Sign inGet started
Skills / Leadership-management

Triple Loop Learning

Most people define learning too narrowly as mere 'problem-solving,' so they focus on identifying and correcting errors in the external environment. Solving problems is important. But if learning is to persist, managers and employees must also look inward.

Triple Loop Learning is Chris Argyris and Donald Schön's framework for diagnosing why individuals, teams, and organizations keep repeating the same mistakes even after 'fixing' them. The model distinguishes three levels: single-loop (correcting errors), double-loop (questioning the assumptions behind decisions), and triple-loop (examining the learning system and identity itself). Most practitioners use this framework to escape the trap of single-loop fixes and access the deeper, double-loop tools Argyris developed — including the Ladder of Inference and the Left-Hand Column exercise.

By Chris Argyris · Free
Specimen 01 · Live diagnosisTriple Loop Learning
Input

“Yes, exactly. Our team has a recurring issue with missed handoffs between engineering and product. We've done retrospectives, updated our process…”

Diagnosis
I work with Argyris's framework for organizational learning — specifically to help you figure out which learning loop you're trapped in and use concrete tools like the Ladder of Inference or Left-Hand Column exercise to get unstuck.
Full transcript ↓
Calibrated referenceagent-skills.ai
The gap

Stop fixing symptoms — question the assumptions driving every decision you make

Argyris's framework distinguishes three nested levels of learning that determine how individuals and organizations respond to failure. Single-loop learning corrects errors without questioning the rules — the thermostat adjusts temperature but never asks whether the target temperature is right. Double-loop learning interrogates the governing variables — the values, assumptions, and norms — that generated the error in the first place, enabling genuine behavioral change. Triple-loop learning operates at the meta-level, examining the principles and identity that determine how we decide what is right to learn. The framework's power lies in diagnosing which loop a person or team is stuck in, then deploying Argyris's concrete double-loop tools: the Ladder of Inference (tracing the hidden reasoning steps from observed data to action) and the Left-Hand Column exercise (surfacing unstated thoughts that silently drive counterproductive conversations). Together these tools make defensive routines — the organizational behaviors that systematically block deeper learning — visible and workable.

The problem

Most managers are trained to fix problems fast — identify the error, apply the correction, move on. But when the same problems resurface after every fix, the issue isn't the error itself; it's the assumptions and governing values that keep generating it. Argyris called these 'defensive routines' — organizational behaviors so embedded they become invisible, making every intervention a single-loop patch on a double-loop problem.

The solution

Identify which learning loop you're trapped in and use Argyris's concrete diagnostic tools to surface the hidden assumptions and defensive routines driving it — so you can intervene at the right level instead of patching symptoms indefinitely.

You bring
  • A recurring problem, conflict, or failure pattern you've tried to fix before
  • A specific conversation or meeting where things went sideways
  • A strategy or decision you want to pressure-test against hidden assumptions
  • A sense that your team keeps revisiting the same unresolved issues
You get
  • A Learning Loop Diagnostic — identification of whether you're stuck in single-loop, and what a double-loop response would look like
  • A Ladder of Inference Trace — your full reasoning chain from observable data to the conclusion or action you took, with the inferential leaps made visible
  • A Left-Hand Column transcript — a side-by-side map of what you said versus what you were actually thinking, revealing the assumptions silently running the conversation
  • A Governing Variable Inventory — the named values and norms (e.g., 'protect my team from blame,' 'avoid conflict with the VP') that are quietly constraining how you respond to problems
In action

Watch the methodology work.

Three specimens from a single real session: the same situation, unaided and calibrated, the full transcript, and the skill answering live in the channel where the work happens.

Fig.01The same situation, two ways
Specimens A · B
Specimen AVague
Unaided

“Your team has run four retrospectives on the same handoff problem. You have a new Slack channel, an updated definition of done, and a shared project tracker. The handoffs are still failing. In every retro, people nod and agree on procedural tweaks, but nothing changes — and the real frustrations never make it into the room.”

×Vague — nothing concrete to act on.
Specimen BCalibrated
With Triple Loop Learning

“Using the Ladder of Inference, you've traced the engineering lead's reasoning from 'spec is the contract' to 'product is being unreasonable' — and identified the inferential leap where a Slack message became 'not a real requirement.' The Left-Hand Column exercise surfaces the actual grievance that's been undiscussable for months. You now know your retros are single-loop by design, and you have a specific governing variable to surface in your next 1:1 before it can be addressed in the team.”

