Mental Models and Paradigms: The Deepest Leverage

By Pritesh Yadav June 22, 2026 14 min read —

So far in this book we have looked at the visible machinery of systems: stocks, flows, and the feedback loops that connect them. But there is a layer underneath all of that — a layer you cannot point to or measure. It is the set of beliefs in people's heads about how the world works. This chapter is about that hidden layer, because it turns out to be the most powerful place to change a system.

This is a counter-intuitive idea, and it is worth stating up front. The things that feel most concrete in a system — budgets, prices, rules — usually move the system the least. The thing that feels most intangible — what people believe — usually moves it the most.

What a mental model is

A mental model is an internal, simplified picture of how some part of the world works. It is the set of assumptions and beliefs you use to make sense of what you see and to decide what to do. You carry thousands of them: how a meeting should run, what a good employee looks like, whether strangers can be trusted, how a printer jams.

Donella Meadows, whose work has guided much of this book, defined them this way: "Mental models are the images, assumptions, and stories which we carry in our minds of ourselves, other people, institutions, and every aspect of the world."

Here is the crucial point. Every human action is based on a mental model — you cannot act without one. So the problem is never that you have models. The problem is that every model is incomplete and often wrong, and that we forget they are models at all. We mistake the map for the territory and stop questioning it.

Analogy: Mental models are like prescription glasses. They make the world legible, but they also distort everything you see — and you don't notice the distortion because you are always looking through the lenses, never at them. Surfacing a model means taking the glasses off and looking at the frame. For a moment the world goes blurry. That discomfort is not a sign you are wrong; it is the sign you are finally seeing the frame.

The iceberg model: four levels of seeing

Systems thinkers use an iceberg to show that what we notice — events — sits on top of three deeper, less visible layers. Most icebergs hide most of their mass below the waterline, and so do systems.

        ~~~~~~~ waterline ~~~~~~~
   |  EVENTS        what just happened
   |--------------------------------------
   |  PATTERNS      what keeps happening
   |  STRUCTURES    rules, flows, feedback
   |  MENTAL        the beliefs that built
   |  MODELS        the structures
   v  (deepest, least visible, most power)

Level 1 — Events: what happened. A product recall, a missed deadline, a traffic jam.
Level 2 — Patterns: what has been happening over time. Recalls becoming more frequent; deadlines repeatedly slipping.
Level 3 — Structures: the rules, incentives, flows, and feedback loops that produce those patterns — cost-cutting that quietly reduces quality checks, scheduling norms that always under-estimate the work.
Level 4 — Mental models: the beliefs and values that created and maintain those structures — "cost is the only variable we control," "a schedule is a negotiation, not a commitment."

Most organisations fight fires at Level 1, because Level 1 is what they can see. They fire the manager after the recall and patch the hole after the breach. But if the structure and the underlying belief are unchanged, the same root cause simply produces a new event in a new form. Sustainable change requires reaching Level 4.

Common mistake: Over-relying on events as the signal. Because events are dramatic and visible, teams treat each one as unique — a separate post-mortem, a separate fix — instead of as a symptom of one structure and one belief. The result is a permanent fire-fighting culture.

Paradigms: the shared mental model

A paradigm is the shared mental model of a whole society, organisation, or field. It is the set of unstated beliefs so widely held that they are presumed to be true rather than believed. Meadows called it "the mindset or paradigm out of which the system — its goals, power structure, rules, its culture — arises."

Some paradigms she named: "Nature is a stock of resources to be converted to human purpose." "One can own land." "Growth is always good." Notice that these don't feel like opinions; they feel like facts. That is exactly what makes a paradigm so powerful — and so invisible.

Analogy: Paradigms are the water fish swim in. A fish does not think about water; it thinks with water. The paradigm shapes which questions get asked and which solutions seem possible. This is why paradigm shifts so often come from outsiders — Tesla was not a car company, Netflix was not a video store — because an outsider is not breathing the same water.

Meadows' leverage points: why paradigm is near the top

A leverage point is a place in a system where a small push produces a large change in behaviour. Meadows ranked twelve of them, from least to most powerful. The ranking surprises almost everyone, because the things that feel most important have the least leverage.

Rank	Leverage point	How it feels
12 (weakest)	Numbers — subsidies, prices, taxes	Concrete, urgent, fought over
5	Rules of the system	Powerful
4	Power to self-organise	More powerful
3	Goals of the system	Deep
2	Paradigm — the mindset itself	Almost invisible
1 (strongest)	Transcending paradigms	Intangible

Numbers — the budgets and prices people argue about endlessly — are leverage point 12, the weakest. Paradigm is leverage point 2. The mindset from which the whole system arises is second only to one thing.

