Revitalizing Double-Loop Learning: From Conceptual Foundations to Organizational Transformation

Listen to a review of this article:

Abstract: Double-loop learning (DLL), introduced by Argyris and Schön in 1974, represents one of the most influential yet underutilized frameworks in organizational learning theory. Despite widespread citation, DLL has left a surprisingly superficial impact on management practice and scholarship. This article examines why this conceptual-practical gap persists and proposes pathways for revitalization. Through synthesis of empirical research and theoretical developments, we identify three critical challenges: definitional ambiguity leading to inconsistent conceptualization, methodological limitations in measurement approaches, and contextual barriers to implementation. We argue that DLL's limited impact stems from two interrelated features—its conceptual complexity and implementation difficulty—which have spawned misconceptions that distance current practice from the framework's original intent. By clarifying DLL's dual cognitive-behavioral nature, establishing rigorous measurement criteria grounded in observable data, and integrating contextual factors (task, social, physical) into intervention design, organizations can unlock DLL's transformative potential for systematic problem-solving and sustainable innovation. This revitalization offers actionable insights for practitioners seeking to move beyond surface-level fixes toward fundamental organizational transformation.

In an era of accelerating disruption, organizations face a paradoxical learning challenge. The faster the pace of change, the more critical it becomes to question fundamental assumptions—yet the more difficult such questioning becomes. Leaders find themselves trapped in what Argyris (1999) termed "skilled incompetence": the better they become at their established ways of working, the more resistant they become to examining whether those ways remain appropriate.

This paradox sits at the heart of double-loop learning (DLL), arguably one of the most cited yet least operationalized concepts in organizational scholarship. Since Argyris and Schön introduced the distinction between single-loop and double-loop learning nearly five decades ago, thousands of articles have referenced the framework. Influential management thinkers—from Peter Senge's learning organization to more recent advocates of theory-based learning—have built upon its foundations. Yet as Lipshitz (2000) observed, there remains "an evident gap between the frequency and the profoundness of references to Argyris and Schön's work in the literature."

The Contemporary Relevance of Double-Loop Learning

The stakes for addressing this gap have never been higher. Consider three converging organizational realities:

First, the rise of theory-based approaches to strategy has elevated experimentation and hypothesis-testing as core strategic capabilities (Ehrig & Schmidt, 2022; Felin & Zenger, 2009). Strategists increasingly design experiments to test beliefs about successful courses of action, learning from results to revise those beliefs. This approach mirrors DLL's core logic—yet strategy scholars rarely engage with Argyris and Schön's conceptual apparatus, missing opportunities to leverage decades of accumulated insight.

Second, the innovation imperative requires organizations to "imagine the new in response to the unimagined" (Grandori, 2020, p. 495). Systematic problem-solving that questions root causes rather than applying quick fixes has become essential for generating breakthrough innovations (Iyengar, 2023). Research confirms that DLL enables "out-of-the-box thinking" critical for innovation (Al-Raqadi et al., 2016; Foldy & Douglas Creed, 1999), yet organizations struggle to cultivate such thinking systematically.

Third, the microfoundations of organizational capabilities increasingly recognize that collective learning requires understanding how individual cognition and behavior aggregate into organizational-level outcomes (Barney & Felin, 2013; Furlan et al., 2019). DLL's explicit focus on the individual-organizational interface positions it as a natural bridge for microfoundational research—a potential that remains largely untapped.

Why This Article, Why Now

This article aims to establish what we term a breaking point: a clear demarcation between conceptually muddled applications of DLL and rigorous approaches faithful to its original foundations. Our systematic review of 128 studies published between 1974 and 2021 reveals troubling patterns: definitional drift where DLL loses its dual cognitive-behavioral character; methodological shortcuts that violate Argyris and Schön's fundamental precepts; and implementation approaches that ignore contextual factors critical to success.

Yet this critique is not an end in itself. By clarifying what DLL is (and is not), how it can be measured rigorously, and how it can be generated effectively, we provide a foundation for renewed scholarship and practice. The framework we propose integrates theoretical clarification with empirical insights and practical guidance, offering researchers and practitioners a platform for advancing both knowledge and application.

The article proceeds as follows: We first establish theoretical foundations, distinguishing DLL from related constructs and clarifying its core components. We then present our systematic review methodology and findings across three domains—conceptualization, measurement, and generation. Next, we develop an integrative framework that addresses misconceptions and wrong assumptions while incorporating contextual factors often neglected in implementation efforts. We conclude with a research agenda designed to revitalize DLL scholarship and practice.

The Double-Loop Learning Landscape

Defining Double-Loop Learning in Organizational Contexts

To understand DLL, we must first distinguish individual learning from organizational learning. While individuals possess cognitive systems and can learn, organizations lack brains yet maintain cognitive systems and memories that store learning encoded in routines, structures, and systems (Hedberg, 1981; Levitt & March, 1998). Organizational learning occurs when "any of its units acquires knowledge that it recognizes as potentially useful to the organization" (Huber, 1991, p. 89) and when organizations "encode inferences from history into routines that guide behavior" (Levitt & March, 1998, p. 319).

Argyris and Schön's theory of action provides the conceptual foundation for DLL. At its core lies a critical distinction: all human action is based on theories of action—propositions or causal representations in the form "if I behave in such and such a manner, then the following consequences should occur" (Argyris, 1999, p. 179). Individuals hold two types of theories of action:

Espoused theory represents the theory to which agents claim allegiance—the set of rules they say they follow when asked how they would act in a particular situation.

Theory-in-use represents the theory that can actually be inferred from agents' behavior—the set of rules they actually follow, often unconsciously, when acting in specific situations.

The gap between these two theories creates the conceptual space in which DLL operates. Most people operate from what Argyris and Schön identified as Model I theory-in-use, characterized by governing variables such as unilateral control, winning and minimizing losing, and suppressing negative feelings. Model I produces single-loop learning (SLL): when errors are detected, individuals search for alternative action strategies to achieve the same desired consequences without questioning or altering the underlying values, beliefs, and assumptions (governing variables) that produced the error.

