Action Science: A Theory of Action Learning

Rex C. Mitchell, Ph.D., Don Rossmoore, Ph.D.

"Action science" (see, e.g., Argyris, Putnam, Smith, 1985) is a development of Lewinian "action research," focusing on how human beings design and implement action in relation to one another. Individuals design action to achieve intended consequences, and monitor themselves to learn if their actions are effective, albeit they tend to do this without much conscious intent. They make sense of their environment by constructing meanings to which they attend, and these constructions guide action. In monitoring the effectiveness of their action, they also monitor the suitability of their construction of the environment.

This effort is very complex; therefore, individuals construct a simplified representation of the environment and a manageable set of causal theories that prescribe how to achieve their intended consequences. It would be very impractical to construct such representations and theories from scratch in each situation. Rather, people learn a repertoire of concepts, schemas, and strategies - and they learn programs for drawing from their repertoire to design representations and action for specific situations. We speak of such design programs as theories of action.


The form of a theory of action proposition is, "In situation S, to achieve consequence C, do action A. For example, "If I am about to criticize someone, in order to do it without being counterattacked, first criticize myself."

There are two kinds of theories of action. Espoused theories are those that an individual claims to follow. Theories-in-use are those that can be inferred from the individual's actions. There is more to this than the well-known discrepancies between what people say and what they do. The distinction is not between theory and action, but between two different theories of action - those that people espouse and those that they use.

Being aware of both sets of our theories of action can help us make more informed decisions - and also to learn from our decisions and actions. Argyris and others are "centrally concerned with the features of theories of action that promote or inhibit learning in behavioral systems." (Ibid, p.84)


The basic concept can be represented as:
Governing variables ----> Action strategies ----> Consequences

Governing variables are values that actors seek to satisfy. Each governing variable can be thought of as a continuum with a preferred range, e.g.: we don't want anxiety to get too high (making people feel tense), or too low (leading to boredom). We live in a field of many governing variables. Typically it is necessary to trade off among governing variables, because actions that raise the value of one may lower the value of another.

Action strategies are sequences of moves by actors in particular situations to satisfy governing variables (e.g., if anxiety seems too high, inject humor) - with intended consequences (e.g., there will be verbal and body movement signs that the anxiety levels have decreased to move within my preferred range)

When the consequences of an action strategy are as the actor intends, then there is a match between intention and outcome, and the theory-in-use is confirmed. If consequences are unintended (especially if negative), there is a mismatch or error.


When the consequences of an action strategy indicate a mismatch or error, there are two main types of responses:

  1. An individual's first response to error is typically to search for another action strategy that will satisfy the same governing variables... e.g., if the agent wants to suppress conflict (governing variable) and to this end avoids saying anything that might be controversial (action strategy), but others raise threatening issues anyway (outcome mismatch), the agent may try the new strategy of talking volubly about issues on which everyone is likely to agree. In such a case, when new action strategies are used in the service of the same governing variables, we speak of single-loop learning.

  2. Another possibility is to change the governing variables themselves, e.g., rather than suppress conflict, the agent might choose to emphasize and value open inquiry. The associated action strategy might be to initiate constructive discussion of conflictual issues. This is double-loop learning. This involves restructuring of organizational norms, assumptions, and culture to be congruent with a radical change in organizational strategy. It involves fundamental change in the direction, climate, and values within the organization.

Several cues by which double-loop problems can be identified in practice (identification is especially difficult since double-loop problems require dealing with the defenses of human beings):

...all these are problems that require inquiry into governing variables if they are to be solved in such a way that they remain solved.


Argyris and Schön have spent many years studying and working with thousands of individuals in various organizations. They have found that essentially all individuals, including highly successful professionals and leaders, use a pattern of reasoning and action that they designated as Model I. Argyris and Schön recommend a better alternative (Model II), which has different, but not opposite, governing variables (values that individuals seek to satisfy) to those of Model I. When most people are asked to describe their behavior (espoused theory), they describe Model II, not Model I. However, when these same people are observed in action (theory-in-use), they are acting in accordance with Model I, but they are unaware of this discrepancy. The two models are summarized in the following table, which is the latest version developed by me and Don Rossmoore (Mitchell & Rossmoore, 2001) from our experience, but based on the conceptualization of Argyris & Schön (e.g., 1996, p.93 & 118).

