Jump to ContentJump to Main Navigation
Pragmatist DemocracyEvolutionary Learning as Public Philosophy$

Christopher Ansell

Print publication date: 2011

Print ISBN-13: 9780199772438

Published to Oxford Scholarship Online: January 2012

DOI: 10.1093/acprof:oso/9780199772438.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2017. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a monograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: null; date: 26 February 2017



(p.104) Chapter 6 Recursiveness
Pragmatist Democracy

Christopher K. Ansell

Oxford University Press

Abstract and Keywords

This chapter argues that recursiveness is a central but unspecified aspect of Pragmatist thought, defining recursiveness as a “continuous and interlocking cycle of perspectives.” Building on complexity theory, the chapter argues that recursiveness develops in organizations with “heterarchical” structures (with many-to-many linkages, rather than the many-to-one linkages typical of hierarchy). It then examines the New York Police Department’s Compstat system as a concrete case of organizational recursiveness. Under Compstat, direct, real-time communication between top executives and street-level bureaucrats supports strategic problem-solving. Generalizing from this example, the chapter develops the concept of a “constitutional hierarchy,” in which top executives primarily act as stewards of their organization’s general rules and values, but uphold and refine these rules and values through close contact with problem-solving operations. Recursiveness can be extended to entire organizational communities using the structural principles developed in Likert’s concept of “linking pin organization.”

Keywords:   pragmatism, recursiveness, heterarchy, NYPD, Compstat, hierarchy, Linking-Pin Organization

Activity is creative in so far as it moves to its own enrichment as activity, that is, bringing along with itself a release of further activities.

—John Dewey, Human Nature and Conduct

By creating a critical mass of creativity, innovation, and dedication at the executive level as well as in middle management and among the rank-and-file, the police chief executive can tip the balance to create an “epidemic” of organizational creativity, innovation, and dedication that yields remarkable results.

—Vincent Henry, The Compstat Paradigm

We live in an age where hidebound bureaucracies are supposed to give way to flexible and entrepreneurial networks, where hierarchical control and planning ought to be replaced by self-organization, and where decentralization is regarded as preferable to centralization. In fact, this state of affairs is just the current round of an old debate between freedom and constraint well known to Dewey, who wrote: “To view institutions as enemies of freedom, and all conventions as slaveries, is to deny the only means by which positive freedom in action can be secured” (1957, 166). Resolving the tension between top-down and bottom-up organization, however, is no trivial matter. The irrationalities of centralized and standardized rule-making are well known and lead us to celebrate bottom-up initiative and local knowledge. But then events like the beating of Rodney King, Hurricane Katrina, or Abu Ghraib come along to renew our sensitivity to the value of “command and control.” Consequently, public organizations seem to cycle between demands for decentralization and centralization.

This chapter explores the recursive quality of evolutionary learning for insights into how to more effectively manage these institutional tugs-of-war. Recursiveness is defined here as a continuous and interlocking cycle of perspectives. Although recursiveness is not a term explicitly used by the founding (p.105) Pragmatists, the idea runs throughout their work.1 As developed in detail in an excursus at the end of this chapter, for example, Dewey analyzed human psychology as a continuous and interlocking cycle of impulse, habit, and reflective intelligence. In much the same way, the tug-of-war between centralization and decentralization can be reconceived as a recursive process.

Applied to organizations, recursiveness is analogous to Follett’s concept of “circular integration,” as described in the previous chapter. However, one of the advantages of using the concept of recu rsiveness is that it allows us to draw a connection to complexity theory, via Douglas Hofstadter’s analysis of recursiveness. In his provocative book, Gödel, Escher, Bach, Hofstadter related recursiveness to the idea of “strange loops” and he analyzes these loops in terms of the concepts of “nesting” and “entangled hierarchy.” These ideas allow us to “loop back” to Herbert Simon and his conception of hierarchy as nested subassemblies.

Finally, armed with an elaborated model of recursiveness, this chapter explores its applicability to the tension between centralization and decentralization. Building on the analysis in chapters 4 and 5, the Compstat system developed by the New York Police Department (NYPD) is examined as a concrete case of organizational recursiveness. Developed during the tenure of Chief William Bratton, Compstat was part of a larger structural reform that decentralized decision-making discretion to precinct commanders. However, through the use of “real-time” crime statistics and frequent strategy meetings, Compstat allowed NYPD headquarters to actively steer and guide organizational problem-solving.

