First Part
Topic One: Ambiguities of Intelligence
What are the major difficulties in assessing the intelligence of action (and thus of learning)?
1. Defining Intelligence -- does one base it on the favorability of the outcome or on the appropriateness of the decision making procedure? Do organizations ex facto "learn" if their performance improves? Or as in prescriptive literature do orgs have intelligence if they follow the right procedures?
2. Gathering and Using Information to Assess Intelligence
Quote: " Decision makers gather information and do not use it; ask for more and ignore it; make decisions first and look for relevant information afterward, and gather and process information that has little or no direct relevance to decisions. "
3. Intelligence and Outcomes are Nested in Time
What may be deemed intelligent in the short-term may not be intelligent in the long-term. Also the preferences and values used to assess intelligence at the time of the decision may change through action and be different once the outcome is realized. Anticipatory rationalism in this situation is both difficult and dangerous.
4. Decisions Often Made Outside Formal Decision Process
Formal decision processes are often inefficient, and sometimes it's best to buffer symbolism from decision making.
Is there any meaning to organizational intelligence in the face of conflict of interest?
Multiple observers with multiple interests create multiple and often conflicting interpretations of performance. In politics it's almost impossible to assess intelligence because there are always people who claim that the action was not intelligent. How do you weight the opinions of others? Maybe if you defined intelligence as some "majority agreement", but this would potentially limit intelligence by reducing variation.
Intelligence at the local level may not be intelligent at the global level too, and vice versa.
But the system of limited attention does seem to help somewhat, by reducing everyone's ability to focus on all issues. Dividing into departments also separates actions and outcomes from full scrutiny by all interested parties. Inconsistencies are ignored through ignorance.
In the face of effects of action that unfold over time?
What may be deemed intelligent in the short-term may not be intelligent in the long-term. Delayed gratification is difficult. Also the preferences and values used to assess intelligence at the time of the decision may change through action and be different once the outcome is realized. Should intelligence be based on values at decision time or outcome time? Anticipatory rationalism in this situation is both difficult and dangerous.
In these terms, does learning from experience have any advantage (or disadvantage) when compared to rational calculation?
Advantages
1. Less Time-Consuming and More Efficient
* rational calculation is data-intensive
2. Past may be better predictor of the future
3. Allows appropriation of others experiences, knowledge and rules as well
Disadvantages
1. Paucity of Information
Ameliorated by:
* rich histories
* multiple observers and multiple interpretations
* hypothetical histories
2. Redundancy of Information
Past actions tend to be repeated . leading to observations based on the same or similar experiences. They repeat successes. Ways to reduce redundancy are:
* increase measurement noise (leading to arbitrary failures)
* reduce retrival of memory
* inrease rate of aspirations adjusting
3. Ambiguity of Information
It's hard to isolate causal effects from random or extraneous forces. Reducing ambiquity is done by:
* making big changes rather than small ones
* slow the rate of adaptation to experience
4. Strategic Nature of Information
Information is power, and can be wielded to unintelligent ends
Topic Two: Complications in learning
What are the major complications in organizational learning?
Experiential learning involves the processes of variation, selection, and retention. These processes require the capability to:
* experiment
* generate inferences about outcomes
* act on the basis of knowledge
* retain knowledge
Some of major complications in each area are:
1. Capability to experiment
Limited by:
* need to be consistent and accountable.
* rules preventing experimentation
* low expected value of return (vs exploitation)
* balancing exploration and exploitation
2. Capability to Generate Inferences About Outcomes
Limited by:
* biased samples
- undersample rare event
- undersample failures (by decision makers in power)
* tendency to see data as confirming prior actions
* simple causal theories about events
* focus on reliability rather than validity of interpretations
* attribute success to ability and failures to luck
* multiple observations
* success/failure traps
3. Capability to Act on the Basis of Knowledge
Limited by:
* restricted authority
* lack of resources
* limited experience with the knowledge
* knowledge substitutes
* substitute power for learning
4. Capability to Retain Knowledge
Limited by:
* difficulty in translating knowledge into routines
* turnover / low socialization
* competing interests in interpreting history (politics)
* non-neutrality of knowledge
Can they ameliorated? How?
1. Facilitating Experimentation
* provide incentives to experiment
- economic/non-economic rewards
- buffer people from failure
* keep aspirations above performance
* slow learning
- via ignorance
- via optimism/over-estimation of success
- via ideology
- via turnover
2. Facilitating Knowledge
* Increase sample size of experience
- via more patience
- via persistance through optimism
* Improve Validity of Interpretation
- via improvement of experimental design of experience (big changes)
- via multiple observers of same events
- via hypthetical histories
* Improve Reliability of Interpretation
- via traditions of civility in debate
* Improve Memory
- via better routines
* Improve Learning From Others
Topic Three: The Learning Organization
The idea of the "learning organization" is currently relatively fashionable. The idea is that organizations adapt by learning will have a competitive advantage over organizations that do not. To what extent is that a meaningful proposition? To what extent is it a true proposition?