In contrast, Model II theory-in-use—characterized by governing variables such as valid information, free and informed choice, and personal responsibility—enables double-loop learning. DLL occurs when "mismatches are corrected by first examining and altering the governing variables and then the actions" (Argyris, 1999, p. 68). The agent does not merely search for alternative actions to achieve the same ends; she also examines the appropriateness and propriety of her chosen ends themselves.

DLL thus involves two explicit, inseparable components:

1. The cognitive component: Changes in the values, beliefs, or assumptions governing one's theory-in-use (not merely one's espoused theory)

2. The behavioral component: Concrete and observable changes in actions or activities that flow from the cognitive changes

This dual nature is critical. As Argyris (2003, p. 1179) emphasized, the only way to know if one has learned something is "when you can produce in the form of action whatever you claim that you know." Cognitive changes alone do not constitute DLL; behavior or action produced as a result of changing governing variables completes the learning loop.

Prevalence, Barriers, and the State of Practice

Despite decades of attention, genuine DLL remains rare in organizational settings. Empirical studies consistently document this scarcity. For example, Mazur et al. (2012) observed 504 instances of improvement behaviors in hospital settings; only 26 involved DLL behaviors. Stavropoulou et al. (2015) systematically reviewed incident reporting systems in healthcare, finding that 33 of 35 examined studies showed evidence only of SLL (direct improvements to clinical settings) with no strong evidence of DLL-associated cultural change or mindset shifts.

Why does DLL remain so elusive? Research identifies several interrelated barriers:

Defensive reasoning and routines represent the primary obstacle. Model I theory-in-use produces defensive reasoning—a reasoning process that encourages people to "keep private the premises, inferences, and conclusions that shape their behavior and avoid testing them in a truly independent and objective fashion" (Argyris, 1999, p. 131). These defensive routines are "any policy, practice, or action that prevents embarrassment or threat to the players involved, and, at the same time, prevents learning" (Argyris, 1999, p. 166).

Defensive routines operate automatically and pervasively. They are "emotional heritage of the primitive fight-or-flight response meant to protect us in times of danger" (Noonan, 2007, p. 57). Crucially, they are organizational rather than merely individual: individuals with different personalities behave similarly when confronting threat, and new members quickly adopt the same defensive patterns (Argyris, 1999).

Unawareness compounds defensiveness. Individuals typically remain "unaware of the contradiction between their espoused theory and their theory-in-use, between the way they think they are acting and the way they really act" (Argyris, 1999, p. 131). This unawareness operates at multiple levels: people are unaware of their biases, unaware that they are not open to having their ideas challenged, and unaware of what they themselves are doing (Mazutis & Slawinski, 2008; McAvoy & Butler, 2007). Even when awareness emerges, individuals often cannot "change their modes or frames of reasoning as they judge the appropriateness of their attitudes and behaviors based on past knowledge" (Wong, 2005, p. 339).

Leadership gaps create structural impediments. Multiple studies document that interventions fail to produce DLL when they do not engage leaders in key positions who hold the power to stimulate, drive, and facilitate significant organizational changes (Hardless et al., 2005; Huang & Shih, 2011; Vashdi et al., 2007). Without committed leadership that models inquiry and authentic dialogue about taken-for-granted assumptions, DLL interventions produce limited impact.

The cumulative effect of these barriers helps explain DLL's superficial impact despite widespread citation. Organizations reference the concept but struggle to operationalize it; scholars cite the framework but often misapply it; practitioners espouse its value but default to single-loop approaches when confronting difficult problems.

Organizational and Individual Consequences of Inadequate Double-Loop Learning

Organizational Performance Impacts

The absence of DLL capabilities creates measurable performance consequences. Organizations locked into single-loop patterns exhibit what Argyris (2005) characterized as "limited learning systems" marked by defensive interpersonal and group relationships, defensive norms, and reduced production of valid information. These systems minimize freedom of choice, internal commitment, and risk-taking—all essential for innovation and adaptation.

Research documents specific performance deficits:

Innovation stagnation: Organizations that fail to examine underlying assumptions struggle to generate "ideas that are different from the ideas that already exist" (Iyengar, 2023, p. 192). DLL enables what Senge (1990) termed "generative learning"—learning that enhances capacity to create rather than merely adapt. Without this capability, organizations become trapped in incremental improvement, applying known solutions to novel problems that demand fundamental rethinking.

Problem recurrence: Organizations employing single-loop approaches "fix problems" through first-order solutions without "diagnosing and altering underlying causes to prevent recurrence" (Tucker et al., 2002, p. 124). This creates what Kululanga et al. (1999) describe as discrete rather than thorough solutions, ensuring the same problems resurface repeatedly. The costs compound over time: resources consumed addressing symptoms; morale eroded by frustration; credibility damaged as stakeholders lose confidence in management's problem-solving ability.

Strategic drift: In the absence of periodic examination of governing assumptions, strategies become increasingly misaligned with environmental realities. What Argyris and Schön (1978) termed "Model O-I" organizational systems produce primary inhibiting loops when confronting embarrassing and threatening problems. Rather than questioning whether the strategy remains appropriate, organizational members distance themselves from responsibility, suppress negative feelings, and engage in bypass and cover-up behaviors that prevent the organization from adapting.

Resource misallocation: Organizations operating primarily in single-loop mode tend to over-invest in efficiency improvements while under-investing in effectiveness questioning. They optimize the performance of outdated strategies rather than allocating resources to explore whether fundamentally different approaches might better serve evolving organizational purposes.

Quantifying these impacts remains challenging given measurement difficulties discussed subsequently. However, the pattern across multiple studies is consistent: organizations that develop DLL capabilities demonstrate superior long-run effectiveness in problem-solving, greater innovation capacity, and enhanced ability to adapt to discontinuous change.

Individual Wellbeing and Stakeholder Impacts

The consequences of inadequate DLL extend beyond organizational performance metrics to affect individual wellbeing and stakeholder experience. The defensive routines that inhibit DLL create psychologically unsafe environments where individuals feel unable to question assumptions, challenge established practices, or admit uncertainty.

