Abstract
Organizational learning is widely seen as a particularly valuable form of change, driven by professionals closest to the work of the agency and all its challenges. However, the growing literature on this process identifies a large and varied set of requisites for learning. The object here is to survey these requisites and show how they are the many guises of a few basic learning processes, and in doing so distinguish the conditions that stimulate or initiate learning from those that support it. Although all of the paths to learning can be encouraged, the stimuli have been less appreciated for their particular role.
Introduction
Organizational learning is widely seen as a particularly valuable form of change, informed by professionals closest to the work of the agency and all its challenges. Virtually all organizations now identify themselves as learning organizations, claiming, often hopefully, that they are open to evidence-based improvements, collaboratively shaped. In contrast, many official public sector reform programs adopt externally designed and imposed change techniques that rely little on the experience of public employees. As more research on learning in public organizations appears, however, an increasing number and variety of conditions, inducements, or prerequisites emerge as necessary or at least important for learning. If these accounts are all accurate, the likelihood of learning is either exceedingly low or administrators must work extraordinarily hard to achieve it.
The object here is to bring together these accounts from the research on public organizations and the even wider theorizing from the private sector to address the question of how to stimulate the learning process and initiate evidence-based change. One of the persistent problems in pinning down how learning occurs in public agencies is that it is not so much a discrete event as a set of ongoing processes with shifting appearances. That is, the thesis examined here is that the various conditions identified as necessary to initiate and maintain learning may not be so much different prerequisites as differences in the ways that core learning processes appear and are implemented in practice. Recognizing the various guises under which learning appears can help encourage this useful indigenous practice. Furthermore, identifying the variety of pathways to learning offers a way to understand the processes which underlie long-term institutional changes (Carpenter, 2001; Ostrom & Basurto, 2011).
The essential idea behind organization learning is that members use some kind of evidence to uncover new connections between actions and the results of their work and then spread this new understanding to others in the organization. Members draw inferences and make connections when they discover previously unrecognized cause–effect relationships or when they come to appreciate the value of practices employed elsewhere (Argote, 2013). But making these connections is not easy, and the literature often focuses on what blocks learning. Members must be looking for connections or at least recognize their desirability when they stumble upon them. Some way of passing this knowledge on to others is necessary if personal lessons are to become organizational, or interorganizational. In some cases, new solutions are created jointly and shared from the start. Learning is portrayed in various ways: as information processing, collaborative dialogue, oppositional dialogue, problem-solving, trial and error search, adaptation, imitation, or knowledge transfer. But across all these accounts, the idea of discovering connections between actions and desired results that are then passed on to other members identifies the endeavor as organizational learning.
The literature on public agency learning is replete with descriptions of learning processes and requirements. Manage-ment research on learning in the private sector has identified an even richer and more diverse range of conditions that contribute to learning (Visser, 2007) so that including them in this assessment is useful. Less research has been published on learning in non-profit organizations (Rashman, Withers, & Hartley, 2009), though the need for learning in that setting is not less.
Bringing together the many conditions and actions that have been identified as contributing to learning begins with a catalog and brief description of each one. Next, we identify the ways in which each one contributes to core learning processes, some to stimulate or initiate learning and others to support it. The role of personal and organizational rewards and disincentives plays no small part in creating the conditions that stimulate learning. From this emerges an organization learning framework that suggests ways to initiate and maintain learning.
Sources of Organizational Learning
The conditions identified in the literature as requisites for learning include the following:
Information about performance.
Access to this information.
Accountability policies that do not punish the acknowledgment of failure.
Cognitive recognition of the risk or severity of dangers.
Crises that reveal problems.
Moderate conflict and tension.
Leader action to explore improvements and undertake change.
Opportunities for transferring lessons across organizations.
Culture values that legitimate the investigation and adoption of alternatives.
The capacity to draw inferences and make connections.
Requisite variety of experiences and results.
The capacity to change policies and practices.
The capacity to preserve useful lessons.
Information About Performance
One of the most commonly agreed-upon requirements for learning is the availability of rich data about results that allows actors to track actions, events, and results (Behn, 2003; Levitt & March, 1988; Moynihan, 2005) to establish, at least roughly, cause and effect. Although theories of organization learning arise in multiple disciplines, management science and sociology tend to emphasize the importance of this requirement by characterizing learning as information processing (Easterby-Smith, Araujo, & Burgoyne, 1999). This is exemplified in Huber’s (1991) four process model: knowledge acquisition, information distribution, information interpretation, and organizational memory. For Levitt and March (1988), learning depends on making inferences from information about an organization’s history to generate new routines. For many others, the information is qualitative, based on stories and reports of past success and failure (Vera, Crossan, & Apaydin, 2011).
Given the importance attributed to information about results for learning, it is not surprising that one of the most heavily investigated potential sources of public organization learning in recent years has been the data collected under the performance movement (Moynihan & Kroll, 2016; Radin, 2006; Taylor, 2014). Learning is a particularly appropriate use for information about unsatisfactory results. Although the movement springs in large part from efforts to increase accountability and provide a business-like bottom line, implicit in the movement is the idea that information about results will lead to agency learning (Kroll, 2015; Newcomer & Caudle, 2011). The evidence for this result is, however, mixed at best (Moynihan & Hawes, 2012; Moynihan & Lavertu, 2012; Newcomer & Olejniczak, 2013).
