Systems Dimensions: Decision Systems, Learning, and Social Calculus
This is post 6 of 23 in a series on Systems Thinking: Managing Chaos and Complexity by Jamshid Gharajedaghi (ISBN: 978-0-7506-7973-2).
In the previous post, we covered throughput, membership, and conflict management. Now we’re getting into some of the meatiest parts of Chapter 3: how decisions actually get made, how organizations learn, and how we measure things that don’t fit neatly into spreadsheets.
Decision Systems: Power Isn’t a Pie
Here’s a question most organizations get wrong: should decisions be centralized or decentralized?
Gharajedaghi says that question itself is the problem. It sets up a false choice. Centralize everything and you get suffocation. Decentralize everything and you get chaos. Try to split the difference by “sharing power” and you get gridlock.
His answer? Power can be duplicated. It’s not a pie where giving someone a slice means you have less. It’s more like knowledge. You can give it away and still have it.
This is the key idea behind real empowerment. It’s not about handing over control. It’s not about micromanaging from a distance. It’s about making sure everyone understands why decisions are made the way they are.
The Boss Story
Gharajedaghi tells a great story here. You start a new job. Boss promises decentralization, says he’ll judge you on results only. You go off and do interesting work. But every time you share your ideas, he clearly hates them. Eventually he starts showing up in your office asking pointed questions. “Decentralization” is dead.
Why? You never aligned on how to think about decisions. You had freedom but no shared framework.
Then you take a different job. New boss tells you exactly what to do, but also explains why. She walks through her reasoning, criteria, values. Over time, you absorb how she thinks. One day you suggest something and she loves it. Decentralization just happened. Naturally.
The lesson: real decentralization comes from shared decision criteria, not from handing over authority and hoping for the best.
Decision Criteria vs. Decisions
This distinction matters more than it sounds. Decision criteria are the rules for making decisions. Actual decisions are just applying those rules to specific situations.
Gharajedaghi breaks criteria into two types:
Policies deal with the big picture. Why are we doing this? What variables matter? What outcomes do we expect? These are value-loaded choices.
Procedures deal with how. They specify the methods for applying policies to real situations.
Here’s the critical part: policies should leave room for flexibility. They need to be specific enough to keep things consistent, but broad enough that people can adapt to their specific context. And procedures should be developed by the people who actually execute them. Not by someone three levels up who has never touched the work.
The Circular Organization
To make all this work at scale, Gharajedaghi describes a structure called the circular organization, originally published by Russell Ackoff in 1981. Every manager forms a “design cell” with their boss and their direct reports. Since each cell spans three levels, it creates a nested network where everyone interacts with at least three levels of the organization.
The point isn’t bureaucracy. It’s creating shared understanding. These cells exist so people learn why the organization does what it does and collectively own the decision criteria.
Getting to Consensus
One thing I appreciate about this section is how honest it is about consensus. Gharajedaghi doesn’t pretend consensus is easy or always possible. He says it’s not majority rule and it doesn’t even require unanimity. It’s an agreement to act.
When consensus can’t be reached, that’s when leadership matters most. The leader needs to create a working synthesis of different positions. People don’t have to agree with it, but they need to feel heard, understand why the decision was made, and be willing to support it.
And if even that doesn’t work? The opposing sides should agree on an experiment. Design a test, set clear criteria, and let the results decide. I really like this idea. Instead of arguing forever, just run the experiment.
Learning and Control Systems
The other half of decision-making is learning. And learning, according to Gharajedaghi (building on Ackoff’s management system framework), starts with being surprised.
You expected something to happen. Something different happened instead. That gap between expectation and reality is where learning lives.
But you only learn if you do two things: figure out why the mismatch happened, and then change what you do so it doesn’t happen again. Otherwise, you just had an unpleasant surprise.
Tracking What You Expected
For this to work, you need a formal process. When you make a major decision, you record the assumptions behind it, the information used, the process followed, and what you expected to happen.
Then, when reality doesn’t match, you check four things: Was the data wrong? Was the implementation wrong? Was the decision itself wrong? Or did the environment change?
This sounds obvious. But most organizations don’t do it. They make decisions, move on, and scramble when things go wrong without any clear record of what they originally expected.
Early Warning Systems
The best version of this includes an early warning system. Instead of waiting for problems to explode, you continuously monitor whether your original assumptions still hold. If the assumptions are no longer valid, you can course-correct before the damage is done.
This is proactive learning versus reactive firefighting. And honestly, most organizations are stuck in reactive mode.
Social Calculus: Measuring What Actually Matters
This is where Gharajedaghi gets really ambitious. He argues that you can’t control purposeful individuals through supervision alone. You need a new way of measuring and aligning behavior. He calls it social calculus.
The idea has three dimensions:
Vertical compatibility asks: do the interests of people at different levels align? If a worker’s success depends on undermining the larger organization (or vice versa), you have a structural problem.
Horizontal compatibility asks: do peer units work well together? This is where he brings up the classic conflict between marketing and production. Marketing wants variety, customization, and flexibility. Production wants standardization and long runs. If marketing is measured on revenue and production is measured on cost, they’re structurally set up to fight each other.
His solution is elegant: make both of them profit centers. Instead of one being a cost center and the other a revenue center, both should maximize the difference between cost and revenue. Now they have compatible goals. A win for one can be a win for the other.
Temporal compatibility asks: are the interests of past, present, and future members aligned? This one is less obvious but really interesting. If current members see that the organization treats former members badly (poor retirement, broken promises), they learn something about their own future. That breeds insecurity, cynicism, and self-serving behavior. On the flip side, if past members’ interests completely dominate (tradition for tradition’s sake), the organization can’t adapt.
The Contribution Ratio
Gharajedaghi proposes measuring economic activities not just by financial return, but by total contribution to society. You factor in how scarce the consumed resources are, how needed the outputs are, and how much employment gets created. The result is a contribution ratio.
His example: bread production might have low margins but a high contribution ratio (people need bread, it creates jobs). Yo-yos might have great margins but a low contribution ratio. Social calculus would shift incentives through differentiated loans, interest rates, and taxes to make bread more attractive to invest in.
Is this practical? Debatable. But the framework forces a good question: what are we actually rewarding, and is that what we should be rewarding?
Putting It All Together
At the end of the chapter, Gharajedaghi summarizes what it takes to manage a multiminded system:
- Align member interests with the purpose of the whole (and vice versa)
- Empower through duplication of power, not abdication
- Separate control from service and turn it into a learning function
- Prevent win/lose struggles and dissolve paralyzing conflicts
These four points tie together everything from both parts of Chapter 3. The systems dimensions aren’t separate topics. They’re interconnected aspects of the same challenge: how do you create an organization where independent, purposeful people can work together without someone standing over them?
The answer is design. Not control, not compromise, not wishful thinking. Intentional design of decision systems, measurement systems, and learning systems that make cooperation the natural path. That’s what makes this chapter one of the most practical in the entire book.
Previous: Systems Dimensions - Throughput, Membership, and Conflict