Levinthal, Daniel, A. March James G. The Myopia of Learning, Strategic Management Journal Vol 14 95-112 (1993)

Strategic management theory has been built without fully considering the issues of imperfect information and limits to learning. o, 96

Experiential learning has become a plausible mechanism subsituting or augmenting calculative rationality. They somewhat replace enthusiasms for long-range planning and rational calculation as bases for organizational survival.

But experience is often a poor teacher. Many of the same cognitive limits that constrain rationality constrain learning. Inferences, observations, memory, and interpretations are fraught with variability between individuals. Interpretations of history are often political.

Two mechanisms of learning

The first is simplification --- organizations seek to simplify experience and minimize interactions. The second effect is specialization. Learning processes tend to focus attention and narrow experience.

Organizations code outcomes into successes and failures and develop ideas about causes for them. Experience is clouded and shaped by many actors simultaneously learning. Simultaneous learning by several sub-units in a noisy environment can be quite difficult.

Organizations can either generage enough experience to create complex models (usually unfeasible) or inhibit learning in one part of the organization to make learning more effective in another part.

Creating departments is one way to buffer and segregate experience. Another way is to seqentialize attention to divergent goals. Buffers between units and goals allow local consequences to be examined. However, others encouage "tight coupling" instead (JIT, etc.). That way problems in one area become known throughout.

Some difficulties suggest tight coupling to faciliate detection of signals, others suggest loose coupling to faciliate their interpretation. The appropriate balance probably depends on the frequency of errors and the difficulty of diagnosis.

Decomposition and Enactment

As long as problems are decomposable, departmentalization may be effective. But organizations also enact their own environments -- by treating problems as separate they make them separable. Problems that are not seen do not exist.

But restricting the flow of information between subunits restricts knowledge of opportunities and actitivities. Reduction in knowledge leads to a reduction in salience. Subunits simultaneously construct private comprehensible worlds where giveing up mental models becomes diffitult.

Specialization and the principle of learning substitution

A learning system can adapt through sevaral different mechanisms at several different points with approximately the same overall effect. The success of adaptation in one part reduces adaptation pressure in another part. Adjustment in one way tends to inhibit adjustment in another. This produces specialization of learning competence.

Multiple actors- fast and slow learners

Rapid adaptation by one group reduces the need for another. When two drivers are on a collision course, the movement of one relieves movement by the other. Highly adapative parents reduce the need for the child to become adapative.

Multiple mechanisms -- targets, search, and slack

Changes in search, slack, and targets function as substitutes for each other. Each can restore the aspiration/performance equilibrium, but the have different effects on the organization.

But preferences develop along with comptencies. Thus preference change becomes an adaptive substitue for search or change in an actitivy.

Multiple responses: Exit, voice, and loyality

Dissatisfied people can either leave or speak up to try to change the situation. Each is a substitue for each other. But from the organizational viewpoint exit leaves only less demanding participants, which will eventually reduce capabilities. Thus organizations want to slow the exit of quality conscious participants long enough to use their influence in improvement. One way is via loyalty. It slows search and adjustment in order to increase pressure on slack.

Multiple, nested options

Learning experience is nested. It occurs on several but interrelated levels at the same time. The organization learns wheter to pursue a strategy at the same time they learn how to pursue that strategy. Which and how happen at the same time.

In nested learning, learning at one level is a substitue for learning at another. Refining an existing technology may be a substitute for developing another one.

Fast adaptation at one level leads to slow adaptation at another level. If marketing finds new markets for existing products, R&D doesn't have to change quickly. But this often means that learning at the operatinve level subsitutes for learning at higher levels. If customers adapt to inefficiencies, manufacturers are less likely to do so. But in the long run first-order learning can not substitue for second-order learning.

Problems of Myopia

Both specialization and simplification improve performance but also lead to limits on improvements.

Overlooking distant times

Short-run success can lead to long-term failure.

Erosion of enactment

Simplication often captures the central elements of past environments than the contingencies of current circumstances. People become expert at their simple world and fail when the simple world is no longer tenable and irrelevant.

Second-order effects of specialization

Learning that leads to a comparative advantage in one domain may decay adapative capabilities in another domain.

Competency Traps

There is a postive feedback loop between experience and competence. They engage in more competent activities more frequently, increasing experience and thus competence.

Traps of power

Power allows an organization to change their environments rather than learn to adapt to them.. They lose their capabilities to respond to change. If they lose their control, they will be severely disadvantaged.