Psychological consequences for organizational members: Working in Model O-I systems characterized by defensive norms produces specific psychological strains. Individuals experience the stress of maintaining contradictions between what they espouse and what they do, the frustration of repeatedly encountering preventable problems, and the demotivation that comes from perceiving that genuine learning is neither valued nor possible. As Argyris (1999) documented, these environments are marked by low freedom of choice and minimal internal commitment—conditions antithetical to psychological wellbeing.

The inability to engage in authentic dialogue about assumptions and beliefs forces organizational members into what Thornhill and Amit (2003) observed: people know they should not use defensive reasoning to deal with work difficulties, yet they still employ such reasoning to avoid losing control and dealing with embarrassment (Mordaunt, 2006). This gap between knowing and doing creates cognitive dissonance that erodes psychological health.

Stakeholder and beneficiary impacts: The absence of DLL capabilities affects not only internal members but also external stakeholders who depend on organizational services. In healthcare settings, for instance, the failure of incident reporting systems to enable DLL means that systemic causes of patient safety incidents remain unaddressed (Stavropoulou et al., 2015). Patients continue experiencing preventable harm because organizations fix specific incidents (SLL) without examining the assumptions and structures that allow such incidents to occur.

Similar patterns emerge across sectors. In public services, citizens receive lower-quality outcomes when agencies cannot question whether their program theories accurately reflect citizen needs and environmental realities. In commercial contexts, customers encounter repeated frustrations when companies fix individual complaints without examining whether underlying product design or service delivery assumptions require revision.

Leadership development deficits: Perhaps most consequentially, organizations that cannot practice DLL fail to develop leaders capable of navigating complexity and ambiguity. Leadership development in Model O-I systems emphasizes execution skills rather than inquiry capabilities. Rising leaders learn to deliver results within existing frameworks but not to question whether those frameworks remain appropriate—a critical capability as individuals assume broader responsibilities requiring strategic judgment.

Evidence-Based Organizational Responses

Table 1: Double-Loop Learning Implementation Strategies and Examples

Organization or Context	Intervention Category	Learning Mechanism or Tool	Governing Assumptions Addressed	Reported Outcomes	Evidence Type (Inferred)
Microsoft	Leadership Modeling	Learn-it-all orientation / Inquiry sessions	Defensive culture of "knowing it all"; strategic assumptions about cloud and mobile	Successful strategic pivot and cultural transformation	Specific empirical case study
NASA Jet Propulsion Laboratory	Rapid Learning Protocols	Briefing-debriefing / Assumption examination	Organizational and cognitive assumptions that allowed flawed technical decisions	Improved project success rates; increased willingness to surface uncertainties	Specific empirical case study
Toyota	President's Diagnosis	Strategic assumption examination	Single-source supplier proximity vs. supply chain resilience	Strategic changes in supply chain design; increased resilience	Specific empirical case study
Virginia Mason Medical Center	Virginia Mason Production System	Assumption audits	Assumptions regarding physician knowledge of medication lists	Redesigned care coordination systems; improved patient safety	Specific empirical case study
Shell Oil	Scenario Planning	System dynamics / Simulation	Energy demand, geopolitical stability, and regulatory evolution	Faster adaptation to energy cost declines compared to competitors	Specific empirical case study
Buurtzorg Netherlands	Organizational Design	Circular organizing / Autonomous neighborhood teams	Traditional hierarchical management vs. patient-needs efficiency	Higher care quality at lower cost; continuous protocol revision	Specific empirical case study
U.S. Army	Systematic Lesson-Learning	After Action Reviews (AAR)	Doctrinal assumptions and participant beliefs leading to problems	Enhanced adaptation capabilities; truth-telling regardless of rank	Specific empirical case study
Pixar	Post-project Reviews	Pixar Post-Mortems	Storytelling, technology, and collaboration assumptions	Maintained creative vitality across decades and transitions	Specific empirical case study
Interpublic Group	Action Science	Immunity-to-change mapping / Left-hand column	Assumptions about client embarrassment and creative risk	Enabled more innovative work; improved client relationships	Specific empirical case study
Healthcare / Hospital Settings	Incident Reporting Review	Systematic literature review	Systemic causes of patient safety vs. surface fixes	Findings showed 33 of 35 studies lacked evidence of mindset shifts (Single-Loop Learning only)	Systematic review finding
General Leadership Development	Double-loop coaching	3R cycles (Reflect, Reframe, Redesign)	Leaders' cognition and mental models shaping behavior	Concrete and observable changes in actions resulting from cognitive changes	Theoretical framework
Software Organizations / Continuous Improvement	Integrated Double Kaizen Loop (IDKL)	5 Whys probing governing beliefs	Validity of underlying improvement targets and paradigms	Culture of continuous learning and sustainable improvements	Theoretical framework

Cultivating Leadership Commitment to Reflection and Authentic Dialogue

The most frequently documented barrier to DLL—lack of committed leadership engaging in critical reflection and authentic dialogue—suggests a primary intervention point. Leaders must develop both the capability and the will to examine their own assumptions before they can foster DLL more broadly.

Multiple intervention studies demonstrate that DLL succeeds or fails based primarily on whether leaders in key positions model inquiry, participate authentically in examining assumptions, and commit organizational resources to support learning processes. Conversely, interventions that engage only mid-level participants while leaving leadership untouched consistently produce limited results (Hardless et al., 2005; Vashdi et al., 2007; Witherspoon, 2014).

Effective approaches for developing reflective leadership:

• Double-loop coaching mechanisms that help leaders recognize how their thinking shapes their actions and results, using structured "3R" cycles of reflect, reframe, redesign to coach on leaders' cognition rather than merely their overt behavior (Witherspoon, 2014)

• Collaborative Developmental Action Inquiry (CDAI) methods that engage leadership teams in examining their collective frames of reference, testing whether their shared assumptions remain valid, and redesigning approaches based on new insights (Kwon & Nicolaides, 2017)

• Reflexive organizational learning protocols adapted from military briefing-debriefing models, where leadership teams systematically examine not only what happened during an initiative but also what assumptions embedded in the system led to unexpected outcomes (Vashdi et al., 2007)

• Problem-based learning interventions combining interactive multimedia, experiential learning, and role-playing to facilitate experience sharing, discussion, and critical reflection among leadership cohorts (Hardless et al., 2005)

Microsoft: When Satya Nadella became CEO in 2014, he explicitly confronted Microsoft's defensive culture of "knowing it all" by modeling what he termed a "learn-it-all" orientation. He began executive meetings by sharing his own uncertainties and assumptions, explicitly inviting challenges. Leadership team meetings shifted from report-outs to genuine inquiry sessions examining whether the company's strategic assumptions about cloud computing, mobile devices, and platform ecosystems remained valid. This leadership modeling cascaded through the organization, enabling Microsoft's successful strategic pivot and cultural transformation.