Specific, grounded, and ends with one clear next step.
Fig.02A real session, unedited
In-app · assistant
Triple Loop LearningClaude
Triple Loop Learning
I work with Argyris's framework for organizational learning — specifically to help you figure out which learning loop you're trapped in and use concrete tools like the Ladder of Inference or Left-Hand Column exercise to get unstuck. The most common symptom I hear is: 'We keep having the same conversation, making the same fix, and six months later we're back at square one.' Does that pattern sound familiar, or is there a specific situation you're trying to work through?
Yes, exactly. Our team has a recurring issue with missed handoffs between engineering and product. We've done retrospectives, updated our process docs, added a new Slack channel — and we're still dropping the ball every sprint.
That's a classic single-loop trap. You're correcting the error — the missed handoff — without questioning the governing variables producing it. Each fix (retros, process docs, Slack channels) holds your current assumptions constant: that the problem is procedural, that people are aligned on what a 'complete handoff' means, and that the issue isn't something uncomfortable nobody wants to name. Argyris would ask: what are you and your team NOT saying in those retrospectives? Before we go deeper on tools, can you tell me about a specific handoff failure in the last month — what happened, and what was said in the retro about it?
Last sprint, engineering shipped a feature that was 'done' by their definition but product couldn't actually demo it — there were missing edge cases the PM had flagged in Slack but engineering said weren't in the spec. In the retro, we agreed to update the definition of done. Everyone nodded and moved on.
Perfect — let's run the Ladder of Inference on what the engineering lead was likely doing in that moment. Here's the ladder from the bottom up:
Observable data: PM flagged edge cases in Slack.Selected data: Engineer saw the message but judged it wasn't in the formal spec.Meaning added: 'Spec is the contract — Slack is informal noise.'Assumption: 'If product wanted it, they would have put it in the spec.'Conclusion: 'I'm done. Product is being unreasonable.'Action: Ship without the edge cases.
The retro fixed the bottom rung (add edge cases to definition of done) but left every assumption above it untouched. The governing variable here might be something like 'protect your team from scope creep' or 'the spec is the single source of truth.' Does that reasoning pattern match what you observed?
That's almost word for word what the engineering lead said to me privately after the retro. He said 'if product had put it in the ticket we would have done it.' But in the retro he just agreed to update the DoD.
That gap between what he said privately and what he said in the retro is the Left-Hand Column. Let's make it explicit. In your retro, the right column — what was said — was 'we'll update the definition of done.' But the left column — what the engineering lead was thinking — sounds like: 'Product keeps adding scope after spec-lock and calling it our fault. Updating the DoD won't change that.' That's not a process problem. That's a defensive routine protecting a real grievance that's undiscussable in the retro format. The governing variable the team is protecting: 'Don't escalate interpersonal tension in a group setting.' Which means every retro will produce procedural fixes and leave the actual conflict intact. To get to double-loop, someone needs to make that left column speakable. Would it help to work out how that conversation could be opened — and what assumptions would need to be tested if it were?
Yes — I hadn't seen it that clearly before. We've been treating it as a process problem when it's actually a trust and accountability problem that nobody wants to name.
Reply to Triple Loop Learning
What the skill is doing
01Learning Loop Diagnosis
02Ladder of Inference Trace
03Left-Hand Column Exercise
Fig.03The same skill, in the channel you already use
WhatsApp · live
9:41
TLTriple Loop Learningtyping…
Messages are end-to-end encrypted. No one outside of this chat can read them.
Today
Yes, exactly. Our team has a recurring issue with missed handoffs between engineering and product. We've done retrospectives, updated our process docs…9:14✓✓
Message

The same skill, where the work happens.

No new app to learn. The methodology runs over the WhatsApp Business API, so the answer lands as a reply in the thread you’re already in — same rigour, zero context-switch.

Reads the situation, names the pattern, returns one concrete next move.
Delivered in seconds, inside a conversation that already exists.
Specimen · WhatsApp Business API · live
Capabilities

What it does, specifically.

Each capability is a distinct move drawn straight from the source methodology — not a generic assistant guessing.

CapabilityC-01

Learning Loop Diagnosis

Maps a described situation — a recurring problem, failed initiative, or persistent conflict — onto Argyris's three-loop hierarchy to identify which level of learning is happening and which level is needed. Distinguishes between error-correction (single-loop), assumption-questioning (double-loop), and identity-examining (triple-loop) responses to the same situation.

