As we discussed above, researchers have generally taken one of two approaches to knowledge loss. On the one hand, a number of writers have adopted Peters’ approach and point to the importance of unlearning (Hedberg, 1981; Lyles, 1988) in breaking the inertia of past learning in the face of environmental change (Hannan and Freeman, 1984; Miller, 1993, 1994; Miller and Friesen, 1980; Romanelli and Tushman, 1986; Rumelt, 1995). Unlearning from this perspective is understood as a voluntary effort to rid the organization of knowledge that is no longer needed. This argument highlights the fact that organizations must somehow unlearn old routines and practices in order to learn new and more appropriate ways of doing things.
On the other hand, writers have also argued that organizations may forget accidentally; that is, knowledge may be lost without any explicit desire to eliminate the knowledge from the organization. Authors have documented how an organization’s pool of knowledge may dissipate rapidly and unintentionally (Argote, Beckman, and Epple, 1990; Darr et al., 1995; Epple et al., 1991), and how this involuntary forgetting can have serious negative effects on productivity, profitability, and competitiveness (Argote, 1999: 60). We will consider these two views in the following subsections.
Unlearning
In discussions of unlearning, it is often argued that organizations must unlearn old practices in order to allow them to learn new ways of doing things. Learning from this perspective involves not just the creation of new knowledge, but also the active elimination of existing knowledge, especially when the new knowledge collides in some significant way with existing knowledge (Martin de Holan, 2006). From this perspective, unlearning is positive; when old knowledge prevents the organization from adjusting to the new requirements of the environment, unlearning is the solution.
This view of unlearning has largely been inspired by Hedberg (1981). In his seminal work, Hedberg claimed that unlearning was a necessary complement to the notion of organizational learning. He argued that unlearning is distinct from learning, but conceptually necessary to understand how organizations learn because ‘knowledge grows and simultaneously becomes obsolete as reality changes’ (Hedberg, 1981:3). As a result, organizations need to unlearn; that is, to engage in a ‘process through which learners discard knowledge’ (Hedberg, 1981: 18).
At the core of this thinking is the idea that organizational learning necessarily happens in a pre-existing organizational context with a pre-existing stock of knowledge. This ‘canvas’ over which learning happens has been conceptualized as a set (a ‘collection of individual elements’), or as a network where ‘knowledge elements are connected to other elements or not’, creating a variety of connections and dependent relationships that vary in strength and direction (Ahuja and Novelli, 2010). Newstrom (1983) was one of the first to challenge the ‘clean slate fallacy’ (the idea that learning occurred in a void and did not interact in a significant way with existing knowledge). He proposed a typology of learning situations where the significance of unlearning depended on the motivations of the change. According to his typology, new knowledge that aimed to create a new behavior would require a limited amount of unlearning, whereas new knowledge whose objective would be to replace one behavior with another would require a considerable amount of unlearning. He defines unlearning as the ‘process of reducing or eliminating preexisting knowledge or habits that would otherwise represent formidable barriers to learning’ (Newstrom, 1983). From this definition, it is easy to see how managing unlearning could be beneficial for the organization.
Many other researchers have utilized this idea that unlearning is necessary for new learning to occur. For example, Anand and colleagues (1998: 806) noted that there are circumstances when ‘the existing memory may be an obstruction rather than an aid.’ The disruption and recreation of parts of the organization’s memory may therefore be required. Similarly, Crossan and associates argue that ‘the tension between assimilating new learning (feed forward) and using what has already been learned (feedback) arises because the institutionalized learning (what has already been learned) impedes the assimilation of new learning’ (Crossan, Lane, and White, 1999: 533).
While these researchers have generally taken a behavioral learning approach focusing on routines and standard operating procedures, other researchers (e.g. Bettis and Prahalad, 1995; Miller 1990, 1994) have adopted a more cognitive view and argued convincingly that the failure to discard or ‘unlearn’ old dominant logics is one of the main reasons why organizations find it so difficult to adjust their behavior to new environmental conditions, even when they see clear evidence of changes in their environment. Noticing the difficulty that organizations have with diversifications, even related ones, and with rapid and or discontinuous change in a core business, these authors argue that it is not necessarily a problem of routines alone, but of collective representations of the world that make alternative views difficult or unlikely, and that prevent organizational members from either noticing the need for change (Freeman, 1999) or interpreting the changes to understand their consequences, making them blind to stimuli from their environment (Kiesler and Sproull, 1982).
Prahalad and Bettis (1995) are particularly notable in this line of thought as they claim that dominant logics, which are ‘the mental maps developed through experience in the core business,’ are often applied inappropriately in other circumstances. Because these dominant logics represent the shared cognitive map and the strategic mindset of the top management team in an organization, they are closely related to the processes and tools used in the organization and, consequently, with the types of behaviors that can be enacted and with those that cannot. (Bettis, Wong, and Blettner, 2010). Consequently, old dominant logics are one of the most important factors preventing organizations from discarding old knowledge, and a crucial part of organizational knowledge to unlearn when circumstances require, because they are ‘inherently adaptive properties as long as neither the domain of application nor the environment changes significantly’ (Bettis et al., 2010).
From this perspective, dominant logics represent the cognitive view of learning, where learning is seen as a lens that allows the organization and its members to understand in a collective way the environment in which it operates and the adequate responses to that environment. Unlearning is seen as the ability to discard an old logic in order to provide room for a new one: ‘strategic learning and unlearning of the kind involved in the dominant logic are inextricably intertwined’ (Bettis and Prahalad, 1995: 10), ‘before strategic learning . . . can occur, the old logic must in a sense be unlearned by the organization’ (Bettis and Prahalad, 1995). This is so because as organizations grow and become more complex ‘it becomes necessary to establish formal structure, procedures, system, routines and processes (which) are designed in at least rough congruence with the dominant logic’ (Bettis et al., 2010).