The highest leverage point: transcending paradigms

Meadows' number one leverage point is the ability to hold any paradigm lightly — to know that every model is partial and none is the final truth. She wrote that "mastery has less to do with pushing leverage points than it does with strategically, profoundly, madly letting go."

This is not relativism (the idea that all views are equally valid). It is epistemic humility — keeping your current best model firmly enough to act on it, while staying genuinely ready to revise it when the evidence demands.

Common mistake: Confusing "transcending paradigms" with having no convictions. It means the opposite of paralysis. A working scientist uses a paradigm every single day and also knows it is provisional. The two traps to avoid are rigid certainty (my model is reality) and paralysed scepticism (no model is trustworthy, so I can't act).

How paradigms change

Drawing on Thomas Kuhn's The Structure of Scientific Revolutions (1962), Meadows described how a paradigm shifts. First, anomalies accumulate — results the existing paradigm cannot explain. (Kuhn's example: Mercury's orbit had a wobble that Newton's physics could not account for, which Einstein's later did.) People first dismiss anomalies as puzzles to be solved later. When too many pile up to ignore, a crisis breaks out, competing theories arise, and eventually one becomes the new paradigm.

Meadows added a practical lever: you can speed a paradigm shift by "repeatedly and consistently pointing out anomalies and failures in the current paradigm to those with open minds."

Common mistake: Treating paradigm change as purely intellectual. Winning an argument almost never changes a paradigm. People change their minds when the old model demonstrably fails in their own experience, not when they are out-argued. Whitepapers and workshops aimed at persuasion usually fail at this level.

Why mental models defend themselves: the ladder of inference

Chris Argyris built the ladder of inference (later popularised by Peter Senge in The Fifth Discipline, 1990) to show how a mental model becomes self-sealing — that is, how it protects itself from being disproved.

  ACTIONS ----> produce results that confirm beliefs
     ^                                |
     |  BELIEFS                       |
     |  CONCLUSIONS                   |
     |  ASSUMPTIONS                   |
     |  ADD MEANING                   |
     |  SELECT DATA <-----------------+
     |  OBSERVABLE DATA   (reflexive loop)

We climb the ladder fast: from all the observable data we select a slice (filtered by what we already believe), add meaning, make assumptions, draw conclusions, adopt beliefs, and act. The sting is in the loop at the bottom: our beliefs decide which data we notice in the first place. So the model grows stronger even when the world contradicts it. Senge called this the reflexive loop.

Common mistake: The ladder cycling faster and faster. Experts are especially exposed — their quick, accurate pattern-matching inside their field becomes quick, inaccurate pattern-matching outside it. The speed itself is the danger: the intermediate steps become invisible even to the expert. This is closely related to the curse of knowledge (named by Camerer, Loewenstein, and Weber in 1989) — once you know something, you cannot un-know it, so you systematically underestimate how opaque your model is to a beginner.

Single-loop versus double-loop learning

Argyris drew a line between two kinds of learning. Single-loop learning adjusts your actions to hit a goal without questioning the goal. Double-loop learning questions the goal itself.

Analogy: A thermostat is set to 68°F. The room is cold, so it turns on the heat — single-loop. It never asks, "Is 68°F the right setting? Who chose it, and why?" — that is double-loop. Most organisations are excellent thermostats: fast, precise, consistent. But if the set-point came from a wrong assumption nobody revisits, they efficiently chase the wrong goal forever.

Double-loop learning is how mental models actually get updated. It is blocked by what Argyris called defensive routines, by fear of looking incompetent, and by authority gradients that silence dissent. It is enabled by psychological safety, devil's-advocate roles, pre-mortems, and anonymous polling before group discussion.

How mental models fail in organisations

Senge catalogued seven recurring "learning disabilities," all rooted in faulty mental models. A few stand out:

"The enemy is out there" — the problem is always blamed on an outsider, never on the system itself. (We saw this temptation throughout the feedback-loop chapters: structure, not villains, drives most behaviour.)
"The fixation on events" — attention stuck at iceberg Level 1, never reaching patterns.
"The boiled frog" — slow, gradual change is not perceived as a threat.
"The myth of the management team" — a group that looks like it reasons together but actually suppresses disagreement.

Common mistake: Citing the boiled frog as biology. Real frogs do jump out of slowly heated water. The parable is only a named metaphor for how human threat-detection is tuned to sudden change, not slow drift. Use it as a metaphor, not a fact.