Although planned and tracked performance information about results is typically what is referred to in public sector research on learning, serendipitous information may hold the best clues to how to make improvements (Kroll, 2013). Everything from “hot gossip” from trusted associates to client feedback delivered ad hoc in face-to-face encounters offers information that can be used to make new connections between actions and outcomes.
Access to This Information
This information must also be available to those who are able to interpret and use it for problem-solving. A dedicated organizational setting for discussing results and possible changes has been proposed as an important condition for learning. Moynihan (2005) and Moynihan and Landuyt (2009) suggest that official forums for communicating information about results and experience and discussing interpretations, or work groups that function that way, encourage the kind of exchanges that lead to learning. These may be after-action reviews, strategic planning sessions, or formal performance assessment gatherings. Jenkins-Smith and Sabatier (1993) note too that professional forums or conferences out of the public eye make it more likely that officials can change their policy positions and adopt the kind of innovations that result from learning.
These observations assume, however, that the forum permits the honest exchange of ideas, reports, and interpretations. A toxic, critical, or belittling atmosphere in a forum makes open exchange unlikely. Team debriefings can become “hot wash-ups” (LaPorte & Consolini, 1991) or “wire brushing” (Goodsell, 2011). A censorious presider or a spirit of disparagement can create a setting that is hostile to revealing shortcomings. “NASA Chicken,” the label for the often-seen reluctance to report problems, is an example (Wald & Schwartz, 2003).
Accountability Policies That Do Not Punish the Acknowledgment of Failure
A related condition affecting learning is an accountability policy that does not impose career-damaging penalties for admitting the existence of problems, errors, or work failures. Instead, of “embracing error” (Korten, 1980) by recognizing unsatisfactory results and digging into their causes, stringent accountability can lead members to ignore problems or blame them on others. “High stakes” accountability that accompanies some versions of performance management (Newcomer & Olejniczak, 2013) in which public embarrassment and budgetary or personnel consequences are threatened (Dubnick, 2005; Mahler & Posner, 2014) represses the serious search for causes of failures that can lead to learning.
Besides fears of career penalties, however, high stakes accountability affects even the cognitive ability to learn. Research on aircraft near-misses shows that higher levels of accountability and the potential for punitive action are clearly associated with lower levels of lesson-drawing and performance improvement over time. Morris and Moore (2000) found that air crews held to stringent accountability were more likely to comfort themselves in incident reports by noting that results could have been even worse rather than speculating about how they might have averted the incident, and thus learning from the event.
The dilemma is how to make the admission of error possible without appearing to condone unsatisfactory results. Weick and others identify the conditions of psychological safely that make learning from errors more likely, including language that emphasizes the accidental character of errors, leaders who permit open discussion of problems, and programs that enable lower level employees to take the lead in identifying and solving problems appropriate to their level of expertise (van Stralen, 2008; Vogus, Sutcliffe, & Weick, 2010; Weick & Sutcliffe, 2006). Similarly, Greiling and Halachmi (2013) suggest that overcoming typical disincentives for acknowledging errors in public agencies is more likely when agency members are allowed to explain why targets were not met, can identify factors underlying undesirable trends, and are permitted to devise new solutions and incentives.
In a cross-national study, Newcomer and Olejniczak (2013) found that performance accountably and learning are often at odds, but they also identified several practices associated with successful learning. These included problem-solving simulations and decision-making games that create a sense of safety in taking risks with new ideas. They note that learning also emerged as a result of after-action reviews and data-driven reviews such as New York City’s COMSTAT analyses that surface clear deficiencies without requiring personal admission of blame. A forward-looking idea from the Census Bureau is a very popular agency-wide contest to identify improvements, and by implication less-than-satisfactory current practices.
Cognitive Recognition of the Risk or Severity of Dangers
Recognizing the risks in and seriousness of a situation is not often identified separately as a key factor in learning; however, without such recognition, indicators of problems will not trigger the learning needed to avoid disasters (Comfort, 2007; LaPorte & Consolini, 1991). According to Comfort (2007), the weak response to the Katrina disaster was not a result of information or communication lapses, but rather a failure of cognition: key actors did not recognize the severity of the threat or its consequences (p. 190).
Significant risk can be downplayed or ignored even when, perhaps ironically especially when, penalties for error are high. Normalization of deviance from rules in the Space Shuttle program (Vaughan, 1996) contributed to the loss of the Space Shuttle Challenger. Signs that could have led to lessons about safety were not taken seriously or acted upon, in part, because they would have delayed the launch schedule.
