Several theoretical models of organizational learning curves and related phenomena have been proposed (e.g., Dorroh, Gulledge, & Womer, 1994; Levy, 1965; Muth, 1986; Roberts, 1983). Muth (1986) provided an excellent review of these models and developed a new one. Muth (1986) generated the power law relationship between unit cost and experience [see Eq. (1.1)] by a model that involves a firm searching randomly for lower cost methods from a fixed population of technological improvements. Muth’s model did not aim to explain the variation in learning rates observed across firms.
By contrast, Huberman (2001) developed a theoretical model of organizational learning that aimed at explaining the variation in learning rates observed across firms. In Huberman’s model, the production process is mathematically represented by a connected graph whose nodes represent stages in the production process and whose links represent the routines connecting them. Because there are multiple ways in which products can be assembled, the goal node representing the finished product can be reached along a variety of paths. The total cost of manufacturing a product is proportional to the number of steps (or links in the graph) that are needed to reach the goal. The more steps, the greater the cost of production. The learning process involves finding increasingly shorter paths from the initial state to the final product.
Learning occurs through two mechanisms in Huberman’s model. First, a shortcut or new routine for going from the initial to the final node can be discovered. For example, the organization might discover a shortcut for painting a product that involves fewer procedures.
Second, the organization can improve at selecting routines (or choosing links in the graph). That is, of the many possible links leading from one node to another, the organization improves its ability to select the most efficacious link. For example, members of an organization might learn who is good at what so they know whom to go to for advice or assistance. When an issue arises in a particular domain, members of the organization go to the person with the most expertise in the domain, and thus save considerable time. These two learning mechanisms lead to a shorter path from the initial to final stage.
Huberman’s model generates a power law decrease in the number of steps to assemble a product. The model also generates variation in organizational learning rates: changes in the effectiveness of the procedure for selecting routines lead to dif- ferences in the learning rates. Huberman’s model also produces other empirical regu- larities found to characterize organizational learning such as “organizational forgetting.” Thus, his model is very consistent with observed empirical regularities.
Research is needed to test whether Huberman’s model corresponds to the process by which learning occurs in organizations. The correspondence of his model with known empirical regularities about organizational learning makes it attractive, and it is intuitively appealing. As organizations acquire experience, they reduce the number of steps or shorten the number of links in a production process. For example, an organization might learn that its layout is inefficient and rearrange equipment to minimize the number of steps required to produce a product. Or an organization might learn that its structure is unwieldy and shift to a structure where individuals who interact on a recurring basis are grouped together. Thus, the number of links required for members to communicate is reduced. Alternatively, members of an organization might learn who is expert about particular domains and therefore choose more effective links in the production process. Thus, in this framework, organizational learning involves building faster and more effective connections for getting work done.
Fang (2011) proposed a “credit assignment” model of organizational learning. Similar to Huberman’s model, the learning mechanism in Fang’s model is the gradual recognition of steps to the goal. As organizations gain experience, they assign credit to successive states. Learning occurs through a mechanism of credit propagation: credit from states closer to the goal propagates to states farther from the goal. As the organization’s model becomes more accurate, it reduces the number of steps it takes to reach the final goal and the organization’s performance improves. Fang’s model generates empirical regu- larities observed in the learning literature, including the variation in learning rates. Further, Fang (2011) tested hypotheses generated from her simulation model in experiments involving human participants. Consistent with the credit assignment model, she found that as individuals or dyads gained experience with the task, they developed increasingly fine-grained mental models that initially identified the values of states close to the goal and then identified the values of farther states. Thus, by providing evidence that participants learn through the psychological process of credit assignment, Fang (2011) linked individual (Fu & Anderson, 2006) and organizational learning and thereby, provided a micro foundation for organizational learning.
Source: Argote Linda (2013), Organizational Learning: Creating, Retaining and Transferring Knowledge, Springer; 2nd ed. 2013 edition.