Double-loop learning refers to the distinction between learning that keeps a behavioral system operating within a field of constancy and learning that changes what the system seeks to achieve or to keep constant. It is related to the distinction between first-order and second-order change. The emphasis on learning rather than change highlights the processes by which members of the system seek to improve how it functions. Double-loop learning is an important concept for action research because it focuses on what it takes for people and systems to make fundamental changes.
The distinction between single- and double-loop learning comes from the cybernetic theorist W. R. Ashby. Ashby used the example of a thermostat that turns heat on or off to keep the temperature near a set point. This is single-loop learning. When someone changes the setting, the system engages in double-loop learning.
Chris Argyris and Donald Schön introduced this distinction to the domain of leadership and organizational learning. They defined double-loop learning as behavioral learning that changes the governing variables (values, norms, goals) of one’s theory-in-use, the theory of action that can be inferred from behavior. They argued that learning processes and research approaches that may be adequate for single-loop learning are inadequate for double-loop learning. They developed the theory of action approach, also known as action science, to create knowledge that is useful for double-loop learning.
In action science the concept of double-loop learning can be re-cast in terms of the epistemology of practice developed by Donald Schön in The Reflective Practitioner. Schön emphasized the activity of framing by which we make sense of a situation, setting the problem that we will seek to solve. Double-loop learning can be seen as re-framing how we define situations, how we construct our role, and what we take to be desirable outcomes.
Single- and double-loop learning can occur at any level of social analysis, including individuals, interpersonal relationships, groups, and organizations. For example, as an organization grows, it undergoes changes that may require double-loop learning at several levels. It may shift from a traditional hierarchy to a matrix structure, requiring individuals to learn how to surface and manage conflict across boundaries. It may need to shift from a technology-driven, “if we build it, they will buy” approach to a customer-focused approach that takes account of different needs in different regions. Customary work practices must change and the changes must become integrated into the professional identities and working relationships of members of the organization.
Double-loop learning is unsettling, almost by definition. When individuals, groups, and organizations face challenges, they typically respond with single-loop learning. When these attempts do not succeed, the most common responses are more single-loop learning and blaming others or the environment. Few individuals and fewer organizations are good at double-loop learning.
We can distinguish between behavioral double-loop learning and double-loop learning for instrumental, technical, or policy issues. Double-loop learning on technical or policy issues may occur when individuals or small groups have breakthrough insights. Creating a culture conducive to breakthrough insights, however, often requires behavioral double-loop learning. And implementing new policies or strategies may require behavioral double-loop learning.
Behavioral double-loop learning entails changes in values and frames governing how people interact. For example, rather than suppressing or avoiding conflict, people may learn to surface and resolve conflict. Rather than assuming their own or their group’s point of view should prevail and strategizing to make that happen, they may learn to invite other perspectives. Rather than leaving difficult or embarrassing issues unspoken, they may learn to raise them. This kind of double-loop learning increases the learning capability of an organization. It makes it more likely that assumptions underlying current ways of dealing with technical, instrumental, and policy issues will be identified and questioned.
Behavioral double-loop learning requires at least three stages. The first is discovering how current values and frames contribute to ineffective behavior and identifying alternative values and frames that could lead to more effective behavior. The second stage is developing the skill necessary to produce the new behavior in actual situations. This can take considerable practice, as initial attempts to produce the new behavior often result in what Argyris has described as “gimmicks,” with the seemingly new behavior used in the service of the old values and frames. Gimmicks are usually ineffective because other people see them for what they are. For example, recognizing that involving others in a decision process can increase their commitment to implementing the decision, people may attempt to “involve” others in ways that do not give them any actual influence. The third stage in behavioral double-loop learning is to integrate the new behavior, as informed by new values and frames, into group norms and relationships so that it becomes the new normal.
It is possible to achieve some double-loop changes in organizations while bypassing behavioral double-loop learning. One approach is to bring in consultants or to convene a task force that is authorized to circumvent normal practices that keep problems hidden. The limitation of this approach is that it leaves in place the behavioral routines that prevented the organization from correcting the problem earlier and that will likely prevent correcting problems in the future. A second approach is to introduce systems and processes that make visible information that drives action, for example total costs across the organization for developing, producing, selling, delivering, and servicing a product. Implementing these systems often runs into barriers rooted in the behavioral routines that are being left untouched, but it can be an effective way to improve some areas of organizational functioning. A third approach is to bring in new management with a mandate to make sweeping changes. Effective implementation of these changes may require behavioral double-loop learning on the part of organizational members, as when it is necessary to work more interdependently across units.
Argyris, C. (1976). Increasing leadership effectiveness. New York: John Wiley & Sons.
Argyris, C. (1977). “Double loop learning in organizations.” Harvard Business Review, September-October 1977, pp. 115-124.
Argyris, C., Putnam, R., & Smith, D. M. (1985). Action science. San Francisco: Jossey-Bass.
Argyris, C., and Schön, D. A. (1974). Theory in practice: Increasing professional effectiveness. San Francisco: Jossey-Bass.
Argyris, C., and Schön, D. A. (1996). Organizational learning II. Reading, MA: Addison Wesley.