Elaborating the Model of Recursiveness

A number of scholars have developed models of recursiveness.2 Anthony Giddens’s model of structuration, for example, uses the concept in a manner quite compatible with Pragmatism. In Central Problems in Social Theory, Giddens writes:

In place of each of these dualisms, as a single conceptual move, the theory of structuration substitutes the central notion of the duality of structure. By duality of structure, I mean the essential recursiveness of social life, as constituted in social practices: structure is both medium and outcome of the reproduction of practices. Structure enters simultaneously into the constitution of the agent and social practices, and “exists” in the generating moments of this constitution. (1979, 5)

Gregory Bateson also developed a recursive model of life in his later work (Harries-Jones 1995). Bateson’s view of recursion sought to understand not (p.106) only how the parts composed the whole but also how the whole “reentered” the parts. In one account, Bateson regarded recursion as a form of causal looping: “a system of causation turned in upon itself and controls itself” (Harries-Jones 1995, 186).3

From the perspective of trying to overcome the dualism of centralization and decentralization, however, we can turn to an important theorist of complexity, the computer scientist Douglas Hofstadter. Recursiveness is a central concept in his wonderful analysis of the deeper linkages among art, language, and mathematics (Hofstadter 1979). The core idea of Hofstadter’s concept of recursiveness is the phenomenon of “strange loops,” which are graphically illustrated by M. C. Escher’s paintings that continuously loop back to where they started (or that shift back and forth in perspective from figure to ground). Hofstadter argues that recursion occurs in two basic types of structural context—a context of levels or “nesting” and a context of tangled hierarchies or “heterarchy.” It is analytically useful to point out that these two ideas exist in a fundamental tension to one other. It is this tension that allows recursion to occur.4

Let’s begin this discussion with Herbert Simon’s analysis of hierarchy as a nearly decomposable system, which was mentioned in the previous chapter. His argument suggests that hierarchy is a nested system, much like a Russian doll.5 The idea of nesting in both Simon (1962) and Hofstadter (1979) implies a clear ordering of the relationship between superordinate and subordinate levels: the superordinate level fully contains the subordinate level within it. Federations, for example, have a clear nested structure. The state of California is encompassed by the larger association of the United States of America. Furthermore, the jurisdictions of each of the states are mutually exclusive. California does not share territory with Nevada.

Since this concept of nesting is so pervasive, there is a tendency for analysts to see this kind of nested hierarchy as a universal principle of all organization (Koestler 1967). Yet nested relationships are not always so clear. Sometimes the nesting does not have such a well-defined ordering. This is where heterarchy comes in. From a set-theoretic perspective, the units of the heterarchy overlap, both vertically and horizontally—imagine California and Nevada sharing authority over a patch of territory. The units of a heterarchy are interpenetrating. In network-theoretic terms, the heterarchy is an organization structured according to the principle of many-to-many rather than many-to-one ties (Kontopolous 1993; Ansell 2000) (see fig. 2). In a heterarchy, both units and levels can still be observed, though they are less distinct than in the nearly decomposable hierarchy imagined by Simon. As a network of relationships becomes more spaghetti-like—“entangled” in Hofstadter’s terms—the distinctiveness of both levels and units is lost.


Figure 2. Hierarchy and Heterarchy Compared.

(p.107) Before proceeding, it is perhaps useful to draw a distinction between structural and behavioral organization. In terms of formal authority, the United States is a nested, clearly ordered hierarchy. However, when it comes to dealing with a tricky problem like control over water from the Colorado River, this nested ordering can often become less distinct. In behavioral terms, the relationship between California and Nevada probably looks more heterarchical, though structurally their jurisdictions are clearly demarcated.

In dynamic terms, the strange loops described by Hofstadter occur when a subordinate level (e.g., nested within a more encompassing level) flips over to become the superordinate level. The prior ordering is inverted. This cannot easily occur in a strictly hierarchical organization, with a clear ordering of relations between the whole and its parts.6 It is more likely to occur in a heterarchical organization, where the relationship between the whole and the parts is not as well defined. On the other hand, an inversion has no meaning in organizations without nested levels. When an organization becomes completely entangled—a spaghetti-like network—there are no distinct levels.

When a level of organization shifts back and forth between being superordinate (the nesting level) and subordinate (the nested level), you have the possibility of a powerful form of recursiveness that possibly overcomes the tension between top-down and bottom-up organization. In most contemporary organizations, this is likely to happen on the behavioral rather than on the structural level. In Dewey’s recursive model of habit, impulse, and intelligence, the inversion is the temporary reordering of perspective. Intelligence, for Dewey, may temporarily stand above and outside habit and impulse, but then perspective shifts back to habit and impulse as one engages in action.7 An “inversion” flips over hierarchy, placing the previously subordinate role in a (temporarily) superordinate position.8 Such temporary inversions of perspective and even command occur, behaviorally, all the time in hierarchies. Rochlin, La Porte, and Roberts (p.108) (1987) offer a good example in the aircraft carriers they studied. They found that a naval officer will defer to an enlisted person when the situation is a technical one and the enlisted person has greater technical or contextual knowledge.

To summarize, a recursive system can develop where organization—structural or behavioral—is heterarchical. This system should have some recognizable units and levels (a relationship of nesting), but these levels will be reciprocally embedded in each other. Under these conditions, recursive “strange loops” are possible. These loops work according to “inversions” in which subordinate levels flip over to become temporarily superordinate levels, only to flip back into a subordinate role. These ideas are consistent with the initial definition of recursiveness as a continuous and interlocking cycle of perspectives.

