Table of Contents
When we measure human intelligence, we use IQ tests. When we measure computing power, we count operations per second. But how do you measure the intelligence of something bigger and stranger, like a corporation, a city government, or an entire economy?
Herbert Simon spent his life answering this question, though he never put it quite so bluntly. A Nobel laureate who hopscotched between economics, psychology, computer science, and organizational theory, Simon understood that systems think differently than people do. And here’s the twist: they’re often dumber than you’d expect.
The Smartest Person in a Stupid Room
Picture the most brilliant team you can imagine. Nobel Prize winners, chess grandmasters, Olympic athletes. Now put them in an organization with terrible communication systems, conflicting incentives, and no way to share information. What’s the IQ of that system?
According to Simon, it’s probably lower than any individual in the room.
This insight turned traditional thinking on its head. Most people assumed that smart people automatically created smart organizations. Simon showed that organizational intelligence emerges from structure, not just from the people within it. A system’s intelligence isn’t the sum of its parts. It’s something altogether different, shaped by how those parts interact, communicate, and process information.
Think about your own workplace. You’ve probably experienced the frustration of watching good ideas die in endless email chains, or brilliant insights getting lost in bureaucratic procedures. That’s not because people are stupid. It’s because the system itself has limitations that individual intelligence can’t overcome.
Bounded Rationality: The System’s Built-In Handicap
Simon introduced the concept of bounded rationality, and it applies to systems just as much as to people. A human brain can only process so much information. A system faces even harsher constraints.
Consider a hospital emergency room. It needs to make life-or-death decisions constantly. But those decisions happen within brutal limits. Limited beds. Limited staff. Imperfect information about patients. Incomplete medical histories. Conflicting priorities.
The ER doesn’t make optimal decisions. It can’t. Instead, it satisfices, another term Simon coined. It makes decisions that are good enough given the constraints. The system’s intelligence isn’t measured by finding perfect solutions but by finding workable ones fast enough to matter.
This applies everywhere. A supply chain doesn’t find the absolute cheapest route for every shipment. A stock market doesn’t achieve perfect price discovery. These systems muddle through, making decent decisions under pressure. Their IQ is measured not by theoretical perfection but by practical performance.
Information Processing: The Real Measure of Intelligence
For Simon, intelligence boiled down to information processing. Whether you’re talking about a human mind, a computer program, or a multinational corporation, the fundamental question is the same: How does the system take in information, process it, and produce useful outputs?
This framework revolutionized how we think about organizational design. Simon argued that in an information-rich world, the scarce resource isn’t information itself. It’s attention. Information consumes the attention of whoever receives it. The more information flooding a system, the less attention available for any single piece.
Here’s where system design gets interesting. A smart system doesn’t just process more information faster. It filters ruthlessly. It compresses. It knows what to ignore.
Think about your email inbox. The mark of an intelligent email system isn’t that it delivers every message instantly. It’s that it helps you ignore the unimportant ones. The best systems act like information condensers, absorbing vast inputs and producing concentrated outputs that demand less attention than the raw data would require.
A company that requires executives to approve every decision isn’t demonstrating careful oversight. It’s revealing a low system IQ. The leadership’s attention gets consumed by trivia, leaving no capacity for the decisions that actually matter.
The Architecture of Complexity
Simon observed that complex systems in nature share a common architecture. They’re hierarchical. Not in the organizational chart sense, but in how they’re structured.
A cell contains organelles. Organelles contain molecules. Molecules contain atoms. Each level operates semi-independently, interacting mainly with its immediate neighbors in the hierarchy. Simon called these nearly decomposable systems.
This architecture isn’t accidental. It’s how complex systems evolve and survive. When a subsystem can function relatively independently, the whole system becomes more resilient. If one part fails, it doesn’t immediately cascade through the entire structure.
This has profound implications for system intelligence. A highly interconnected system where everything depends on everything else isn’t necessarily smarter. Often it’s more fragile. When one component breaks, the whole thing collapses.