In highly reliable organizations, in which members have learned how to anticipate and cope with failures, mindfulness is the habit that anticipates and responds to subtle cues about potential failures. Weick and Sutcliffe identify this organizational quality as spending: (a) more time examining failure as a window on the health of the system, (b) more time resisting the urge to simplify assumptions about the world, (c) more time observing operations and their effects, (d) more time developing resilience to manage unexpected events, and (e) more time locating local expertise and creating a climate of deference to those experts. (Weick & Sutcliffe, 2006, p. 516)
Mindfulness training by physicians in a new pediatric intensive care unit allowed nursing staff to better identify signs of potentially fatal deterioration in the infants. Staff learned to make better decisions about bedside treatment and when to alert hospital physicians. Based on their military and emergency medical services (EMS) experiences, physicians used examples and stories to make staff aware of danger signs that could lead to negative outcomes for the infants, and for “that caregiver” (emphasis in the original, van Stralen, 2008, p. 81). The process eventually created a sense of trust and confidence among staff that they would not be criticized even for calling in a false alarm. “This gave caregivers internal, personal pride in their role in saving a child’s life” (p. 81) and led to better treatment outcomes. Similarly, Comfort (2007) argues that cognition and risk awareness in a disaster setting means that participants can recognize discrepancies between normal operations and changes in key indicators that signal danger that should lead to further investigation of vulnerabilities.
Crises That Reveal Problems
There is also some evidence that a disaster or a crisis sparks learning in its aftermath. The logic behind this is that the event imposes unmistakable proof of the need for change. The sense of dismay among agency professionals, lost agency prestige, and external public pressures for finding remedies all motivate special efforts to confront agency problems and learn from disaster. Birkland (2006) finds that disasters act as focusing events that bring together new support for preexisting ideas, and mobilize groups and political actors in support of learning and policy change. This support is typically short-lived, however, and there is a limited window of opportunity for resources and support for change before other priorities reemerge (Birkland, 1997). Similar learning can occur in agency procedures and interorganizational systems (Comfort, 1994; Comfort & Kapucu, 2006) in response to disasters. In contrast, however, Moynihan (2008) demonstrates the difficulties of learning during a crisis because of time pressures and the anxiety over possible further disaster. The effort to understand the unfolding disaster, the absence of relevant past experience, and the possibility of defensiveness or political posturing all limit learning during crises. Thus, learning from disasters is not easy, but is especially important.
Less catastrophic difficulties may also stimulate learning. Case studies of three Department of Labor programs illustrate learning in response to significant agency pressure (Mahler, 2013). In the Trade Adjustment Assistance program, legislation greatly enlarged program eligibility and created a backlog of over 250,000 cases. The Department’s Office of Performance Monitoring used work records and re-engineering techniques to streamline work processes to reduce the backlog. Budget cuts in the Wage and Hour Division led to less capacity to identify underpaid workers. Going outside the agency, staff contracted with Boston University and Mathematica to analyze data from the Bureau of Labor Statistics to pinpoint the industries and geographic areas in which violations of wage, overtime, and child labor laws were most common, allowing staff to better target their more limited case investigation time. Cuts in budgets and staff for investigations in the Office of Labor-Management Standards led them to conduct a detailed analysis of past audits of union elections and assets. This enabled them to identify the “red flags” that were most likely to yield convictions and again better target their reduced resources. Field staff were often initially reluctant to make changes in their routines, but their expressed pride in the results and their successes led them to overcome their earlier reticence. In each of these cases, analysis staff responded to new pressures, often budgetary, and used evidence-based analyses of past work results to craft improved policies and procedures.
These are classic examples of information processing as a means of organizational learning and should seem almost mundane. But that is the point: learning is, or should be, a common, indigenous process for remedying performance gaps. That is perhaps why so much of the literature on learning describes the ways in which learning is blocked.
Moderate Conflict and Tension
Moynihan (2005) notes the importance of collegiality and a focus on agency goals in successful learning forums. But conflict at moderate levels has also been identified as promoting agency learning. Bouwen and Fry (1991) studied learning in industrial organizations and found that the greatest innovations were seen when actors met head-on repeatedly over attempts to make basic changes in work organizations, but were also aware of their conflict and its effect on their progress toward solutions. Tension, struggle and confrontation fostered learning when they led members to negotiate solutions. Hedberg, Nystrom, and Starbuck (1976) argue that a residual level of discontent can lead to the next innovation. Furthermore, Moynihan and Landuyt (2009) note that excessive cohesion, such as groupthink processes, can short-circuit innovation thinking that may lead to learning. Jenkins-Smith and Sabatier (1993), in examining the role of learning in policy change, identify a moderate level of conflict among members of policy-making bodies as one condition that fosters learned change.
What this suggests is that even when organization members hold divergent views, learning and agency problem-solving is possible. Manageable levels of conflict may even spur debates about performance that lead to new lessons. As noted above, heavy-handed suppression of disagreements clearly blocks learning.
Leader Action to Explore Improvements and Undertake Change
Leadership too plays an important part in agency learning. Leaders can support, obstruct, or neglect the open exchange of information and the exploration of opportunities for improvement that are the hallmarks of learning (Moynihan, 2005). Acting strategically, leaders can adopt transformational styles to foster major changes and new learned solutions, whereas transactional styles are more suited to institutionalizing improvements and lessons (Vera & Crossan, 2004). Moynihan and Lavertu (2012) observe that “there is general agreement that supportive leadership fosters performance information use” (p. 596) that is a source of evidence-based improvements. Support can appear in the design of structures for collecting and sharing performance information or in the attention leaders pay to the use of such information for making decisions (Moynihan & Landuyt, 2009). Also important is the perceived likelihood of sustained support for the collection and use of performance information. Dull (2009), for example, argues that “Managers who do not believe an agency’s leadership is committed to performance measurement are . . . less likely to report using performance measures, and less likely to report they find performance measures useful in accomplishing managerial tasks” (p. 273).