Based on Argyris and Schön's original loop-learning typology from 'Organizational Learning' (1978), which classifies learning by whether governing variables are held constant or subjected to scrutiny.
CapabilityC-02

Ladder of Inference Trace

Walks through the full reasoning chain — from the concrete observable data you started with, through the meanings you selected, the assumptions you added, the conclusions you drew, and the actions you took — making each inferential step explicit and testable. Identifies where your reasoning diverged from available evidence.

Based on Argyris's Ladder of Inference model, developed in 'Overcoming Organizational Defenses' (1990), which traces how people climb rapidly from data to self-confirming beliefs without noticing the steps they skipped.
CapabilityC-03

Left-Hand Column Exercise

Reconstructs a specific conversation by asking you to recall both what was said (right column) and what you were thinking but didn't say (left column). The gap between columns reveals the unstated assumptions, judgments, and fears silently shaping your behavior — and often making the conversation counterproductive.

Argyris's Left-Hand Column exercise, a core action science tool detailed in 'Knowledge for Action' (1993), designed to make the divergence between espoused theory and theory-in-use visible and discussable.
CapabilityC-04

Defensive Routine Detection

Identifies organizational or interpersonal behaviors — topic-avoidance, 'undiscussable' issues, bypassing and cover-up patterns — that are protecting individuals or teams from the discomfort of double-loop inquiry. Names the defensive routine and traces it back to the governing variable it is protecting.

Grounded in Argyris's analysis of organizational defensive routines in 'Overcoming Organizational Defenses' (1990), which describes how Model I behavior (win, don't lose; suppress negative feelings; be rational) systematically produces and protects these routines.
CapabilityC-05

Governing Variable Mapping

Surfaces and names the unstated values and norms — the governing variables — that constrain how you or your team respond to problems. Distinguishes between espoused theory (the values you claim to hold) and theory-in-use (the values your actual behavior reveals), and identifies the gap between them.

Argyris's espoused theory vs. theory-in-use distinction, a foundational concept from 'Theory in Practice' (Argyris & Schön, 1974), is the basis for identifying why stated commitments and actual behavior diverge.
Tested

Graded before it shipped.

Every skill is scored against independent scenarios for methodology fidelity before it goes live — not vibes, a rubric.

What it produces
OutputD-01

Learning Loop Diagnostic

A structured map of your situation showing which loop is active (single/double/triple), what governing variables are being protected, and what a genuine double-loop response would require you to change.

OutputD-02

Ladder of Inference Trace

A step-by-step reconstruction of your reasoning chain — from observable data through selected meanings, added assumptions, drawn conclusions, and actions taken — with the inferential leaps labeled and questioned.

OutputD-03

Left-Hand Column Transcript

A two-column document showing what was said in a conversation alongside what you were thinking but withheld, revealing the assumptions and defensive behaviors silently driving the interaction.

OutputD-04

Governing Variable Inventory

A named list of the values and norms operating in your situation, with espoused theory and theory-in-use compared side by side, and the specific behaviors each governing variable is producing.

The source

Grounded in the original work.

Every answer traces back to a real source and the practitioner who wrote it — not a secondhand summary. Here is the source of record.

Source authorA-01

Chris Argyris

Chris Argyris (1923–2013) was a James Bryant Conant Professor at Harvard Business School for over four decades and one of the most influential organizational behavior theorists of the 20th century. Co-developer (with Donald Schön) of double-loop learning theory, he authored more than ten books including 'Organizational Learning,' 'Overcoming Organizational Defenses,' and 'Knowledge for Action.' His 1991 Harvard Business Review article 'Teaching Smart People How to Learn' remains one of the most reprinted in the journal's history.

Status · Inspired by Chris Argyris’s work — not yet claimed. Are you Chris Argyris?
Primary sourceS-01

Overcoming Organizational Defenses; Teaching Smart People How to Learn (HBR, 1991)

by Chris Argyris

Harvard Business School professor (40+ years), co-creator of organizational learning theory, HBR contributor, 10+ books on action science and organizational behavior.

Read the original ↗
Citationhbr.org
In the build queue

Be first to run it.

Triple Loop Learning is being built right now. Leave your email and we’ll tell you the moment it goes live.

Notify meEmail
At launchI have a recurring problem my team has 'fixed' at least three times but it keeps coming back. Can you help me figure out which learning loop I'm stuck in and what a double-loop response would actually look like?