NASA Jet Propulsion Laboratory: Following high-profile mission failures, JPL leadership implemented "rapid learning" protocols where project leaders examine not only technical failures but the organizational and cognitive assumptions that allowed flawed decisions to proceed. Leadership participation proved critical: when senior leaders authentically shared their own assumption failures and modeled non-defensive inquiry, team-level DLL increased measurably. Project success rates improved as teams became more willing to surface inconvenient uncertainties early rather than defensively minimizing concerns.

Integrating Double-Loop Learning into Continuous Improvement Systems

Organizations with mature continuous improvement programs possess infrastructure—regular problem-solving rhythms, improvement tools, measurement systems—that can be leveraged for DLL if appropriately adapted. The challenge is ensuring these systems facilitate assumption examination rather than merely optimizing existing approaches.

Studies demonstrate that continuous improvement tools and methods can generate DLL, but success is neither automatic nor common. The critical factor is whether improvement protocols require participants to examine root causes at the assumption level, not merely the procedural or technical level. Organizations achieving high maturity in continuous improvement show greater propensity for DLL (Stålberg & Fundin, 2018), but only 5% of improvement behaviors in less mature systems involve DLL (Mazur et al., 2012).

Effective continuous improvement approaches that enable DLL:

• Integrated Double Kaizen Loop (IDKL) models that combine daily small improvements with periodic deep dives into whether the assumptions underlying improvement targets remain valid, using "5 whys" protocols that explicitly probe governing beliefs rather than stopping at procedural causes (Al-Baik & Miller, 2019)

• DMAIC-DLL frameworks that integrate Define-Measure-Analyze-Improve-Control cycles with collaborative and experiential learning at each stage, ensuring teams question not only how to achieve objectives more efficiently but whether the objectives themselves remain appropriate (Kolawole et al., 2021)

• Plan-Do-Check-Act protocols explicitly designed to generate five types of improvement behavior—three related to SLL (Quick Fixing, Conforming, Expediting) and two to DLL (Initiating behaviors that challenge existing paradigms and Enhancing behaviors that build new capabilities based on revised assumptions) (Mazur et al., 2012)

Toyota Production System Evolution: While TPS is often portrayed as exemplifying continuous improvement, less recognized is how Toyota periodically engages in assumption examination. The company conducts "president's diagnosis" sessions where senior leaders examine whether fundamental operating assumptions remain valid. Following the 2011 Japan earthquake and tsunami, Toyota questioned its long-held assumption that just-in-time delivery required single-source suppliers located near assembly plants. This DLL process led to strategic changes in supply chain design, increasing resilience without abandoning just-in-time principles.

Virginia Mason Medical Center: The hospital integrated DLL into its "Virginia Mason Production System" by adding "assumption audits" to standard improvement activities. When addressing patient safety issues, teams now explicitly identify and test governing assumptions (e.g., "physicians always know patients' full medication lists"). This integration has enabled the hospital to move beyond fixing individual safety incidents toward redesigning systems based on more realistic assumptions about information availability and care coordination.

Leveraging Technological Tools to Surface and Test Assumptions

Technology offers distinctive capabilities for DLL: simulation enables experimentation with mental models; machine learning surfaces patterns that challenge existing beliefs; digital platforms capture and make visible the assumptions embedded in decisions.

Technological tools facilitate DLL primarily by helping participants see their mental models and assumptions, compare them with empirical data, and experiment with revised models in low-stakes environments. The key is not the technology itself but how it is designed to promote reflection and assumption-testing (Bohanec et al., 2017; Kim et al., 2013; Tsuchiya, 1998).

Technological approaches that support DLL:

• System dynamics modeling where teams build explicit causal maps representing their mental models, then use simulation to test whether those models generate observed organizational behavior—discrepancies prompt questioning of mapped assumptions (Kim et al., 2013; Sterman, 1994)

• Machine learning models for decision support that make explicit the assumptions embedded in historical decision patterns, allowing teams to examine whether those assumptions remain valid and experiment with different scenarios based on revised assumptions (Bohanec et al., 2017)

• Business process simulation for reengineering that enables teams to articulate the assumptions underlying current process designs, then test redesigns based on alternative assumptions before implementing changes (Tsuchiya, 1998)

• Digital twin technologies that create virtual representations of operations, allowing managers to test "what if" scenarios that challenge core operating assumptions without disrupting actual operations

Shell Oil Scenario Planning: Shell pioneered using scenarios not merely for strategic planning but as a technology for surfacing and challenging assumptions. By building detailed simulations of alternative futures based on different assumption sets (e.g., about energy demand, geopolitical stability, regulatory evolution), leadership teams identify which current strategic assumptions are most vulnerable. This process has enabled Shell to adapt strategy more quickly than competitors when major assumption violations occur (e.g., rapid renewable energy cost declines).

Amazon Machine Learning for Demand Forecasting: Amazon uses machine learning models not only to improve forecast accuracy but to identify when human forecasters' assumptions (embedded in overrides to algorithmic predictions) prove systematically wrong. By surfacing these assumption failures and prompting examination of the beliefs driving them, Amazon helps category managers engage in DLL about demand patterns, seasonality, and cross-elasticities. This has improved both forecast accuracy and forecasters' mental models.

Designing Organizational Structures and Operating Contexts for Learning

Organizational designs and operating contexts can either facilitate or impede DLL. Structures that enable open inquiry, consent-based decision-making, and multi-directional communication create fertile ground for assumption examination; hierarchical structures that concentrate authority and limit information flow impede DLL.