Governing Variables - Exercise control unilaterally over goals, decisions, tasks, resources, processes, and projects, as well as other's actions, even thoughts and feelings
- Avoid emotional discomfort and pain
- Be rational
- Promote my interests
- Be "right" and avoid being "wrong"
- Optimize exchange of relevant, valid information
- Optimize free & informed choice, and internal commitment to the choice plus constant monitoring of its implementation
Action Strategies - Exercise control in ways that exclude others from control
- Try to convince others that I am "right" and they are wrong
- Control information and communications to benefit my own interests
- Pursue my interests, even when at the expense of the group's interests
- Discover more productive ways of working together, building trust, improving communication, cooperation, & accountability.
- Optimize accountability & relevant, timely feedback.
- Optimize the exchange of: (a) different relevant opinions & (b) relevant facts, in ways that clearly separate facts & opinions.
Tactics - Confuse facts and opinions in taking in, processing and communicating information
- Discount facts and opinions that are upsetting
- Withhold information that might weaken my positions
- Search for and exploit weaknesses in others' points of view
- Bypass differences and disagreements unless there is an advantage to me
- Avoid or delay dealing with difficult issues
- Blame others, not myself
- Unilaterally "protect" others from being hurt by withholding information, censoring actions, avoiding public discussions
- Modify my actions to avoid experiencing emotional discomfort, and rationalize the modifications
- Do not question my own thinking or decisions b- Rationalize and camouflage my efforts to control and bypass, and pursue self-interests.
- Design situations where participants can originate action and experience high causal influence.
- Have tasks controlled jointly.
- Strive to develop strongest possible commitment of each individual to team and objectives without compromising anyone's appropriate authority.
- Strive to encourage appropriate participation and discussion of all relevant issues in ways thatoptimize the exchange of relevant facts and opinions
- Create norms and rules of good conduct that support timely relevant feedback and optimal accountability
- Establish norms where protection of self and others is shared and oriented toward growth. - When discussing my point of view, share the thinking and facts behind it
- Encourage others to challenge my thinking and facts.
- Strive to understand others' concerns, doubts, and questions about my point of view.
- Help others to articulate the thinking and facts behind their points of view.
- Share my concerns, doubts, and questions about other points of view, articulating my thinking and facts. Strive to do this in ways that minimize defensiveness without compromising clarity and completeness.
- Articulate disagreements, differences in thinking, facts, and dilemmas that must be resolved before more cooperation is possible.
- Jointly search for ways to resolve disagreements, rectify differences in thinking and facts, and resolve dilemmas.
Perception of Others - See others as: self-interested, misinformed, unwilling to reason or cooperate. - See others as: striving to act with integrity, thinking in ways I don't understand, but, if I did, I might learn from.
Consequences for Learning & Effectiveness - We remain unaaware of the differences between our professed intentions and our actions (between our "talk" and our "walk")
- Our maps and perceptions of events and situations are faulty
- We avoid issues, decisions, and actions
- Our efforts to control cause resentment
- Some problems and conflicts remain unsolved
- Our learning produces marginal improvements at best.
- Disconfirmable processes
- Double-loop learning
- Frequent public testing of theories
- Increased long-term effectiveness

The following comments about Models I and II summarize my current viewpoint, which is consonant with Argyris' ideas.

I like to think of these two theories of actions as two contrasting, but not totally different ways of conceptualizing and approaching interpersonal interactions. Model II represents a clearly better, more effective way of acting, in my opinion. I prefer to think of the two models as existing on a continuum, rather than being a discontinuous binary duality - representing distinguishably different frames of reference, with somewhat different underlying values about theories of action. However, both share many common values and assumptions about individuals (which depend somewhat on the values and underlying assumptions of the individual(s) involved), e.g., I believe both models are consistent with the following:

  1. Individuals can and wish to learn and grow
  2. They can and desire to make "good" decisions
  3. They want to and should be involved in decisions that affect them
  4. They are unique, with different abilities, limitations, preferences, and needs
  5. They operate from unique internal frames-of-reference
  6. They need to feel reasonably competent, needed, recognized, accepted, respected
  7. They have only an incomplete and not completely accurate perception of themselves, their actions, the effects and implications of their actions, and how they are perceived by others
  8. They will try to avoid or guard against situations in which they feel embarrassed, inadequate, incompetent, or threatened
  9. They are influenced by reinforcements, more than recognized or admitted

Some ways in which Model II is different from Model I - the latter involves or requires:

  1. Different values underlying the theory-in-use
  2. Among the differences are a greater degree of belief about others that: (a) they don't wish to hurt others, (b) they wish to be honest, and can be trusted, until proven otherwise, and (c) they wish to be reasonable, "fair," even helpful, with others
  3. Explicit and extensive use of double-loop learning, including considerably more skill in thinking beyond the usual, single-loop questions and considerations
  4. Greater recognition of and emphasis on limitations and unknowns in individuals' awareness of their assumptions, implications of their actions, and errors
  5. Dealing explicitly with defensive routines typical of Model I (and to be reduced in Model II)
  6. More inquiry to balance the usual advocacy, with better skills in each
  7. A qualitatively different paradigm regarding what it means to learn, to be helpful to others, to not "hurt" them
  8. More active, engaged interactions
  9. Greater responsibility to make use of the new, greater skills
  10. A different conceptualization of individuals' relationship with others and with organizations

Some implications of advocating and trying to produce more Model II behavior are that it:

  1. Provides a conceptual framework useful in self-development, and in consulting
  2. Is a normative view, which I find to have much value
  3. Offers a way to reduce defensive behaviors, reduce double-binds and dilemmas, take more responsibility for situations and dealing with them, reducing anger, fear, frustration, avoidance, blaming others, withdrawing
  4. Gives better opportunities to discuss key issues (that are undiscussible under Model I)
  5. Reduces errors in several ways: tests/verifies communications and assumptions, makes assumptions, values, beliefs, sets explicit
  6. Provides ways to learn about learning - to learn from our experience rather than just repeat it
  7. Therefore, increases flexibility and adaptability

I like to think of Model II as a desirable path or direction, rather than as a binary contrast with Model I. It is important to avoid a "good/bad" orientation or way of relating to individuals operating far on the Model I side. It is difficult to learn to act consistently on the Model II side, especially when there are time and emotional pressures. It even may not be desirable to operate always far on the Model II side, even if this were possible.


There are some interesting problems with managerial decision making even by very competent, experienced decision-makers. For example, Rossmoore (1989, p.18) reports data on 69 action research projects in which he found (cf. similar observations by Argyris & Schon, 1974, 1996; and Culbert & McDonough, 1980):

  1. Most of the managers, most of the time, took action before they found a strategy alternative that minimally satisfied both their individual and organizational objectives simultaneously.
  2. When they acted before they found such an alternative, they were observed comparing two unsatisfactory alternatives (i.e., they framed their decision as a dilemma): one alternative contained an organizational objective to be achieved only if the individual's objective was not achieved. The other alternative contained an individual objective that could be achieved only if the organizational objective was not achieved.
  3. Systematically, the decision-makers chose the alternative favoring their individual objective over the organizational alternative, and then camouflaged their choice.

Why would capable, motivated executives act in this way - with negative effects on performance? The action science concepts from the last few sections can help us understand the rather discouraging data reported by Rossmoore. Why would capable, motivated executives make decisions in the way described in his data? Rossmoore (1989, p.19) comments:

Several problems exist in producing these conditions: (a) it is difficult, (b) it is counter-intuitive and uncomfortable, (c) it requires the explicit discussion of dilemmas and errors. We tend to hide our dilemmas and withhold information on errors - the very information we need to discover our dilemmas. Two pervasive dilemmas are:

  1. "If I say I don't know, I will be disqualified to lead. If I don't say I don't know, we can not improve our current performance."> The basic underlying assumption: Not knowing automatically disqualifies a leader. Most leaders manage this dilemma by claiming knowledge publicly while also remaining vague and ambiguous in public. Their search for knowledge is conducted privately. Paradoxically, the variance between such public and private talk by a leader bars timely access to the knowledge he/she seeks.
    * A plausible alternative assumption is: Leadership can enhance itself by acknowledging ignorance and producing productive, timely, learning.

  2. "If I tell my subordinates how to do it, they may resist. If I don't they will do it wrong."> The basic underlying assumption: To lead requires unilateral control. Leaders often manage this dilemma by talking around the issues as much as they dare without being direct. Paradoxically, their lack of directness undermines their influence.
    * A plausible alternative assumption is: A leader's influence is enhanced by clarifying each subordinate's directions and discretion.


Each of us has a potent opportunity to improve our decision making and organizational performance through incorporating the following elements into our actions, as much and as often as possible:


| My Consulting Web Site | My University Site Home Page |
* Or press the BACK button on your browser to return to your access point *
Last modified June 4, 2006 Copyright 1988-2006 Rex Mitchell