Another perspective on leader influence is research on leader ambition. Teodoro (2011) finds that policy innovation is strongly associated with leaders’ career paths and job mobility. Those managers and executives who have come to the agency from the outside, on a diagonal career path, are more likely to innovate than those who have moved up through the ranks to executive agency positions. A professional’s reputation in the field and a hiring mandate for change helps account for this as does her or his ambition and desire for influence (Teodoro, 2010). In contrast, managers who rise through the ranks have been socialized into the established agency routines. He also finds that those executives with more mobility because of family situation and labor market opportunities are more likely to take risks (Teodoro, 2011). Ambition is no guarantee of learning, but it may lead executives to initiate efforts to find new ways for the agency to improve.
Opportunities for Transferring Lessons Across Organizations
Transferring knowledge that is the product of past problem-solving by other divisions or organizations has been well studied in private sector service and manufacturing organizations (Argote, 2013). In general, the more similar the units or organizations, the more likely transfers will occur. Some manufacturing firms, for example, set up factories in identical ways down to color schemes to make importing solutions seamless. Absorptive capacity, the extent to which units share some basic knowledge, makes transfers simpler because cause–effect relationships are already understood. Research also shows that lessons from high-status organizations are more likely to be emulated, and that some degree of social cohesion motivates sharing solutions (Argote, 2013). Transfer learning, like imitative learning (Levitt & March, 1988), is different from learning based on direct experience and from vicarious learning, in which events elsewhere are observed and lessons are drawn from them by the receiving organization.
Although there is generally less research on transfer learning in public organizations, Dull (2009) suggests that technology transfer or results diffusion is a process similar to transfer learning. Diffusion and adoption is described by Tolbert and Zucker (1983) who researched civil service reforms in cities in the early decades of the 20th century. The pattern of adoption they found illustrates several of the tendencies for transfer learning described for firms. They found that cities within states that mandated the reforms adopted them early on. “In contrast, when no state-level legitimation occurred, civil service procedures were adopted gradually, diffusing largely through social influence among cities” (p. 35) based on such factors as the age of the city, its population, and the proportion of better educated and professional citizens. Halachmi and Woron (2013) find transfer learning, facilitated by shared science models, among food safety agencies as they respond to a series of food-poisoning outbreaks.
Culture Values That Legitimate the Investigation and Adoption of Alternatives
The role of culture in supporting learning has already been noted in passing. Organizational culture, shared beliefs about the agency, its people and its work, can foster or thwart efforts to identify and communicate about deficiencies or opportunities for improvement. Moynihan (2005) found that “a culture of openness and experimentation” was behind the most successful examples of learning from performance information. Other characteristics include “high employee empowerment, participation, and organizational openness” (Moynihan & Landuyt, 2009, p. 1098). In contrast, embedded values of defending turf and contending subcultures can limit information sharing and imitation. Prevailing beliefs about the superiority of current practice can block recognition of opportunities for improvement (Barzelay, 1992; Mahler, 1997). Culture values also affect the willingness of managers to make serious effort to track performance and use the information for learning (Newcomer & Caudle, 2011). Taylor (2014) found, for example, the use of performance information for making evidence-based changes could be traced to cultural acceptance of performance regimes not only as an espoused value but also as an operating belief.
However, the place of culture as a requisite for learning is also sometimes invoked in a form of circular explanation, in which instances of learned improvements themselves are taken as evidence of a “learning culture.” It may be more helpful to identify what specific beliefs, assumptions, or values-in-use directly affect elements of the learning process. In that way, research would help uncover the importance of particular beliefs in fostering learning.
The Capacity to Draw Inferences and Make Connections
The heart of learning is the ability to draw inferences from information and make new connections. Many kinds of cognitive and inferential capacities can come into play to make learning possible. Other conditions described here support this ability or make it likely, but some form of this connection-making process is essential. That is, some means or procedure for drawing new conclusions is a necessary but not sufficient condition for learning to occur.
There are a number of ways described in the organization learning literature to make new connections. Analytic or calculative approaches identify improvements based on tracking results over time to find what actions led to the best consequences. In studies of these processes, researchers often point to learning based on trial and error experimentation (Greiling & Halachmi, 2013; Miner, Bassoff, & Moorman, 2001). Efforts that seem likely to succeed are tried, and those that lead to desired results are noted and preserved. Learning from failure is particularly important (Gressgard & Hansen, 2015), but trials without error are sought in high-risk settings (LaPorte & Consolini, 1991). Formal analyses of results data such as strategic planning, after-action reviews, benchmarking, and performance management all have the potential to support trial and error learning. Significant problems with these analytic approaches arise when members are unwilling to seek out the data available, acknowledge inadequacies, and actually use the information to make improvements (Moynihan & Lavertu, 2012; Moynihan & Pandey, 2010; Newcomer & Olejniczak, 2013; Poister & Streib, 2005). Here, again the role of culture in supporting the use of such data is critical (Taylor, 2014).