Specific organizational design features—particularly those that enable free inquiry, distribute decision-making, and create multiple feedback channels—correlate with higher DLL capability. The most effective designs combine structural elements with governance processes that encourage questioning (Lewis & Moultrie, 2005; Romme & Van Witteloostuijin, 1999).

Structural and contextual enablers of DLL:

• Circular organization designs featuring organizing into circles (autonomous teams), consent-based decision-making that requires open debate, double-linking between circles that promotes both upward and downward communication, and consent-based election of representatives (Romme & Van Witteloostuijin, 1999)

• Innovation laboratory spaces providing dedicated facilities and dynamically reconfigurable resources designed to encourage creative behaviors and enable teams to create and enhance organizational routines supporting new value-creating strategies (Lewis & Moultrie, 2005)

• Organic organizational structures characterized by low formalization, high participation in decision-making, and fluid role definitions that enable employees to question whether established procedures remain appropriate rather than merely executing them (Sitar & Škerlava ˇ, 2018)

• Psychologically safe team contexts where members can take interpersonal risks (admitting uncertainty, challenging others' views) without fear of embarrassment or retribution—a necessary precondition for the authentic dialogue required for DLL (Edmondson, 1999)

Buurtzorg Netherlands Home Care: The organization eliminated traditional hierarchical structures, organizing instead into autonomous neighborhood teams of 10-12 nurses with no formal hierarchy. Teams make their own decisions through consent processes, consult coaches (not managers) when needed, and regularly examine whether their operating assumptions serve patient needs effectively. This structure has enabled continuous organizational learning: when teams discover that certain care protocols don't work as assumed, they revise the protocols and share learnings across the network. The result is higher care quality at lower cost than traditionally structured home care organizations.

Google's "20% Time" and Project Aristotle: Google's structural allowance for employees to spend 20% of time on self-directed projects created a context for assumption-challenging experimentation. Equally important, the "Project Aristotle" research into team effectiveness explicitly examined Google's assumptions about what makes teams successful. By discovering that psychological safety mattered more than the variables Google had assumed critical (team composition based on individual brilliance), the company engaged in organizational DLL that reshaped management practices.

Creating Systematic Lesson-Learning Processes

Organizations generate vast experience, yet much potential learning is lost because lessons remain tacit, localized, or uncaptured. Systematic processes for extracting, codifying, and disseminating lessons can support DLL—but only if designed to surface assumption failures rather than merely document what happened.

Lesson-learned systems and incident reporting mechanisms are widespread, yet research demonstrates most generate SLL (direct improvements to specific situations) without enabling DLL (examination of underlying assumptions). The critical design choice is whether the system prompts assumption examination and enables organizational responses that address systemic causes (Deverell, 2009; Stavropoulou et al., 2015).

Lesson-learning processes that enable DLL:

• Crisis analysis protocols that require teams to examine not only what happened during the crisis but what organizational beliefs, values, and assumptions created vulnerability to the crisis type, using structured frameworks to distinguish surface causes from deeper governing assumptions (Deverell, 2009)

• Post-project review processes mandating that teams identify and document not only technical lessons but assumption failures—instances where beliefs about scope, risk, stakeholder behavior, or resource requirements proved incorrect—and examine why those beliefs were held (Williams et al., 2021)

• Near-miss analysis systems designed explicitly to examine assumptions rather than merely document events, using protocols that ask "What assumptions would have had to be true for our approach to work perfectly?" and then examining which proved false (Stavropoulou et al., 2015)

• Assumption registries where teams log critical assumptions underlying major decisions, then periodically review which assumptions were validated and which violated, treating violations as opportunities for DLL

U.S. Army After Action Reviews (AAR): The Army institutionalized systematic lesson-learning through AAR protocols used after training exercises and combat operations. Critically, effective AARs go beyond identifying what went wrong to examine why participants held the beliefs and assumptions that led to problems. By creating psychologically safe contexts where rank temporarily matters less than truth-telling, AARs enable examination of doctrinal assumptions that might otherwise go unquestioned. This system contributed significantly to the Army's adaptation capabilities.

Pixar Post-Mortems: Following each film's completion, Pixar conducts intensive "post-mortem" sessions examining not only what went well or poorly but what assumptions about storytelling, technology, collaboration, or creative process proved accurate or inaccurate. These sessions explicitly surface governing assumptions that leaders and creative teams held, examining why those beliefs seemed reasonable at the time and what evidence eventually contradicted them. By treating assumption failures as learning opportunities rather than occasions for blame, Pixar maintains creative vitality across multiple decades and leadership transitions.

Addressing Defensive Routines through Immunity-to-Change Interventions

Since defensive routines represent the primary barrier to DLL, interventions explicitly designed to help individuals and groups recognize and overcome their defensiveness merit special attention. These approaches combine elements of action science with developmental psychology to surface and address the hidden commitments that sustain defensive behaviors.

Immunity-to-change frameworks grounded in Argyris and Schön's theory of action plus Kegan's developmental psychology show promise for helping individuals recognize the competing commitments and big assumptions that keep them locked in Model I theory-in-use despite espousing Model II values (Bochman & Kroth, 2010). The interventions require skilled facilitation and participants' willingness to examine uncomfortable contradictions.

Interventions addressing defensive routines:

• Immunity-to-change mapping where individuals identify an important change goal, observe their behaviors that work against that goal, uncover the hidden competing commitment those behaviors serve, and examine the big assumption that makes the competing commitment seem necessary (Kegan & Lahey, 2009)

• Left-hand/right-hand column case method where participants write what was said in a difficult conversation (right column) alongside what they were thinking but not saying (left column), then examine the reasoning and assumptions reflected in both columns and the discrepancies between them (Argyris, 1993)

• Public testing of inferences protocols teaching participants to make their reasoning explicit, explain the data and assumptions underlying their conclusions, and actively seek disconfirmation rather than confirmation of their hypotheses (Argyris et al., 1985)

• Team defensive routine mapping where teams collectively identify their characteristic defensive patterns, trace them to underlying collective assumptions, and design experiments to test whether those assumptions remain valid (Clarke, 2006)

Interpublic Group Agency Transformation: When a major advertising agency within Interpublic confronted declining creative effectiveness, leadership engaged consultants trained in action science to help address defensive routines blocking innovation. Through immunity-to-change mapping with creative leaders, the intervention surfaced competing commitments (e.g., desire to innovate vs. commitment to never risk client embarrassment) and big assumptions (e.g., "Clients cannot handle creative work that challenges their categories"). By designing small experiments to test these assumptions—pilots where clients were presented with genuinely boundary-pushing creative—the agency discovered many assumptions were outdated. This DLL enabled more innovative work and improved client relationships.