Knowledge inventories and problem of timing

The time between the anticipation of a problem and its arrival may not be adequate for an organization to identify and develop the knowledge, or acculumulate the experience required to respond effectively. Thus organizations build knowledge inventories to react to potential problems. If there are numerous potential situations, its hard to gain knowledge before it is needed.

Broader or deeper knowledge is less likely to have immediate pay-off but results in a greater ability to adapt to changes.

Overlooking distant places

The best survival strategy for one part of the organization is usually not the best strategy for another part. In fact, the best survival strategies may not be the best for the social structure or the individuals.

Selection among learners

Organizations that become well adapted to their environments will often die when the environments change. But these self-destructive properties of learning make the replacement of obsolescent organizations easier. Also, the fruits of succesful exploration tend to diffuse across other organizations. But the risks and costs of exploration are private goods.

Knowledge diffusion

Knowledge builds capability to detect, evaluate, and exploit other's ideas. But if everyone only exploits without innovating themselves there will be no innovations in the pool to exploit anyway.

Thus at the population level there are increasing returns to investing in learning. But there can also be a downward spiral where less investment brings reliance of knowledge of less value which reduces capability, which in turn reduces knowledge (especially vs competition).

Overlooking failures

Learning is also likely to be misleading if the experiential record on which it is based is a biased representation of past reality. Past successes create an overestimation of future successes. Over-confidence leads to learning from expectations of consequences before they are observed, and they reinterpret results to seem more favorable. On the other hand, lack of self-confidence is self-reinforcing too.

Most people attribute failures to luck and success to ability. Successful people underestimate risks, and unsuccessful people overestimate risks.

The exploitation/exploration balance

Learning leads to dynamics of accelerating exploitation or exploration

The traps of learning

The failure trap

Failing organizations can be driven to frenzies of exploration which drives out exploitation. They abandon new ideas before they can exploit them enough to find their benefits. Aspirations adjust downward more slowly and exhibit a consistent optimistic bias.

The success trap

Past success rewards exploitation over the low potential success of exploration. The trap can be broken by rapid adjustment of aspiration or false feedback about the power of exploration.

Learning and competitive advantage

Learning generally increases average performance and reliability and reduces variability. While higher average performance is a competitive advantage, higher reliability and lower variability is a mixed blessing. In a competitive environment higherconsistency may reduce the chances of success from deviant behavior. (see the promotion model of ability and relability).

We know that while work experience and knowledge of standard beliefs are fair predictors of individual sucess, highly conspicuous sucess in competitive situations is not that related to experience or knowledge. Exploration is on average unfruitful but it is the only way to finish first. Of course, with success exploitation will allow you some gain that will be eroded over time by aggregate learning.

Sustaining exploration

The most common situation is where exploitation drives our exploration. Exploitation generates clearer, earlier, and closer feedback than exploration. It yields more positive returns in the near term. Three processes can help maintain exploration:

The role of incentives

Assignment of property rights to successful searches (e.g., patents, bankruptcy laws, stock offerings,

The role of organizational structure

Failures to recall past lessons, implement past solutions, communicate about current problems all can lead to increased experimentation. Most attempts are unrewarding, but occaisionally one or two "contain the seeds to a new direction".

One can also avoid excessive socialization of new members. This allows the organization to learn from the new members before they are molded.

The role of beliefs

One can also change risk preference or perceived risk.

Influencing risk preference

Risk taking is affected by the ralation between current or expected outcomes and aspirations. Slowly raising aspiration levels can help promote more risky behavior by maintaining slack (in self-referntial situations) and reference to more superior performance increases risk too.

Influencing Perceived risk

Ignorace can also produce more exploratory behavior. Success can foster more risky behavior. It can compensate for the risk aversion that can follow success. It can help induce the self-sacrificing risk taking that serves the organization and larger society. Only the overconfident will ever be heroes.

The role of internal selection

Selection processes in most organizations typically over-sample successful people and under-sample failure. Thus executives often have the illusion of control.

Learning and Strategic Management

Strategic management deals with three grand problems of decision-making

1. Ignorance

2. Conflict

3. Ambguity

The contributions of learning to inteiligence are constrained by three major problems of myopia:

1. Temporal Myopia

One option to relieve myopis is for individuals or sources of capital to move in and out of organization as entrepreneurs (good for them, bad for those left behind).

2. Spatial myopia

3. Failure myopia

Mostly, these myopias lead to less exploration and suggest a conservatism of expectations.