Exploration and Exploitation, J. G. March, Organization Science, Vol 2 1 Feb 1991p. 71 87
The vulnerability of exploration
Adaptive processes tend to improve exploitation before exploitation.
The social context of organizational learning
A model of mutual learning
4 Key Features
1. There is an external reality independent of beliefs. Reality has m dimensions, each with value 1 or -1 with independent probability 0.5
2. Beliefs about reality are held by each of n individuals and the organizational code as -1, 0, or 1.
3. In each period in which the organizational code differs from the individual, the individual will change his belief with probability p1 (effectiveness of socialization).
4. But the organizational code can change based on the relative number of people whose beliefs are more reflective of reality than the organizational code. P2 is the effectiveness of learning of the code.
Initial conditions are :
A reality m-tuple (m dimensions, each is 1 or -1), organizational code m-tuple (all initially 0) and n individual m-tuples (m dimensions with values equal to 1, 0, or -1 with equal probability). Thus initially the organizational code is neutral and people have no knowledge.
Thus the individuals nor the code directly experience reality. Both are influenced by each other. Improvement in knowledge comes by the code mimicking the beliefs (including the false beliefs) of superior individuals and by individuals mimicking the code (including its false beliefs).
Basid properties of the model in a closed system
Given m-30, n-50, and simulations-80, One can calculate the proportion of reality in the organizational code and the proportion of reality in the individuals.
Results
Eventually the individuals and code converge, though not always at reality.
Slower socialization leads to greater knowledge at equilibrium than faster socialization. But faster code adaptability leads to greater knowledge when socialization is low and less knowledge when socialization is high. The highest knowledge equilibrium occurs when the code learns rapidly from individuals whose socialization to the code is slow.
Effects of learning rate homogeneity
Having both fast and slow socialization learners improves the overall knowledge equilibrium Knowledge gains are disproportionally due to slow learners, but the are disproportionally realized by the fast learners. It seems that the fraction of slow learners is a significant factor in organizational learning. Since individuals are at a disadvantage to be "slow", it may be hard to maintain an optimal number of slow learners.
Basic properties of the model in a more open system
Effects of personnel turnover
Slow learners stay deviant long enough for the code to learn from them. If P1 is low, knowledge declines with increasing turnover (inadequate exploitation). If p1 is high, then moderate amounts of turnover are beneificial. Introducing more new people increases exploration and improves aggregate knowledge. The gains don't come from the superior knowledge of the new recruit but from their diversity. Old-timers have more knowledge but it is more redundant with the code.
Effects of environmental turbulence
If reality changes from 1 to -1 with probability p4, environmental changes make adaptation essential but learning difficult. As individuals and beliefs converge, change becomes more difficult in the face of exogenous changes. The equilibrium value becomes a random walk of knowledge. However, with turnover higher levels of knowledge is maintainted.
This assumes that you recruit people with some deviance from the organizational code. The balance point optimal for the organization is not necessarily optimal for the individuals. When there is turbulence, there is particular individual advantage to having tenure in an organization that has turnover.
3. Knowledge and Ecologies of Competition
Competition and the importance of relative performance
Consider a case for primacy between an organziation . As the number of competitors increases, the contribution of performance variance to competitive advantage increases.
Learning, knowledge, and competitive advantage
The model suggests that learning that increases the mean and varaiance is beneficial in a competition for primacy. In fact, the relative position is more affected by variance as the number of competitors increases. Some learning processes affect both the average performance and variability A new technology will have higher average performance but higher variance due to inexperience.
4. Little models and old wisdom
Thus a major inefficiency of optimized socialization is that individuals may adjust to an organizational code before the code can learn from them.
The vulnerability of exploration
Adaptive processes tend to improve exploitation before exploitation.
The social context of organizational learning
A model of mutual learning
4 Key Features
1. There is an external reality independent of beliefs. Reality has m dimensions, each with value 1 or -1 with independent probability 0.5
2. Beliefs about reality are held by each of n individuals and the organizational code as -1, 0, or 1.
3. In each period in which the organizational code differs from the individual, the individual will change his belief with probability p1 (effectiveness of socialization).
4. But the organizational code can change based on the relative number of people whose beliefs are more reflective of reality than the organizational code. P2 is the effectiveness of learning of the code.
Initial conditions are :
A reality m-tuple (m dimensions, each is 1 or -1), organizational code m-tuple (all initially 0) and n individual m-tuples (m dimensions with values equal to 1, 0, or -1 with equal probability). Thus initially the organizational code is neutral and people have no knowledge.
Thus the individuals nor the code directly experience reality. Both are influenced by each other. Improvement in knowledge comes by the code mimicking the beliefs (including the false beliefs) of superior individuals and by individuals mimicking the code (including its false beliefs).
Basid properties of the model in a closed system
Given m-30, n-50, and simulations-80, One can calculate the proportion of reality in the organizational code and the proportion of reality in the individuals.
Results
Eventually the individuals and code converge, though not always at reality.
Slower socialization leads to greater knowledge at equilibrium than faster socialization. But faster code adaptability leads to greater knowledge when socialization is low and less knowledge when socialization is high. The highest knowledge equilibrium occurs when the code learns rapidly from individuals whose socialization to the code is slow.
Effects of learning rate homogeneity
Having both fast and slow socialization learners improves the overall knowledge equilibrium Knowledge gains are disproportionally due to slow learners, but the are disproportionally realized by the fast learners. It seems that the fraction of slow learners is a significant factor in organizational learning. Since individuals are at a disadvantage to be "slow", it may be hard to maintain an optimal number of slow learners.
Basic properties of the model in a more open system
Effects of personnel turnover
Slow learners stay deviant long enough for the code to learn from them. If P1 is low, knowledge declines with increasing turnover (inadequate exploitation). If p1 is high, then moderate amounts of turnover are beneificial. Introducing more new people increases exploration and improves aggregate knowledge. The gains don't come from the superior knowledge of the new recruit but from their diversity. Old-timers have more knowledge but it is more redundant with the code.
Effects of environmental turbulence
If reality changes from 1 to -1 with probability p4, environmental changes make adaptation essential but learning difficult. As individuals and beliefs converge, change becomes more difficult in the face of exogenous changes. The equilibrium value becomes a random walk of knowledge. However, with turnover higher levels of knowledge is maintainted.
This assumes that you recruit people with some deviance from the organizational code. The balance point optimal for the organization is not necessarily optimal for the individuals. When there is turbulence, there is particular individual advantage to having tenure in an organization that has turnover.
3. Knowledge and Ecologies of Competition
Competition and the importance of relative performance
Consider a case for primacy between an organziation . As the number of competitors increases, the contribution of performance variance to competitive advantage increases.
Learning, knowledge, and competitive advantage
The model suggests that learning that increases the mean and varaiance is beneficial in a competition for primacy. In fact, the relative position is more affected by variance as the number of competitors increases. Some learning processes affect both the average performance and variability A new technology will have higher average performance but higher variance due to inexperience.
4. Little models and old wisdom
Thus a major inefficiency of optimized socialization is that individuals may adjust to an organizational code before the code can learn from them.
No comments :
Post a Comment
Note: only a member of this blog may post a comment.