Bianchi and Williams (2015) take this kind of experimental approach further to identify the multiple, long-term cycles of experimentation and change that lead over time to improved results. Learned changes can result from one-time breakthroughs, but more often are the product of multi-year trial and error learning cycles until successful solutions emerge. Lebovic (1995) illustrates this process in arms-control tracking. Similarly, Sabatier (1993) notes the importance of seeing policy learning as a long-term, even decadal process.
A number of researchers have identified the capacity for technical analysis of quantitative results in particular as critical for learned improvements. Behn (2003), Moynihan and Kroll (2016), and Hatry and Davies (2011) have noted that the quality of data-driven reviews, including the capacity to analyze the reviews, is associated with the use and implementation of review findings and evidence-based improvements. Moynihan and Hawes (2012) find similarly that the analytic capacity of the central administration of schools is associated with higher levels of performance information use. Case examples from the Department of Labor described earlier also illustrate the importance of research capacity. In those cases, agency researchers used business process analysis and sophisticated statistical analysis by contractors to devise improved agency policies.
Statistical analysis for learning in industry goes back to the last decades of the 19th century. Learning curves are still used as a way of measuring and calculating the reduction in errors that occurs as experience increases. Manufacturers calculate the reduction in cost, time, or errors with increasing number of items produced. The resulting analysis shows a steep curve if improvements emerge quickly with experience or a gentle slope if learning occurs slowly (Argote, 2013). Thus, the use of the phrase “steep learning curve” to describe a difficult learning task is a misnomer!
Many forms of qualitative analysis can also lead to learning. Improvisational learning, learning in real time to solve an immediate problem (Miner et al., 2001), is a rapid variant of a trial and error process requiring access to a stable of possible solutions. Coping is a form of this improvisational learning, and Tummers, Bekkers, Vink, and Musheno (2015) identify some forms of behavioral coping among public servants that involve devising “long-lasting solutions to overcome stressful situations and meet client’s demands” (p. 1108). J. Q. Wilson (1989) describes the work of police officers as sometimes coping in undefined and rapidly changing situations in which the task definition can be an unhelpful “handling the situation” or taking charge (p. 38) and staying alive. Following Lipsky, Hupe and Hill (2007) note that coping choices of street-level bureaucrats become institutionalized, though unofficially, as policy.
Both trial and error strategies and coping imply that the organization members have some alternatives in mind when current practices are deemed inadequate. Barzelay (2007) summarizes several means of generating such alternatives including inventing, contriving, vicarious learning, and using design exemplars, that is, models found to solve similar problems elsewhere. Past experience in other organizations, ideas derived from exchanges at professional meetings or in journals, and new solutions generated imaginatively in-house provide options. Random experimentation or “grasping at straws” may occur. Nutt (1984), however, among others, suggests that ideas about alternatives are typically borrowed from similar, high-status organizations, imported by consultants, or invented in-house based on professional ideas or personal experiences, though these are often not successful (Nutt, 2004).
Crossan, Lane, and White delve into discourse-based learning practices. They describe cognitive processes underlying learning with their 4Is model, starting with “intuiting,” or recognizing the possibilities in an experience for change, “interpreting” or explaining the insight, “integrating” this insight with others’ understanding through dialogue, and “institutionalizing” the lessons into the organization’s systems, structures, procedures, and strategy (Crossan, Lane, & White, 1999, p. 525). The second and third elements of these processes especially describe the ways members go about finding new solutions. Brown and Duguid (1991) illustrate the interpreting and dialogue elements particularly well in their ethnographic study of technicians searching for solutions as they compare stories about past experiences.
Requisite Variety of Experiences and Results
Drawing inferences also requires some variation in conditions, actions, and results that can be compared with allow members to infer what leads to what (Levitt & March, 1988; Weick, 1987). This is why garbage can decision-making is inimical to learning, because a clear connection between cause and effect is obscured by the random character of attaching solutions to problems (March & Olsen, 1979). In the absence of direct experience with variation in efforts and results, demonstration programs or experiments with alternative routines becomes a rich source for learning (Korten, 1980). Moynihan (2005) found that “double loop” learning that creates fundamental change was made possible by experimentation with policies in the Vermont Department of Corrections. In contrast, rapid decisions that prematurely close off such sources of information limit the capacity of officials to make inferences (Landau & Stout, 1979).
When experiments are lacking (March, Sproull, & Tamuz, 1991) or when they pose a risk (LaPorte & Consolini, 1991), observing close calls, simulations, and plausible scenarios can substitute. Thus, learning can also occur from analyzing events that do not occur, or counterfactuals.