Paradigms in the wild

Example — Kodak: Kodak's own engineers invented the digital camera, and internal forecasts showed film would be replaced. Yet the operating paradigm was "we are a film company; film is our business." That belief decided what counted as a threat. Embracing digital meant dismantling a hugely profitable film business, so protecting it felt rational. Kodak filed for bankruptcy in January 2012. The lesson is not lack of information — it is that the paradigm dictated how information was read. (Blockbuster told the same story when it passed on buying Netflix around 2000 for roughly $50 million.)

Example — Toyota vs. Taylorism: Frederick Taylor's scientific management rested on "workers are interchangeable executing units; engineers design the work." The Toyota Production System reversed it: "workers are thinking partners; improvement comes from the people doing the work." Toyota even lets any worker pull a cord to stop the line for a quality problem. In the 1984 GM–Toyota NUMMI plant, the same workers, building, and equipment that had produced GM's worst quality became one of America's most productive plants within two years under Toyota's management. Nothing physical changed — only the paradigm.

Example — The Goal (Goldratt, 1984): Eliyahu Goldratt's Theory of Constraints attacked the paradigm "efficiency everywhere is good; a busy machine is a productive machine." His counter-paradigm: throughput is limited by one constraint, so improving anything that is not the constraint does not help the system — it just piles up inventory and lengthens lead times. "A balanced plant is a bankrupt plant," he wrote. An hour lost at the bottleneck is an hour of throughput lost for the entire system. A genuine paradigm shift, because it overturned a belief that felt obviously true.

Analogy — the iceberg and the ocean liner: The event is the collision. The pattern is days of ice warnings. The structure is the routing policy and watch schedule that kept the ship at full speed. The mental model is the paradigm of owner and captain: "this ship is unsinkable; the schedule is non-negotiable." Investigations dwell on the collision and the watch schedule. The deepest layer — the belief — is almost never the headline.

Practical techniques for surfacing mental models

Because models hide, you need deliberate tools to drag them into the light. Here are five drawn from the masters of this field.

The left-hand column: (Argyris and Schön, 1974.) Divide a page. On the right, write what was actually said in a hard conversation. On the left, write what you were thinking but did not say. The gap exposes your operating assumptions.
Ladder-of-inference walk-back: Trace a decision backward, rung by rung: "What data led to that conclusion? What meaning did I add? What assumption am I making?"
Pre-mortem: (Gary Klein.) Before a project starts, imagine it is six months later and has failed badly. Everyone writes, independently, every plausible reason. This externalises and tests the model before it produces a costly outcome.
Balancing advocacy with inquiry: (Senge.) For every opinion, add your reasoning and invite challenge: "Here is what I think and here is why — what is your view?" This makes the model visible and testable.
Surfacing anomalies: (Meadows.) Find results the current paradigm cannot explain, and keep raising them with open-minded people.

Tip: Watch behaviour, not self-report. Argyris distinguished espoused theory (what we say we believe) from theory-in-use (what our actions reveal we believe). They diverge constantly. To find a real mental model, look at what people actually do.

Common mistake: Mistaking good intentions for paradigm change. Announcing "we now put customers first" changes nothing if the incentive system still rewards cost-cutting. Structure follows paradigm — so if the structure has not changed, neither has the real paradigm. (Compare the "customer is a cost" mindset, which minimises service spend, with the "customer is an asset" / lifetime-value mindset that Amazon, Costco, and subscription businesses built entirely different structures around.)

Key takeaway: The deepest leverage in any system is not in its numbers or even its rules — it is in the beliefs that produced those rules. Change what people take to be obviously true, and the goals, structures, and behaviours reorganise themselves around the new belief.

Key Takeaways

Everyone acts on mental models; the danger is operating on ones you have never examined — mistaking the map for the territory.
The iceberg model shows four levels — events, patterns, structures, mental models. Fire-fighting lives at Level 1; durable change requires reaching Level 4.
A paradigm is a shared mental model so widely held it feels like fact. In Meadows' ranking, paradigm is the second-highest leverage point, far above prices and budgets.
The highest leverage point is transcending paradigms: holding your best model firmly enough to act, lightly enough to revise — humility, not paralysis.
Mental models defend themselves through the ladder of inference; updating them needs double-loop learning, which questions the goal, not just the action.
Paradigms change when anomalies accumulate until the old model visibly fails — Kodak, Nokia, and Taylorism all show that information alone is not enough; the operating belief decides what the information means.

Continue reading

🔄 Systems Thinking

Systems Thinking vs. Linear Thinking

🔄 Systems Thinking

What Is a System? Parts, Connections, and Purpose

🔄 Systems Thinking

Stocks and Flows: What Builds Up and What Moves