March, James G., Lee S. Sproull, and Michal Tamuz, "Learning from Samples of One or Fewer", Organization Science, 2 (1991) 1-13.

They explore how organizations convert infrequent events into interpretations of history, and how they balance the need to achieve agreement on interpretations with the need to interpret history correctly.

Organizations often have to learn without having much experience. How does the military learn when it isn't at war? How does an airline learn to avoid fatal accidents when it hasn't had many?

1. Experiencing History Richly

Historical events are unique, and accumulation requires pooling across diverse contexts. Besides pooling, organizations try to maximize the information extracted from incidents.

Experiencing more aspects of experience

For example, evaluating all the collateral consequences of a decision long before the outcome of the decision is known can be useful. When early collateral experiences are positive, organizations, like individuals are prone to exhibit self-reinforcine decision behavior. Organizations often repeat decisions because they made them before (and don't know yet or can't discern the outcomes). Learning is embedded richly in the taking of action, rather than considering ultimate consequences.

Expectations of a particular decision also are learning opportunities and tend to reinforce the behavior (in absence of outcome data). Hoever, in opposite cases where coercion is used, negative expectations may reduce the propensity to repeat.

Decision processes usually lead to optimistic expectations and are vulnerable to disappointment. Short-run lessons are more reinforcine than long-term lessons. Also, actual experience is to be both delayed and more ambiguous than anticipations, thus allowing optimistic expectations about experience to be confirmed by retrospective sense-making of it.

Organizations focus intensely on critical incidents, often learning in surprising ways. To be critical an events that have a place in history and cause branches of development to occur. Critical events change what is believed. The surprise increases its information value. Critical events also have metaphorical power, with strong meaning, interest, and attention to participants. They create stories. They are simple and representative, becoming "vehicles of meaning".

Experiencing More Interpretations

Often the consequences of an action are different in an organization due to differences in interests or culture. People thus learn different lessons from the same experience. There may be multiple "story lines".

There are also efforts to enhance consistency among these lessons via formal proccedings, informal conversations, and diffusion of stories. Interpretations of responsibility are shared. Yet conflicts between groups serve to perpetuate different lessons.

Experience More Preferences

Organizations discover values, aspirations, and indentities while experiencing action consequences. They distinquish between success and failure. By acting, reflecting, and interpreting, organizations learn what they are. By observing their own actions, they learn what they want.

Aspirations change through experience. which loosens the link between performance and outcomes. Definitions of success or failure depend as much on aspiration level as with the outcome itself.

2. Stimulating Experience

Organizations can see history in two ways. One way is that realized events are necessary consequences of previous conditions and decisions. Another way is as a draw from a wealth of possibilities, and one must attend to what didn't occur as well as what did.

Near-histories is one way to described likely but did not happen events. A safety incident could be called and event that could have happened and become an accident. Near-histories, like "near-misses", can be useful learning opportunities in themselves (because there are often many more of them than real events). The near-histories of genius and foolishness are more similiar than their realized histories.

Hypothetical Histories

They are like mental models or simulations with individual learning. Organizations use small samples of specific historical events to construct theories about events, and then construct hypothetical histories with similar interpretive significance. Modern spreadsheets and scenario analysis are similar too.

However, all these above techniques can lead to learning false lessons, superstitious learning, and exaggerated confidence in historical understandings.

3. Assessing and Improving Learning From Small Histories

These processes can both be reliable and valid if used properly. A reliable learning process is one where an organization develops common understanding and shares its interpretations. A valid process is one where the organization can understand, predict, and control its environment.

If individual beliefs converge, shared reality results in high reliability and validity. Yet much of learning results from contrary interpretations. Thus, exploration of new ideas conflicts with maintanence of of stable, shared ones. We need a balance.

The reliability of learning: The construction and sharing of belief.

Stability in shared understanding is important for organizational effectiveness and survival. Buts its hard to sustain shared beliefs. Many different interpretations are both supportable and refutable. The retrival of history from memory exxagerates the consistency of experience. Incorrect predictions are forgotten or ignored. Missing data is remembered as real during recollection.

Often information is gathered to interpret decisions than to inform them. Meetings often are held to share stories than to take action. Organizations crate robust understandings resistant to contradictory information. Social contruction of aspirations tends to be conservative, reinforcing shared behavioral preferences.

Emphasizing critical incidents will improve reliability. People will focus on them. Yet their interpretations may be more based on convenience, cognitive proximity, and political expedience and any sense of validity. Experience is used to strengthen and elaborate previously believed theories of life.

Botha success and failure contain the seeds of change. Persistent success leads to confidence and a willingness to experiment. Persistent failure leads to instability in beliefs and disagreement in preferences and action.

Validity of Learning: The construction of causal belief.

Its hard to make accurate inferences for a few events in history. Events are stochastic (random), requiring understanding of the underlying distributions of potential events. There is also error in interpreting historical events. as well. While some error can be reduce with a more thorough analysis, or by aggregating multiple events.

Every unique historical event is a series of micro-events, each which can be experienced to increase overall richness of the event. However, this quest of richness requires the absorption of detail without molding it. Openess is an important assett.

These micro-events provide "scraps of reality" that help expose the underlying truth about an event.

Pooling the observations of many observers can also help increaase interpretive ability. Adding events or observers is a question of balance.

One can also make efforts to experience the events that did not happen. These near-histories can often be interpreted more easily than realized history. There are less contextual constraints to push interpretion toward a certain way. But constructing hypothetical histories can be expensive. Finally, hypothetical histories can't easily compare to realized histories. Its hard to match the power of real events and the vividness of their stories. These near-stories may also be ambiguous and unpersuasive, or criticized for painting a overly pessimistic picture.

For a low probability of success, near-histories may paint a too optimistic picture. Organizations often unsuccessfully search for new technologies and use near-histories as support for more investment.

Finally, do near-histories emphasize the fact that they almost happened or the fact that they didn't?

4. Four Questions

Organizational efforts to learn from small histories are balanced by feelings of "can do and do so" and "cannot do and do not" learn. We don't have a shared conception on how to do it or how to define critical events.

Beliefs that these techniques will work rest on four fundamental questions:

1. What is the evidential understanding of imagination? Interpreting near-histories is a mix of theoretical and empirical knowledge in ways that are not normal inference theory.

2. What is the proper process for combining prior expectations and interrelated, cummulated aspects of a rich description into an intrepretation of history? How do you decide between good and bad stories?

3. What is the proper trade-off between reliability and validity in historical interpretation? Often organizations will sacrifice validity to gain reliability.

4. What are the relative values of multiple observations of events and multiple interpretations of them?