billions of neurons firing simultaneously, with feedback loops strengthening successful patterns and weakening failed ones.

The handshake: Psychology explains how individual minds work as connectionist networks. Behavioral-mechanics explains how the same principles scale to collectives. Both operate through identical logic: parallel exploration, feedback-driven amplification and dampening, distributed problem-solving. This reveals that intelligence is not a property of individuals; it is a property of connectionist systems at any scale.

Practical implication: You think your intelligence comes from your individual brain's reasoning power. But your brain is not smart because of its reasoning—it is smart because it can explore thousands of possibilities simultaneously and pick the winners. You can improve your own connectionist processing by: (1) increasing parallel exploration (brainstorming, consulting diverse perspectives), (2) clarifying feedback (testing ideas against reality rather than defending them), (3) preventing lock-in (allowing yourself to abandon failed paths rather than defending them).

History: Connectionist Systems and Civilizational Power

Market Systems Versus Command Systems Across History documents how civilizations that allow connectionist processing outthink civilizations with centralized planning, and vice versa.

The handshake: History documents when connectionist systems have succeeded and failed. Behavioral-mechanics explains why—because parallel exploration and clear feedback are computationally superior to serial planning. Together they show that civilizational power is not mysterious—it follows from the structure of information flow. Civilizations that constrain information flow and feedback become brittle and fail. Civilizations that allow parallel exploration and clear feedback become robust and powerful.


The Live Edge

The Sharpest Implication

You believe you are thinking rationally, but you are experiencing the output of a connectionist network that is exploring thousands of possibilities unconsciously.

You feel like you are deliberating carefully and reaching conclusions through logic. The reality is messier: your brain is running thousands of neural processes in parallel. Most never reach consciousness. The few that do are the ones that the network determined were winners. You experience the winner as "my conclusion," but you did not consciously evaluate all the alternatives—your unconscious network did.

This is not a weakness. Parallel processing is more powerful than serial reasoning. But it means you are not in control of your thinking in the way you believe you are. Your conscious deliberation is more like post-hoc rationalization of what your connectionist network already decided.

Generative Questions

  • What parallel processes are running in your brain right now that you are completely unaware of? (Most of them. Consciousness only perceives the winners. The losers are invisible.)

  • If you could design the feedback loops your brain receives, what would you optimize for instead of what it currently optimizes for? (This reveals that feedback structure determines outcomes; change the feedback and you change the system.)

  • Are you exploring enough alternative possibilities, or have you locked into a single path? (Most people converge too quickly. Parallel processing is most powerful when you keep multiple paths alive longer.)


Connected Concepts


Footnotes

domainBehavioral Mechanics
stable
sources1
complexity
createdApr 27, 2026
inbound links3