The Capacity to Change Policies and Practices
Learning becomes organizational when its lessons or solutions are somehow shared. Many descriptions of learning emphasize the joint or communal character of the process from the start as when professionals dialogue about solving problems or search for improved practices. In other cases, lessons created by some become distributed to all formally as policies, standard operating procedures, specifications, and structures. Lessons are often preserved informally as stories, habits, or cultural beliefs. Schein (2010) emphasizes in particular the importance of organization-wide cultural beliefs as repositories for preserving lessons learned. The literature identifies a number of ways that these formal or informal processes for preserving lessons occur.
In some cases, leadership is key. If the lessons are accepted by those with the authority to make changes, the lessons can be preserved formally (Khademian, 2002), but this may not happen. Self-interest or professional ideologies can lead to rejecting evidence-based change (Kaufman, 1973). Furthermore, lessons learned at the technical level may not be understood or appreciated by upper level managers so that the complexity of steps or actions at the heart of the new practice are not implemented (Garvin, Edmondson, & Gino, 2008). Legal constraints, internal cultural clashes, and external political pressures (Warwick, 1975) also threaten the acceptance of new evidence-based lessons. Lessons that clash with members’ cultural values about the right way to perform tasks threaten acquiescence (Barzelay, 1992).
In other instances, members appear to learn and spread lessons without official sanction or even knowledge. Brown and Duguid describe the process by which technicians, hampered by useless, deskilling repair manuals, are able to find solutions by comparing stories about past experiences with similar problems. The lessons created are shared in communities of practice with other technicians, without the new changes becoming official policy. Similarly, Raelin (1997) observes, “As practitioners come together by being involved with one another in action, they may become a community of practice wherein they learn to construct shared understanding amidst confusing and conflicting data” (p. 563). This learning-by-doing often results in tacit knowledge, another kind of shared lesson. Similarly, Evans, Grudniewicz, and Tsasis (2018) find that implementing an integrated hybrid health care system depended on the formation of shared mental models of tasks and relationships. These shared models are the product of learned associations and consensus around these new understandings.
The Capacity to Preserve Useful Lessons
Even lessons that are implemented and accepted, however, can fade as a result of turnover in organization staff or top leadership or when new administrations and policy or budgetary priorities emerge. Organizational memory, stories, and records are ways to preserve and pass on information about why changes were made as they were. Intentional “unlearning” or unintentional forgetting can have complex and difficult-to-anticipate effects. Unlearning can be a useful and necessary condition if new lessons are to be adopted (Hedberg, 1981; Meier & Hicklin, 2008), and strong ties to old lessons can frustrate new learning (Levitt & March, 1988). However, when the conditions that generated these lessons still exist, hard-won lessons that are forgotten or discarded can lead to a replay of past disasters (Mahler & Casamayou, 2009). Detrimental unlearning, however, is often difficult to identify in practice. It may be understood only after-the-fact. A repository for past lessons may reside in personnel with long histories.
Framework for Stimulating and Supporting Agency Learning
This catalog of conditions, capacities, and actions that promote learning implies that learning is a demanding and complex process with numerous requisites. However, looking at the specific effects of these requisites on building and sharing new connections offers a way to see them as different means for achieving a small number of basic learning processes. Furthermore, this framework for learning processes makes it possible to see that some of the requisites act as inducements or stimuli to learning while others play a role in supporting, facilitating, or maintaining connection-making. A few seem to do both. In particular, the more parsimonious framework described here identifies stimuli to learning by isolating the conditions or factors that lead members to identify current problems or the potential for improvements and motivate them to undertake the effort to find and share solutions.
The core learning processes revealed by this review include some that draw attention to the need for change or available opportunities for improvements. Other conditions identify the ways members can use evidence to make new inferences, that is, the methods for analyzing experience, drawing new inferences, and fitting solutions to problems. Yet, others are means for spreading the inferences to others in the organization. These three core processes offer a framework for learning and highlight the particular conditions that motivate the search for improvements.
Thus, the framework proposed here suggests that the extensive list of requirements found in the literature operate to engender learning in one of the three ways and in doing so identifies the motivators for learning:
1. By drawing attention to the need for or opportunity for improvements.
Information about the effects of agency actions, distributed across organizational actors, including effects that are impossible to ignore such as crises, draws attention to needed improvements and creates opportunities for change. Moderate levels of conflict can encourage this attention as members debate or argue about results and problems. Leaders can draw attention to evidence of failings and direct efforts to find improvements. Ambitious leaders may look for opportunities for innovation to make their reputations. Mindfulness, risk awareness, and cultural values that reward and motivate rooting out the sources of errors or failings also do so. Attractive exemplars of how unsatisfactory results have been overcome elsewhere can spur efforts to transfer lessons out of agency pride or even envy. All of these need not be present for learning to occur, but some way to drawing attention to the need for or opportunity for improvements is necessary. Without this, the motivation to adopt new ideas for improvement, even those stumbled upon, is lost. Thus, the actions or conditions that draw attention to needs and opportunities are often those that stimulate learning by motivating members to undertake the effort to find and share solutions.