Building Long-Term Double-Loop Learning Capability

Recalibrating the Psychological Contract Around Learning

Beyond specific interventions, sustaining DLL requires fundamentally reshaping the psychological contract—the implicit beliefs employees and employers hold about their mutual obligations. Traditional contracts emphasize execution (employees deliver results, employers provide compensation); learning-oriented contracts emphasize development (employees continuously question and improve ways of working, employers invest in capability building).

This recalibration addresses a core tension: individuals will not engage authentically in examining assumptions if they fear such examination will be used against them. The implicit contract must shift from "you are responsible for having the right answer" to "you are responsible for questioning whether our current answers remain right."

Creating psychological safety through contracting: Leaders must explicitly address the perceived risks of assumption-challenging. This requires more than declaring "we value learning from failures"—it requires concrete commitments about how assumption failures will be treated, how they differ from competence failures or integrity failures, and what individuals can expect when they surface inconvenient uncertainties. Organizations successfully building DLL capability develop explicit learning charters or covenants that make these commitments tangible (Edmondson, 2019).

Embedding learning in performance systems: Performance evaluation and promotion criteria must reward not only results achieved but quality of learning demonstrated. This means assessing whether individuals (a) examine their own assumptions, (b) seek disconfirming evidence rather than confirmation, (c) change their approaches when assumptions prove invalid, and (d) help others recognize assumption failures. Without alignment between espoused learning values and actual evaluation criteria, employees rationally focus on defensive impression management rather than authentic inquiry.

Modeling vulnerability by leadership: Perhaps most critical, leaders must visibly engage in the vulnerability that DLL requires. When leaders publicly share their own assumption failures, actively invite challenges to their views, and demonstrably change their positions based on evidence, they signal that the organization's espoused learning values are genuine. Conversely, leaders who espouse learning but defensively reject challenges to their assumptions teach employees that the real rules differ from the stated ones.

Developing Distributed Leadership for Inquiry

While committed senior leadership is necessary for DLL, it is not sufficient. Organizations need inquiry capabilities distributed throughout the hierarchy, with individuals at multiple levels capable of recognizing when assumptions warrant examination and skilled in facilitating such examination.

This distributed leadership model differs fundamentally from traditional hierarchical assumptions where strategic thinking is concentrated at the top while execution dominates lower levels. DLL-capable organizations recognize that assumption failures can be spotted anywhere—often frontline workers encounter contradictions between official assumptions and ground-truth realities—and that organizational learning depends on surfacing and examining these contradictions regardless of where they originate.

Developing inquiry skills broadly: This requires investing in developing action science capabilities—particularly skill in making reasoning explicit, testing inferences publicly, and seeking disconfirmation—across organizational levels. Rather than restricting such development to senior leadership programs, DLL-capable organizations build inquiry capabilities into onboarding and ongoing development for all employees.

Creating inquiry roles and rituals: Some organizations formalize distributed inquiry leadership through designated roles (learning officers, improvement facilitators) and rituals (regular assumption audits, quarterly learning forums). These structures legitimize inquiry activities and provide scaffolding for employees less experienced in assumption examination.

Enabling bottom-up strategic insight: Organizations committed to distributed leadership create mechanisms for surfacing strategic insights from frontline sources. This includes reverse mentoring (where junior employees help senior leaders question assumptions about technology, markets, or social trends), skip-level listening sessions (where executives hear directly from frontline workers about assumption failures), and innovation portals (where anyone can propose assumption-challenging experiments).

Cultivating Purpose and Shared Identity Around Learning

Organizations sustain DLL capability most reliably when learning is deeply embedded in organizational identity and purpose. When members define the organization as fundamentally about continuous questioning and improvement—when this is "who we are" rather than merely "something we do"—DLL becomes self-sustaining.

This identity dimension helps explain why some organizations (certain research laboratories, innovation-driven technology companies, military special operations units) maintain DLL capability across leadership transitions while others lose capability when key champions depart. Organizations where learning is identity-central reproduce DLL capability through socialization and selection, continuously reinforcing it through stories, symbols, and rituals.

Articulating purpose that elevates learning: Purpose statements that explicitly incorporate learning—"we exist to continuously improve how we serve patients," "we are committed to challenging conventional thinking about sustainability"—provide legitimacy for assumption-questioning that more static purpose formulations lack. These purpose formulations frame DLL not as an organizational improvement technique but as essential to fulfilling the organization's mission.

Reinforcing identity through stories and symbols: Organizations building learning-centered identity deliberately craft and share stories highlighting assumption-challenging episodes: the time a junior engineer's question led to redesigning a major product, the project where early assumption testing prevented costly late-stage failures, the strategic pivot enabled by questioning conventional industry wisdom. These stories, particularly when they feature protagonists at various organizational levels, teach what the organization values and who "people like us" are.

Selecting for learning orientation: Organizations serious about learning identity increasingly incorporate learning orientation into selection criteria. This means assessing candidates' willingness to question their own assumptions, their response to evidence contradicting their beliefs, and their skill in helping others examine assumptions—not just domain expertise and execution capability.

Conclusion

Double-loop learning stands at a crossroads. Despite five decades of development since Argyris and Schön's foundational work, despite thousands of citations, despite widespread recognition of its importance, DLL has left surprisingly little impact on management practice. Organizations continue defaulting to single-loop fixes for problems that demand assumption examination. Scholars continue citing the framework while drifting from its conceptual foundations. Practitioners continue struggling to generate DLL despite significant investments in learning interventions.