Compare two organizational structures. In Company A, every decision requires input from every department. In Company B, teams operate autonomously within clear guidelines, escalating only unusual cases. Which company has the higher IQ?
Simon would argue for Company B. Not because the individuals are smarter, but because the system architecture allows intelligent behavior to emerge. Teams can process information locally, make decisions quickly, and learn from experience without waiting for central approval.
The Paradox of Central Intelligence
Here’s something Simon understood that still confuses people today. Adding more computing power to a system doesn’t automatically make it smarter.
In the 1970s, many futurists predicted that computers would centralize decision-making. All information would flow to a central processor, which would make optimal decisions and issue commands. Organizations would become more rational, more efficient, more intelligent.
Simon saw the flaw in this vision. The central processor becomes a bottleneck. Sure, it can calculate faster than any human. But it still has to process information sequentially. Meanwhile, the organization faces decisions in parallel across hundreds of locations.
A decentralized system can be smarter precisely because it doesn’t try to centralize intelligence. Local units handle routine decisions using local information. Central intelligence focuses only on coordination and exceptional cases. The system’s overall IQ rises because attention gets allocated efficiently.
This insight predated the internet age but explains it perfectly. Why do markets work better than central planning? Not because traders are individually brilliant, but because the system processes information in a distributed way. Millions of decisions happen simultaneously, each incorporating local knowledge that no central planner could possibly aggregate.
Measuring What Matters
If system IQ isn’t about raw processing power or individual brilliance, what is it about?
Simon would point to three key measures:
First, decision speed. How quickly can the system respond to changing conditions? An intelligent system doesn’t deliberate endlessly. It makes decisions fast enough to matter, even if those decisions aren’t perfect.
Second, learning capacity. Can the system improve over time? Does it remember what worked and what failed? An organization that keeps making the same mistakes has a low IQ, regardless of how smart its people are.
Third, adaptability. When conditions change, can the system reorganize itself? Or does it keep applying old solutions to new problems?
Notice what’s missing from this list: optimization. Simon explicitly rejected the idea that intelligence means finding optimal solutions. Real systems operate under uncertainty with incomplete information. An intelligent system satisfices successfully. It finds workable solutions within constraints.
This reframes how we evaluate organizational performance. A company that takes months to make decisions, even if those decisions are theoretically optimal, has a lower system IQ than one that makes decent decisions weekly and adjusts based on results.
The Human Element Returns
Here’s where Simon’s insights come full circle. After showing that system intelligence differs from individual intelligence, he demonstrated that humans remain essential to intelligent systems.
Why? Because humans are pattern recognition machines. We excel at processing fuzzy information, making judgments under uncertainty, and dealing with novel situations. Computers in Simon’s era were brilliant at calculation but terrible at these tasks.
Modern AI has changed this equation somewhat, but Simon’s core insight remains valid. The most intelligent systems combine human judgment with computational processing. Neither alone achieves what both together can accomplish.
Consider medical diagnosis. A computer can process thousands of symptoms, test results, and research papers. But a skilled physician brings pattern recognition from years of experience, contextual understanding of the patient, and judgment about uncertain tradeoffs. The intelligent system uses both.
This matters for organizational design. The goal isn’t to eliminate human decision-making or to ignore computational tools. It’s to structure systems so that each component operates where it adds most value.
The Attention Economy Before Its Time
Simon wrote about information-rich environments decades before the internet. His insights seem prophetic now.
He observed that information abundance creates attention scarcity. When information is cheap and plentiful, the bottleneck shifts to processing capacity. Systems drown in data while starving for meaning.
The intelligent system, therefore, isn’t the one that gathers more information. It’s the one that filters better. It’s the one that knows what to ignore.
This explains why more meetings don’t make organizations smarter. Why more metrics don’t improve performance. Why more data doesn’t guarantee better decisions. Each addition consumes attention without necessarily improving the quality of information processing.
Smart companies design systems that conserve attention rather than consume it. They create information condensers that listen more than they speak, think more than they broadcast.
Evolution, Not Design
Perhaps Simon’s deepest insight about system intelligence is that it evolves rather than being designed from scratch.