2. By exercising a capacity for analysis and inference.
Organization members use and draw new inferences about how to improve in many ways. Some emphasize calculative methods while others see reflection, dialogue, and interactive story-telling as means for examining past results and finding solutions. Many cite the importance of sequential trial and error learning over a significant variety of experiences. Others note the significance of immediate coping, practice-based learning-by-doing, and tacit knowledge. Yet, others emphasize the importance of recognizing and borrowing attractive solutions from elsewhere. Making these connections depends on some level of quantitative or qualitative analytic capacity, but there are many ways that individuals and groups can go about this. The result, however, is a new connection, a new inference. Access to information about past results, experienced or observed, provides the material for these inferences.
3. By supporting the wide adoption of lessons for improving the organization.
Establishing and preserving the new connections can be accomplished officially by implementing policies, procedures, and structural arrangements that implement them. Cultural beliefs, communities of practice, institutional habits, and stories also embed past lessons less formally. Leaders can play an active or passive role in any of these.
This three-part framework is not a novel characterization as shown in the review above, but strips learning to its essential elements and makes it possible to see the alternative ways that these elements are accomplished. It is tempting to suggest that they are steps or stages, but there appear to be cases in which linkages become clear before a particular opportunity to use the lesson to solve a problem becomes apparent (Miner et al., 2001). Lessons lie in readiness for an opportunity to use them, for example. Solutions used elsewhere may be admired and remembered even though there is no immediate application. These are solutions in search of a problem, but unlike the garbage can decision-making case, the assignment of solutions to problems is not random or fortuitous, but is rather based on experience or evidence of some kind.
Stimulating and Supporting Learning
This view of learning highlights the ways that some requisites, perhaps better termed antecedents now, act to stimulate learning, while others support or facilitate it. All of the factors can make contributions, but a close reading of the research reviewed above in the context of learning theories generally argues that they play different roles. Distinguishing those conditions suggests ways to spur learning and foster its maintenance.
The difference appears to be that stimulants create a motivation at the personal, professional, or organizational politics levels to seek out and adopt evidence-based changes. These motivations can range from professional conscientiousness and feared loss of agency status to personal ambition and organizational competition. Like individual learning theories in psychology or expectancy theories of organizational behavior, prospects for a reward or threatened withdrawal of something valued acts to stimulate learning. The possibilities for rewards or threats are so varied and numerous in complex organizations that this motivational element of learning can take on many forms. Antecedents that support learning, however, make finding and adopting new lessons possible by providing the evidential material and analytic capabilities to identify potential improvements, but they do not in themselves create the motivation to seek improvements and search for new solutions. Conditions that make sharing and institutionalizing the new connections possible similarly support rather than initiate learning processes. These conditions are summarized in Table 1.
Actions to Stimulate or Support Learning.
Several examples noted earlier illustrate the difference between stimulating and supporting learning. NASA staff and contractors had long collected information about the O-Ring failures and the foam strikes as well as other dangers to NASA’s shuttles, but these indicators did not generate sustained efforts to learn from the accident and find solutions. Rather contractors and managers sought to play down the evidence (Vaughan, 1996; Mahler & Casamayou, 2009) until after the accidents, when analyses of past results was used as a base for making some, but not all, needed evidence-based improvements. Action to make use of the data long collected at the Department of Labor for finding new solutions emerged when serious resource constraints and massive work backlogs confronted the offices. Although less than an organizational disaster, these conditions put pressure on the offices to define problems and find solutions or suffer loss of valued professional accomplishments. Learning in a newly established pediatric intensive care unit took the form of mindfulness training for caregiver staff allowing them to identify signs of dangerously deteriorating conditions in infants early enough that life-saving treatments could be administered avoiding adverse consequences for the patients and their caregivers (van Stralen, 2008) . In these case examples, information about results was collected, but not used to make improvements until personal or organization pressures motivated the assessment of the information and the work of crafting and implementing solutions. Few of the published research studies on learning in public agencies look at what preceded the learning episodes to observe what might have initiated the learning process. Future cases might look in more detail at antecedent conditions.
These cases and the literature reviewed earlier help identify several conditions that act to stimulate and initiate learning by motivating the search for improvements. Crises and disasters draw attention to the practices that produced, at minimum, professional embarrassment and may also threaten serious political or budgetary repercussions. This damage to personal careers, profession reputations, and agency status may motivate efforts to avoid or resolve the crisis. Poor performance is difficult to ignore and resources may be available to aid in backing better solutions (Birkland, 2004; Gilson, Dunleavy, & Tinkler, 2009). However, severe accountability costs that may be associated with disasters may also inhibit learning when members try to disguise or ignore failings.