The crux of this approach, regardless of whether it is behavioral or cognitive, can be succinctly summarized as follows: ‘firms that can unlearn and reframe their past success programs to fit with changing environmental and situational conditions will have a greater likelihood of survival and adaptation’ (Lyles, 1988: 87). Ultimately, this approach sees unlearning as a fundamental dimension of change, because, as Tsang and Zahra (2008) argue in their exhaustive review, ‘unlearning refers to the discarding of old routines to make way for new ones’ (Tsang and Zahra, 2008). From this perspective, unlearning is best defined as the act of eliminating or discarding knowledge voluntarily, without necessarily the creation of new knowledge, although there is often a close association.
Forgetting
A parallel stream of research has focused on the negative consequences of forgetting. This perspective is concisely summarized by Day (1944: 44): ‘Organizations without practical mechanisms to remember what has worked and why will have to repeat their failures and rediscover their success formulas over and over again,’ wasting resources in the process. In sharp contrast with the unlearning view, researchers here emphasize the importance of not losing knowledge, claiming that avoiding forgetting is critically important in order to maintain performance levels previously reached by the organization.
The notion of knowledge dissipating is particularly at odds with the standard learning curve theory and models, which establish a positive relationship between experience and productivity. Although learning curve studies are generally limited to production settings, the theory has been extrapolated to other dimensions of organizational learning, perhaps excessively (Abernathy and Wayne, 1974). In spite of the robust findings supporting learning curves in operations research (for a detailed review, see Adler and Clark, 1991; Argote and Epple, 1990), involuntary loss of knowledge (forgetting) has been well documented in intermittent production settings (Carlson and Rowe, 1976). Evidence that interruptions (either predictable interruptions, such as a national holiday, or random ones, such as a faulty machine) introduce considerable knowledge loss and consequently reduce learning rates has been documented for over half a century (Hirsch, 1952). This strongly suggests that in situations where changeovers and other interruptions make cumulative production non-continuous, learning arising from experience is followed by forgetting, which is followed by relearning (Carlson and Rowe, 1976). Bailey, for example, studies how interruptions in a process could impact the rate of forgetting and the rate of future relearning (that is, the amount of learning needed to achieve a state similar to the one reached in the past by the same organization) (Bailey, 1989).
These findings could, of course, simply be attributed to the interruption. However, similar results were found in continuous production settings where no interruption or resetting took place. Benkard (2000), for example, found that only sixty-one percent of the ‘stock of experience’ existing at the beginning of the year survived at its end. In other empirical studies available (Argote, 1990; Darr et al., 1995), knowledge retention (conceptually, the complement of knowledge deterioration) was even lower, in the range five to fifteen percent. The third, and most recent paper (Thompson, 2007) found lower, though ‘still significant,’ rates of forgetting, highlighting the difficulty of rigorously assessing forgetting.
Nevertheless, these studies show that the stock of knowledge of the organization diminishes as time passes, because organizational forgetting depletes it in a way that organizational learning cannot compensate. More importantly, they hint that knowledge retention is far from automatic as a naïve view of the learning curve would suggest. Furthermore, discovering how to perform an activity at a certain level of performance and being able to sustain the activity over time are far from the same thing.
In addition, researchers have begun to explore the important strategic ramifications of these observations. For example, Besanko et al. (2010) claim that ‘if learning-by-doing can be “undone” by organizational forgetting, this raises the question whether organizational forgetting is an antidote to market dominance.’ This important observation grows out of the fact that to the extent the market leader has more to forget than market followers, organizational forgetting should affect the market leader more than the followers and therefore work to equalize differences among firms. As a result, organizational forgetting suggests that improvements in competitive positions that grow out of organizational learning will be more transitory than those based in other aspects of the firm.
While work on the causes of involuntary forgetting does not provide a full account, some simulations and experiments have been run focusing on the role of turnover and structural design on rates of learning and forgetting (Carley, 1992; Devadas Rao and Argote, 2006). Carley (1992), for example, theorizes that personnel turnover has an effect on organizational performance because knowledge is lost as personnel leave. Bailey (1992) studies how interruptions in a process could impact the rate of forgetting and the rate of future relearning (that is, the amount of learning needed to achieve a state similar to the one reached in the past by the same organization); and in a similar vein, Argote hypothesizes that knowledge depreciation may happen because ‘products or processes change and render old knowledge obsolete . . . , organizational records are lost or become difficult to access . . . [or] member[s] turnover’ (Argote, 1999: 52–53).
Among the theoretical explanations for the cause of involuntary forgetting, we find the inability to codify knowledge in a way that can be captured by the organizational memory system, as is the case when there is significant individual or collective tacit knowledge involved that has not been made explicit by the organization (Nonaka, 1994). In these cases, organizational learning involves not simply the creation of new knowledge, but also the capacity to crystallize knowledge into routines (Nelson and Winter, 1982).
In that vein, Martin de Holan and associates (2004a), propose that ‘knowledge entering an organization must be introduced in the organization’s memory system, else the organization will rapidly forget the new knowledge.’ They go on to argue that the degree of retention of new knowledge depends on the effort put into integration, hinting that knowledge acquisition and the integration of that knowledge into the organization are two distinct activities that require managerial attention, and probably different tools to manage effectively.
These studies suggest that in addition to unlearning as a positive event that helps the organization adapt to its environment, forgetting is an organizational phenomenon that can also have negative consequences: very much like learning, context matters in order to evaluate the outcomes of forgetting. In addition, available research shows that even in the most formalized of knowledge settings, knowledge retention and learning-by-doing is far from perfect and/or automatic.