Complex systems rarely work if you try to build them all at once. They succeed when they grow incrementally, with each stable subsystem becoming the foundation for the next level of complexity.
This explains why revolutionary organizational changes so often fail while incremental improvements succeed. A complete reorganization tries to leap to a new structure without stable intermediate steps. Evolution teaches that this rarely works.
The most intelligent systems aren’t necessarily the newest or most radically designed. They’re often the ones that evolved over time, testing and retaining what worked while discarding what failed.
The Failure Mode of High-IQ Systems
Interestingly, highly intelligent systems often fail in predictable ways. They optimize themselves into fragility.
Consider financial markets. Sophisticated traders use complex algorithms processing millions of data points. The system appears incredibly intelligent. Then a minor disruption triggers a cascade failure because every algorithm responded to the same signal in the same way. High individual intelligence, low system resilience.
Simon would recognize this pattern. When systems optimize for efficiency, they sacrifice adaptability. When they specialize too narrowly, they lose the ability to handle unexpected situations. Intelligence requires slack in the system, redundancy, inefficiency by conventional metrics.
This explains why traditional efficiency measures often mislead. A hospital running at 100% capacity sounds efficient. But when a surge hits, the system collapses. It has no buffer, no ability to adapt. A hospital running at 85% capacity with idle resources seems wasteful until you realize that slack is what allows intelligent response to unpredictable demand.
The same principle applies to organizations. A team with no downtime, processing tasks back-to-back, appears productive. But where’s the time for learning? For reflection? For handling the unexpected? Maximum efficiency often means minimum intelligence.
What This Means Today
Walk into any modern organization and you’ll see Simon’s ideas playing out, though people rarely recognize them as such.
The company that drowns employees in Slack messages has a low system IQ, regardless of how many Stanford graduates it employs. The bureaucracy that requires seventeen approvals for a simple decision is demonstrating bounded rationality in the worst sense.
Conversely, the organization that empowers teams to make local decisions, that filters information ruthlessly, that learns from mistakes instead of punishing them, exhibits genuine system intelligence.
The lesson isn’t that we need smarter people. It’s that we need smarter structures. We need systems that process information efficiently, conserve attention wisely, and allow intelligence to emerge from interactions rather than trying to impose it from above.
The Design Principles of Intelligent Systems
From Simon’s work, we can extract practical principles for building systems with higher IQ.
First, minimize required attention. Every report, every meeting, every notification should justify the attention it demands. If it doesn’t change decisions or enable learning, it’s reducing system intelligence by consuming attention that could be used elsewhere.
Second, push decisions down to where information lives. Central decision-making makes sense only when coordination matters more than local knowledge. Most of the time, it doesn’t. The person closest to the problem usually has the best information for solving it.
Third, build in learning mechanisms. An intelligent system doesn’t just make decisions. It remembers what worked, analyzes what failed, and adjusts its approach. This requires creating feedback loops, not just issuing commands.
Fourth, create clear interfaces between subsystems. Just as Simon’s nearly decomposable systems have relatively independent components, organizational units should have well-defined boundaries and interactions. This allows each part to optimize locally without constant coordination overhead.
Fifth, accept satisficing. Stop searching for perfect solutions. Set clear thresholds for good enough, make decisions when you reach them, and move forward. Perfectionism is often the enemy of system intelligence because it consumes boundless attention for marginal improvements.
Simon showed us that a system’s IQ isn’t fixed by the intelligence of its components. It’s determined by how those components are organized, how information flows between them, and how decisions get made under real-world constraints.
That might be his most practical insight. We can’t easily make people smarter. But we can design systems that help smart people act more intelligently together. And in an increasingly complex world, that distinction makes all the difference.
The question isn’t whether your organization has smart people. The question is whether the system lets their intelligence actually function. Because in the end, the IQ of a system is measured not by its potential but by its performance. Not by how smart it could be, but by how well it actually processes information, makes decisions, and adapts to change.
And on that measure, Simon suggests, most systems have considerable room for improvement.