Further, personal commitment to a mission or to professional values and beliefs makes solving problems and making improvements for clients rewarding for members as described in the large and growing literature on public service motivation (Rainey, 2014). Tangible rewards, promotions, co-worker recognition, and professional awards may be forthcoming (Katz & Kahn, 1978). The mindfulness of professional staff who are attentive to lapses and take them seriously reflects this commitment and its rewards. Mindfulness in the case of organizational learning means that members have developed the habit over time of looking for and expecting to find indicators of problems such as failures of technologies or dangerous, fast-changing medical conditions. Mindfulness and awareness of impending risk, like crises that have already occurred, leads to the search for remedies and local adaptations (LaPorte & Consolini, 1991) and the professional and personal satisfactions that follow. It may also be possible that some moderate conflict or disagreement about the desirability of current practices can stimulate debate. Such disagreements themselves can sometimes reflect commitment to particular personal or professional values, and motivate action to achieve improved results, as found in policy learning (Jenkins-Smith & Sabatier, 1993). Leaders can take a direct role in making decisions based on performance-based lessons or implement innovations transferred from elsewhere out of professional pride, personal ambition, or conscientiousness. Some particular culture values or beliefs can act to spur evidence-based problem-solving, such as values that inspire dedicated service for clients or pride in rescuing clients, or that institutionalize expectations for program enhancement. Other cultural values may play a support role as noted below. In some cases, attractive exemplars may stoke or activate professional pride or even envy (Wilson & Lilly, 2016) and so spur lessons transfer.
All of these conditions will not be present, but any one of them can stimulate the kind of evidence-based examination that leads to new connections and lessons. Especially important, however, in the public sector is the role of conscientious, professional civil servants (Mahler & Casamayou, 2009). This links the initiation of learning to the vast literature on public service motivation. Developing this relationship is beyond the scope of this article.
Other conditions identified in the literature as requisites serve more to support than to initiate learning. Information about the organization’s past successes or performance gaps and its availability across the organization support the search for possible causes of problems and alternative ways of operating if members are looking for them. Leader encouragement of performance reviews and open discussion also supports the search for evidence-based improvements. However, data alone do not appear to be enough to spark the learning process, as illustrated in the many performance management research findings described earlier. Informed accountability policies further support a willingness to examine results and search for improvements, but on their own do not act to initiate them. Information that supports learning need not be quantitative as noted in many cases earlier. Many sources carry clues for solving problems. Some variation in experiences over time is helpful, but not necessarily a prerequisite for learning. Vicarious learning opportunities, counterfactuals, and scenario-building provide alternatives. The opportunity for achievement demonstrated by very successful and admired exemplars can lead to adopting lessons employed elsewhere if members are motivated to see them as solutions to problems or identify in them an opportunity for improvement. Thus, these exemplars can act as stimuli when they activate pride or other motivators for making and adopting new connections and as supports when they offer alternatives to consider in that process.
A facility with a particular methodological capacity to make connections also supports learning. These capabilities are the means for carrying though on making connections rather than initiating them. Many different exploratory or problem-solving methods can serve to push the process forward once started. Analytic, dialogic, tacit, and intuitive methods have all been identified as possible techniques that lead to making new connections. Trial and error, based on historical data or shared stories, can uncover solutions. Ongoing dialogues and reviews that create incremental adaptations over years are less visible instances of learning. All these support learning by making it possible to establish new connections, but are not in themselves the impetus to do so.
Cultural assumptions that permit or encourage members to voice problems, report unsatisfactory results, and identify possible remedies foster learning by supporting the commitments of agency leaders and professionals to try to solve problems and make program improvements. The ability to report problems, however, does not necessitate action based on the reports. Thus, some cultural assumptions act as supports for the search for improvements, whereas others act more directly to stimulate or initiate the process. This illustrates why it is necessary to identify the particular cultural values that are linked to learning rather than invoking a general culture of learning or change.
Finally, the ability to make and preserve the changes that were identified makes the learning organizational, and supports the entire process, but, of course, does not instigate it. Note too that several antecedents can act to stimulate or support learning depending on the particular form and intensity of the condition. Leadership action, particular agency-specific culture values, moderate levels of conflict, and attractive exemplars can either prompt learning or support the search for alternatives and the preservation of lessons.
The distinction between stimulating and supporting learning shows that the many requisites and hallmarks of learning identified in the literature do not all operate in the same way with regard to the core learning process of making new connections. All of them are possible supports, only some also act to initiate.
Conclusion
What this examination of research on organization learning reveals is the outline of the learning process and the conditions that stimulate and support it. Although all of the paths to learning can be encouraged, the stimuli have been less appreciated for their particular role. Attention to identifying the opportunities that these stimuli offer can improve the prospects for learning. Adding data, for example, is supportive but not instigative unless the data reveals a threat or potential for valuable achievement. Crises, one hopes, cannot be manufactured, but mindfulness, professionalism, alertness for successful models, a cultural dedication to improvement, and leader commitment to experimentation are features that can be instilled and the literature shows can prompt learning.
To further explore the role of these and other stimuli, future research on agency learning might include observations about the antecedent conditions that led to the learning process. Although there is no shortage of case studies of learning or literature on the virtues of and roadblocks to learning, very few look at the specific conditions that prompted the efforts to undertake evidence-based change. Looking to what events or conditions led to examples of learned improvements would add to our ability to foster learning.
Organizational learning is recognized as a particularly valuable means of agency improvement. It emphasizes change designed by agency professionals for the specific agency and its setting based on deep knowledge of the agency, it mission, and the results of its past programs and structures. This is in contrast to external, generic reform movements or popular management remedies. These are not necessarily inappropriate, but there is a certain value to exercising the professionalism of public administrators. They can be a repository of understanding of how to operate within the policy constrains in which they exist.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