There are various aspects of adaptation via learning that can result in a competitive advantage for organizations:
1. Absorptive Capacity
Organizations that build learning capabilities are better able to appropriate and exploit knowledge from other competitors. Those with the most absorptive capacities have a distinct advantage. Those that fail to build absorptive capacity may never catch up.
2. Adaptive Aspirations
Changing aspirations through learning can stimulate enough failures to maintain exploration and ensure new ideas are potentially available to adapt to new environmental changes.
3. Focus on Aspirations Better Than Focus on Survival
Aspiration focus allows more risk-taking and exploration, leading to the greater possibility of competitive advantage.
4. Success Increases Slack and Risk-Taking
Continued success can cause managers to overestimate the chances for success and thereby take more risks leading to competitive advantage.
Competitive advantage via learning is only part of the total picture. Other perspectives downplay the role of learning in adaptation and competitive advantage:
1. Population Ecology Suggests Organizations Aren't Very Adaptive
From this perspective organizations are rarely adaptive, and will have competitive advantage only as long as their capabilities are favored by the environment. Eventually they will lose their competitive advantage and dissappear.
2. Excessive Learning Reduces Variability and Experimentation
Learning may cause an organization to explore less as they exploit their existing knowledge. Learning can reduce variability and thus affect overall success, especially if primacy and position vs other competitors is important.
3. Fast Learning Can Cause Specialization in Inferior Alternatives
This may initially give a competitive advantage but may be eliminated once the environment changes.
4. Learning often Ignores Long-term Effects
Adaptive learning (due to human cognitive constraints) tends to focus on short-term outcomes and ignores long-term outcomes.
5. Learning Subsitutions May Cause Loss of Competitive Advantage
Organizations often substitute learning in one area for learning in another, which also may cause long-term adaptive difficulties.
Second Part
In their discussion of differential survival, March and Shapira write "...insofar as relative position with respect to resources matters to survival, a focus on an adaptive aspiration level is strongly favored over a focus on survival." Why is this true?
In this situation the zero point becomes rather irrelevant. A survival focus helps in the short term (as some of the high-risk takers are quickly eliminated), but hurts in the long-term. According to their models, focusing on survival usually leads to taking less risks as one nears the survival point. Fewer risk-taking means fewer chances to get a good enough draw from the population to make a major improvement. As a result those that focus on a survival strategy tend to get bunched up near the survival point over time. While they increase their chances of overall survival, in competitions where position is important they have a much lower chance of maintaining a high relative position with such a strategy.
If one focuses on aspiration level, they engage more in risk-taking when they are below the aspiration level, increasing the chance they may rise above it. In effect, they simultaneously increase their chances of a high relative position as well as the chances of being eliminated due to excessive risk-taking and a big loss. As "losers" drop out, the "winners" that remain adjust average aspirations upward over time. If organizations are above the aspiration level, they will also increase risk taking, leading to chances for higher performance. If aspiration levels are tied to other's performance, such a strategy is needed to maintain good standing with an increasing aspiration level (as poor performers drop out, average performance increases).
What are the general implications for the survival advantages of adaptive aspiration levels?
Over time, aspiration levels of the population tend to increase (as poor performers are eliminated). Maintaining an adaptive aspiration level ensures that an organization takes the necessary risks to potentially keep them near or above aspiration levels. As adaptive aspiration levels get far from the survival point, survival becomes less of an issue.
From a population perspective, if the reproduction of a certain type is dependent on amount of resources controlled by that type, adaptive aspiration levels become important to ensure survival of that type. If reproduction is mainly dependent on just numbers, adaptive aspiration levels are at a disadvantage vs survival focus strategies.
Therefore the consequence of emphasizing adaptive aspiration levels depends on the replacement rules.
Third Part
What are the major unresolved issues in the study of learning in organizations? To what extent can they be resolved by empirical research? What is needed?
This is primarily conjecture but:
1. Assessment of Intelligence: Outcome vs Process Based -- Which is Better?
Is intelligence good outcomes or good processes that on average lead to good outcomes?
2. Is the Exploration/Exploitation Balance Point Stable or Unstable? Does it exist?
3. Exploration during Slack: Mechanisms and Processes
What is the transition between exploitation to exploration when you are above the aspiration level? Does it exist?
4. How can an organization tell when they are at a local optima vs a global optima?
5. What are Realistic forms of Opportunity Distributions?
Are they really normal? Does intelligence improve only the right-hand tail or both? Does the left-hand tail really matter?
6. What Kinds of Learning Do Not Reduce Variability?
Empirical Research to Explore Issues:
1. Rich histories and hypothetical histories of events.
2. Better information about actual decision making processes.
3. Multiple actors making same decision (to increase sample size of experience)
4. Long-term outcome data of decisions (not only short-term outcomes).