Yet this analysis should inspire neither pessimism about DLL's prospects nor abandonment of the framework. The challenges we have documented—conceptual confusion, methodological limitations, implementation barriers—are addressable. By clarifying what DLL actually entails (cognitive and behavioral changes), establishing rigorous measurement approaches (grounded in observable data), and incorporating contextual factors (task, social, physical) into implementation design, organizations can significantly enhance their DLL capabilities.

The stakes for doing so have never been higher. In an era of accelerating change, where yesterday's strategic assumptions become tomorrow's vulnerabilities, where innovations emerge from recombining knowledge in ways that challenge categorical boundaries, where resilience depends on rapidly learning from unimagined shocks, DLL capability has become essential for organizational survival. Organizations that develop genuine DLL capability—that can systematically question governing assumptions, test revised assumptions experimentally, and institutionalize new approaches when experiments succeed—will outperform those trapped in defensive single-loop patterns.

This article has synthesized evidence and proposed a framework to support that development. We have shown that DLL is neither as mystical nor as impossibly difficult as sometimes portrayed. It is challenging, certainly—changing deeply held assumptions never comes easily, and overcoming defensive routines requires sustained commitment. But challenge is not impossibility. Organizations can build DLL capability by following evidence-based approaches: developing reflective leadership, integrating assumption-questioning into improvement systems, leveraging technology to surface mental models, designing structures that enable inquiry, creating systematic lesson-learning processes, and addressing defensive routines explicitly.

The research agenda we have outlined provides pathways for advancing both scholarship and practice. By conducting rigorous qualitative research grounded in observable data, exploring multilevel mechanisms linking individual to organizational learning, examining contextual factors that enable or impede DLL, and developing validated measurement approaches, researchers can strengthen DLL's theoretical foundations while making the framework more practically applicable.

For practitioners, the message is clear: DLL capability is neither automatic nor impossible—it is developable. Organizations committed to building such capability must make conscious choices about leadership development, organizational design, performance systems, and psychological contracts. They must invest in developing inquiry skills, creating safe contexts for assumption-challenging, and aligning espoused learning values with actual incentives. Most fundamentally, they must embrace the vulnerability that genuine learning requires.

The opportunity before us is to transform DLL from an often-cited but rarely-operationalized concept into a practical organizational capability. This revitalization requires that scholars, consultants, and practitioners commit to conceptual clarity, methodological rigor, and contextual sophistication. By doing so, we can finally realize the transformative potential Argyris and Schön envisioned nearly five decades ago.

Research Infographic

Mastering Double-Loop Learning Slide Deck

References

1. Al-Baik, O. & Miller, J. (2019). Integrative double kaizen loop (IDKL): Towards a culture of continuous learning and sustainable improvements for software organizations. IEEE Transactions on Software Engineering, 45(2), 1189–1210.

2. Al-Raqadi, A. M. S., Abdul, R. A., Masrom, M. & Al-Riyami, B. S. N. (2016). System thinking in single- and double-loop learning on the perceptions of improving ships' repair performance. International Journal of Systems Assurance Engineering and Management, 7(1), 126–142.

3. Argyris, C. (1991). Teaching smart people how to learn. Harvard Business Review, 4(2), 4–14.

4. Argyris, C. (1993). Knowledge for action. Jossey-Bass.

5. Argyris, C. (1999). On organizational learning. Blackwell.

6. Argyris, C. (2003). A life full of learning. Organization Studies, 24(7), 1178–1192.

7. Argyris, C. (2005). Actionable knowledge. In H. Tsoukas & C. Knudsen (Eds.), The Oxford handbook of organizational theory (pp. 423–452). Oxford University Press.

8. Argyris, C., Putnam, R. & Smith, D. (1985). Action science: Concepts, methods, and skills for research and interventions. Jossey-Bass.

9. Argyris, C. & Schön, D. (1974). Theory in practice: Increasing professional effectiveness. Jossey-Bass.

10. Argyris, C. & Schön, D. (1978). Organizational learning: A theory of action perspective. Addison-Wesley.

11. Argyris, C. & Schön, D. (1996). Organizational learning II. Addison-Wesley.

12. Barney, J. & Felin, T. (2013). What are microfoundations? Academy of Management Perspectives, 27(2), 138–155.

13. Bochman, D. J. & Kroth, M. (2010). Immunity to transformational learning and change. The Learning Organization, 17(4), 328–342.

14. Bohanec, M., Robnik-Šikonja, M. & Kljajić, M. (2017). Decision-making framework with double-loop learning through interpretable black-box machine learning models. Industrial Management & Data Systems, 117(7), 1389–1406.

15. Clarke, J. G. I. (2006). Transcending organisational autism in the UN system response to HIV/AIDS in Africa. Kybernetes, 35(1/2), 10–24.

16. Deverell, E. (2009). Crises as learning triggers: Exploring a conceptual framework of crisis-induced learning. Journal of Contingencies & Crisis Management, 17(3), 179–188.

17. Edmondson, A. C. (1999). Psychological safety and learning behavior in work teams. Administrative Science Quarterly, 44(2), 350–383.

18. Edmondson, A. C. (2019). The fearless organization: Creating psychological safety in the workplace for learning, innovation, and growth. Wiley.

19. Ehrig, T. & Schmidt, J. (2022). Theory-based learning and experimentation: How strategists can systematically generate knowledge at the edge between the known and the unknown. Strategic Management Journal, 43(7), 1287–1318.

20. Felin, T. & Zenger, T. (2009). Entrepreneurs as theorists: On the origins of collective beliefs and novel strategies. Strategic Entrepreneurship Journal, 3(2), 127–146.

21. Foldy, E. & Douglas Creed, W. E. (1999). Action learning, fragmentation and interaction of single-, double-, and triple-loop change: A case of gay and lesbian workplace advocacy. The Journal of Applied Behavioral Science, 35(2), 207–227.

22. Furlan, A., Galeazzo, A. & Paggiaro, A. (2019). Organizational and perceived learning in the workplace: A multilevel perspective on employees' problem solving. Organization Science, 30(2), 235–445.

