Literature/1975/March

Reviews

 * 2001/Schulz
 * http://www.unc.edu/~healdric/Classes/Soci245/Schulz.pdf (pp. 5-6)

At this point, organizational learning theory conveyed a creative tension between two images of adaptation. On the one hand, learning could be seen as a rational organizational trait, compatible with rationalistic assumptions of economic theories. In this view, learning processes were rational from the perspective of the organization -- they were directed toward performance improvement, and, in the long run, could result in an improved match between organizational arrangements and environmental constraints. On the other hand, learning could be seen as contributing to non-rational. In this view, adaptation processes were regarded as complex and slow and sensitive to small variations in organizational parameters -- characteristics more compatible with notions of limited rationality (Simon, 1955), or even absence of rationality. Each notion continued to play an important role in subsequent work on organizational learning, sometimes as premise, sometimes as a possibility worth exploring, and sometimes as a thesis to be tested.

Strong support for the limited-rational adaptation perspective emerged with the introduction of the concept of "ambiguity". March and Olsen, 1975 noted that the rational adaptation inherent in learning models, including Cyert and March's (1963) learning cycle, is probably unrealistic. Rather, ambiguity prevails -- goals are ambiguous or in conflict, experience can be misleading, and interpretations are problematic. The authors explored four situations in which ambiguity enters the learning cycle (see Figure 1): (1) In "role-constrained learning," individual role definitions prevent that individuals bring their learning to bear on their actions, e.g., when rigid bureaucratic rules inhibit changes of individual behavior. The outcome is inertia. (2) In "superstitious learning" (the term probably originates in Lave and March's (1975) lucid exploration of mathematical models in the social sciences) the connection between organizational action and environmental response is severed. In this situation the organization learns from an apparent environmental response even though it was not caused by the organization. (3) In "audience learning" the connection between individual action and organizational action becomes problematic, e.g., when individuals in a staff unit develop new, powerful solutions which, however, are not implemented due to cultural inertia. (4) The last situation is "learning under ambiguity" -- it is not clear what happened or why it happened. Individuals develop interpretations about causal connections on the basis of insufficient or inaccurate information about the environment, and instead draw on myths, illusions, or ideology. Together, these four possible disconnects suggest that improvement is not a necessary outcome of learning, even though learning is intendedly adaptive. Instead, when ambiguity is present, beliefs, trust, and perceptions determine what happens.

Please insert Figure 1 somewhere here.

Levinthal and March (1981) introduced a comprehensive, thoroughly formalized learning model which incorporated learning under ambiguity. The model focused on search for new technologies. Ambiguity entered the model at two places, first, the effect of an adopted technology on performance was seen as uncertain. Second, adopted technologies were seen as evolving -- they could improve or decay over time. By varying the levels of uncertainty the authors could explore its effects in the model. Simulations of the model revealed that (i) Returns from search depend on the time horizon. (ii) In ambiguous environments, identical organizations will learn to specialize on a search strategy. (iii) Organizational trajectories were characterized by intense path dependence. (iv) Fast learners would adapt quickly to correct signals but they would also adapt quickly to false signals. Conversely, slow learners would not as easily be confused by false signals, but they were also slow to respond to correct signals.