Double-loop learning is a way of solving problems by questioning the underlying beliefs and assumptions that caused the problem, not just adjusting your actions. The concept was developed by organizational theorists Chris Argyris and Donald Schön in the late 1970s, and it remains one of the most influential ideas in management and organizational learning. Where most problem-solving focuses on trying a different approach to reach the same goal, double-loop learning asks whether the goal itself is the right one.
How the Two Loops Work
Every action you take is shaped by a set of “governing variables,” the values, beliefs, and assumptions you hold about how things should work. These variables operate like internal rules: “If I do X, then Y should happen.” When Y doesn’t happen, you have two options for how to respond.
In single-loop learning, you change your action. You try a different tactic, tweak the process, or adjust the execution. The governing variables stay untouched. You’re still operating within the same framework, just picking a new route to the same destination. Think of it as fixing the map when you get lost.
In double-loop learning, you go further. Before changing your action, you examine the governing variables themselves. You question whether your assumptions are correct, whether your goals are appropriate, and whether the rules you’ve been following still make sense. This is less like fixing the map and more like asking whether you’re headed to the right destination at all.
A useful shorthand comes from framing the two approaches as questions. Single-loop learning asks: “Are we doing things right?” Double-loop learning asks: “Are we doing the right things?”
A Concrete Example
Consider Netflix in its early days as a DVD-by-mail company. A single-loop approach to improving the business would involve optimizing shipping logistics, negotiating better postal rates, and refining the envelope design. All of those actions accept the same underlying assumption: customers want DVDs delivered to their homes.
The double-loop question was different: “Do people actually want DVDs, or do they just want easy access to movies?” That reframing challenged the governing variable, the belief that the physical disc was the product. The answer led to streaming, and eventually to original content production. The company didn’t just improve its existing strategy; it replaced the assumptions the strategy was built on.
Single-Loop vs. Double-Loop in Practice
The difference shows up clearly in how organizations handle errors. Research on healthcare safety culture illustrates the contrast well. A single-loop response to medical errors involves correcting individual operational mistakes without significantly changing the overall safety culture. A double-loop response means questioning the governing variables of that culture: the mindset around errors, the hierarchical structures that discourage reporting, and the assumptions about how safety is maintained.
In day-to-day business, single-loop learning looks like employees following directives from senior management, monitoring progress toward set targets, and adjusting their methods when they fall short. It’s a do, check, and adjust cycle. Double-loop learning looks like employees identifying barriers that originate in the organizational structure itself, questioning whether the performance metrics they’ve been given are the right ones, or pushing back on policies that no longer serve their intended purpose.
Both types of learning are necessary. Single-loop learning keeps operations running efficiently. It aligns activities with strategy and catches execution errors. But it can only improve performance within the boundaries of existing assumptions. When those assumptions are wrong, or when the environment has changed enough to make them obsolete, no amount of tactical adjustment will solve the problem.
Why Double-Loop Learning Is Hard
Argyris spent much of his career studying why people and organizations resist double-loop learning, even when they need it most. The answer lies in what he called “defensive routines,” habitual ways of thinking and behaving that protect us from perceived threats but also block genuine learning.
Argyris identified a default operating mode he called Model I, which most people use without realizing it. Model I is driven by four governing variables: stay in unilateral control of situations, maximize winning and minimize losing, avoid generating or expressing negative feelings, and appear as rational as possible. The action strategies that follow from these variables are predictable: manage the environment on your own terms, own and control the task, and protect yourself and others from uncomfortable truths.
The problem is that Model I behavior makes double-loop learning nearly impossible. If your priority is to stay in control and avoid looking wrong, you will not question your own assumptions. You won’t invite others to challenge them either. Argyris argued that people are often unaware of the extent to which their own actions inhibit learning in group settings. They believe they’re being open-minded while their behavior signals the opposite.
Moving Toward Model II Thinking
The alternative Argyris proposed is Model II, a set of governing variables designed to make double-loop learning possible. Model II is built on three principles: use valid information (seek out data that could prove you wrong, not just data that confirms your position), promote free and informed choice (let people make decisions based on accurate information rather than manipulation or control), and take personal responsibility with constant monitoring of your own implementation.
In practical terms, this means creating environments where people can surface uncomfortable questions without penalty. It means treating a mismatch between expected and actual results as a signal to examine your assumptions, not just your tactics. It means asking “why did we think this would work?” before jumping to “what else should we try?”
Organizations that study their own performance metrics illustrate this well. A single-loop approach treats key performance indicators as fixed targets and focuses on aligning activity to hit them. A double-loop approach periodically examines whether the KPIs themselves are measuring the right things, and changes them when they no longer reflect what actually matters.
Where Triple-Loop Learning Fits
Some frameworks extend the concept further into a third loop. If single-loop learning asks “are we doing things right?” and double-loop learning asks “are we doing the right things?”, triple-loop learning asks “how do we decide what’s right?” It moves beyond questioning specific assumptions to examining the deeper values, purpose, and decision-making processes that shape those assumptions in the first place.
Triple-loop learning gets to the core of organizational identity: why do we have these systems, processes, and goals at all? It’s less commonly discussed and harder to implement, but it builds directly on the foundation that Argyris and Schön established. You can think of the three loops as a progression from adjusting behavior, to rethinking beliefs, to transforming the way you form beliefs in the first place.