23. Grandori, A. (2020). Black swans and generative resilience. Management and Organization Review, 16(3), 495–501.

24. Hardless, C., Nilsson, M. & Nuldén, U. (2005). Copernicus experiencing a failing project for reflection and learning. Management Learning, 36(2), 181–217.

25. Hedberg, B. (1981). How organizations learn and unlearn? In P. C. Nystrom & W. H. Starbuck (Eds.), Handbook of organizational design (pp. 8–27). Oxford University Press.

26. Huang, Y. C. & Shih, H. C. (2011). A new mode of learning organization. International Journal of Manpower, 32(5/6), 623–644.

27. Huber, G. P. (1991). Organizational learning: The contributing processes and the literatures. Organization Science, 2(1), 88–115.

28. Iyengar, S. (2023). Think bigger: How to innovate. Columbia University Press.

29. Kegan, R. & Lahey, L. L. (2009). Immunity to change: How to overcome it and unlock the potential in yourself and your organization. Harvard Business Press.

30. Kim, H., MacDonald, R. H. & Andersen, D. F. (2013). Simulation and managerial decision making: A double-loop learning framework. Public Administration Review, 73(2), 291–300.

31. Kolawole, O. A., Mishra, J. L. & Hussain, Z. (2021). Addressing food waste and loss in the Nigerian food supply chain: Use of lean six sigma and double-loop learning. Industrial Marketing Management, 93, 235–249.

32. Kululanga, G. K., McCaffer, R., Price, A. D. F. & Edum-Fotwe, F. (1999). Learning mechanisms employed by construction contractors. Journal of Construction Engineering and Management, 125(4), 215–233.

33. Kwon, C. & Nicolaides, A. (2017). Managing diversity through triple-loop learning: A call for paradigm shift. Human Resource Development Review, 16(1), 85–99.

34. Levitt, B. & March, J. G. (1998). Organizational learning. Annual Review of Sociology, 14(1), 319–340.

35. Lewis, M. & Moultrie, J. (2005). The organizational innovation laboratory. Creativity and Innovation Management, 14(1), 73–83.

36. Lipshitz, R. (2000). Chic, mystique, & misconception: Argyris and Schön and the rhetoric of organizational learning. The Journal of Applied Behavioral Science, 36(4), 456–473.

37. Mazur, L. M., McCreery, J. K. & Chen, S. J. (2012). Quality improvement in hospitals: Identifying and understanding behaviors. Journal of Healthcare Engineering, 3(4), 621–648.

38. Mazutis, D. & Slawinski, N. (2008). Leading organizational learning through authentic dialogue. Management Learning, 39(4), 437–456.

39. McAvoy, J. & Butler, T. (2007). The impact of the Abilene paradox on double-loop learning in an agile team. Information and Software Technology, 49(6), 552–563.

40. Mordaunt, J. (2006). The Emperor's new clothes: Why boards and managers find accountability relationships difficult. Public Policy and Administration, 21(3), 120–134.

41. Noonan, W. R. (2007). Discussing the undiscussable: A guide to overcoming defensive routines in the workplace. Jossey-Bass.

42. Romme, A. G. L. & van Witteloostuijin, A. (1999). Circular organizing and triple loop learning. Journal of Organizational Change Management, 12(5), 439–453.

43. Senge, P. (1990). The fifth discipline: The art and practice of the learning organization. Doubleday/Currency.

44. Sisaye, S. & Birnberg, J. G. (2010). Organizational development and transformational learning approaches in process innovations. Review of Accounting and Finance, 9(4), 337–362.

45. Sitar, A. S. & Škerlava ˇ, M. (2018). Learning-structure fit part I: Conceptualizing the relationship between organizational structure and employee learning. The Learning Organization, 25(5), 294–304.

46. Stålberg, L. & Fundin, A. (2018). Lean production integration adaptable to dynamic conditions. Journal of Manufacturing Technology Management, 29(8), 1358–1375.

47. Stavropoulou, C., Doherty, C. & Tosey, P. (2015). How effective are incident-reporting systems for improving patient safety? A systematic literature review. Milbank Quarterly, 93(4), 826–866.

48. Sterman, J. (1994). Learning in and about complex systems. System Dynamics Review, 10(2/3), 291–330.

49. Thornhill, S. & Amit, R. (2003). Learning about failure: Bankruptcy, firm age, and the resource-based view. Organization Science, 14(5), 497–509.

50. Tsuchiya, S. (1998). Reengineering management: A cognitive approach to reengineering. International Transactions in Operational Research, 5(4), 273–283.

51. Tucker, A. L. & Edmondson, A. C. (2003). Why hospitals don't learn from failures: Organizational and psychological dynamics that inhibit system change. California Management Review, 45(2), 55–72.

52. Tucker, A. L., Edmondson, A. C. & Spear, S. (2002). When problem solving prevents organizational learning. Journal of Organizational Change Management, 15(2), 122–137.

53. Vashdi, D. R., Bamberger, P. A., Erez, M. & Weiss-Meilik, A. (2007). Briefing-debriefing: Using a reflexive organizational learning model from the military to enhance the performance of surgical teams. Human Resource Management, 46(1), 115–142.

54. Williams, R. I., Clark, L. A., Clark, W. R. & Raffo, D. M. (2021). Re-examining systematic literature review in management research: Additional benefits and execution protocols. European Management Journal, 39(4), 521–533.

55. Witherspoon, R. (2014). Double-loop coaching for leadership development. The Journal of Applied Behavioral Science, 50(3), 261–283.

56. Wong, M. M. L. (2005). Organizational learning via expatriate managers: Collective myopia as blocking mechanism. Organization Studies, 26(3), 325–350.

---

Jonathan H. Westover, PhD is Chief Research Officer (Nexus Institute for Work and AI); Associate Dean and Director of HR Academic Programs (WGU); Professor, Organizational Leadership (UVU); OD/HR/Leadership Consultant (Human Capital Innovations). Read Jonathan Westover's executive profile here.

---

Suggested Citation: Westover, J. H. (2026). Revitalizing Double-Loop Learning: From Conceptual Foundations to Organizational Transformation. Human Capital Leadership Review, 34(4). doi.org/10.70175/hclreview.2020.34.4.5