The NYPD and Compstat

With these ideas about how to describe recursiveness, we can analyze a concrete example—the NYPD’s Compstat system. Compstat, which stands for “compare statistics,” is a crime-fighting strategy developed by the NYPD under the leadership of former NYPD Police Chief William Bratton and refined and extended by subsequent chiefs (Howard Safir and Raymond Kelly). The centerpiece of Compstat is the use of real-time statistics to analyze crime patterns and to develop customized crime control strategies. It is often credited with a dramatic reduction of crime that led to New York’s current claim to be the “safest big city in America,” though the precise causal relationship between Compstat and crime reduction is contested.9 This chapter treats Compstat as an “existence proof” that recursiveness is a plausible mechanism for describing organizational life. The recursive qualities of Compstat helped the NYPD overcome persistent bureaucratic pathologies.

In his introduction to Phyllis McDonald’s study of Compstat, former NYPD Police Chief William Bratton succinctly describes the problem of hierarchy in the pre-Compstat NYPD:

The CompStat system and its results reflect a sea change in the way the New York Police Department does business. Like many large bureaucracies, the NYPD had been organized around avoiding risk and failure. For years, precinct commanders had been constrained on every side by regulations and procedures. Many police operations, such as prostitution sweeps and executing search warrants, could only be conducted by centralized units, reflecting an abiding distrust of precinct personnel and resources. Yet, despite all the micromanagement, the department was providing very little in the way of strategic direction. It was clear what precinct commanders and personnel weren’t (p.109) allowed to do, but much less clear what they ought to be doing to combat crime, disorder, and fear. (McDonald 2002, xiii)

Earlier attempts to deal with these bureaucratic pathologies sought to decentralize discretion to “beat cops,” a strategy closely linked to the NYPD’s broader agenda of encouraging community policing. Bratton also sought to partially decentralize decision making, but the beneficiaries of this decentralization were middle managers rather than street-level bureaucrats. Precinct commanders were given much greater authority and responsibility, while being held accountable to top police executives through the Compstat process.10 From one perspective, the Compstat process looks almost like a Tayloristic centralization of decision making. But as Greenberg notes: “The crime control model is not a contradiction to the ‘bottom-up’ theory espoused in community policing. Rather this model is where the ‘bottom-up’ and ‘top-down’ theories meet” (McDonald 2002, 50).

Within the literatures on policing, the Compstat model is often described from the perspective of problem-oriented policing, “broken windows” policing, or community policing; within the management literature, it shares some features with management-by-objectives, total quality management, strategic planning, and process reengineering. Compstat’s emphasis on performance measures and results-oriented accountability means that it also shares a family resemblance to New Public Management. Additionally, Compstat’s reliance on computer technology leads some people to view it as primarily a management information system.11 These descriptions are often apt, but from a recursiveness perspective they often overstress the use of Compstat to achieve top-down accountability.12

Overall, the Compstat process has some striking similarities with Dewey’s analysis of the relationship between habit, impulse, and intelligence. The Compstat meetings operate like high-level reflective intelligence, aiming to reorganize the routines of local units. Much traditional policing is incident-oriented and reactive, leading patrol officers (and others) to work according to rather unreflective routines.13 The Compstat meetings add the kind of reflective intelligence into the mix that Dewey describes as essential for continuous readaptation of habit. As McDonald writes:

The questioning is also designed to ensure that the patrol commander is evaluating all these tactics and strategies for effectiveness and altering them or experimenting with different applications if the desired results are not being achieved. In addition, the questions should provide the patrol commander with new approaches to consider in designing solutions to problems, ensure that needed specialized units (p.110) are cooperating with patrol, and ensure that the crime reduction objectives are being adequately addressed. (McDonald 2002, 44)

Compstat’s four basic principles—accurate, timely intelligence; rapid deployment; effective tactics; and relentless follow-up and assessment—call for a continuous looping cycle of experimentation.14

The core technology of Compstat—real-time crime statistics analyzed collectively in frequent Compstat meetings—is by no means a completely novel strategy in policing. Police departments have long collected and used crime statistics, and they have also used push-pin maps.15 Command meetings to plan crime control strategies are certainly not a novelty. But traditionally, crime control statistics became available after a long delay (say, six months), and command meetings were more infrequent and exclusive (limited to top executives). Most importantly, real-time data were not closely integrated into time-sensitive operational decision-making. The Compstat process uses real-time data collated several times a week to refine operational strategies in frequent, elaborate, and inclusive strategy meetings.16

Dewey’s analysis of habit, impulse, and intelligence can be read as an attempt to overcome the dualities of theory/practice and thought/ action. The equivalent in policing is to overcome the duality between central command planning (theory, intellectual work) versus street-level operations (practice, experience, routine).17 In fact, this tension was expressed as a tension between the administrative (program maintenance) role of executives and middle-level managers and the active on-the-ground role of beat officers. As McDonald notes:

In the past, most commanders in patrol viewed their job as primarily administrative. Their administrative roles were reinforced by community policing as the beat officer became the central element of problem solving and crime resolution in the police department while community policing encouraged partnership with the community, but it also reinforced an isolation from the patrol commander. (McDonald 2002, 73)

Like Dewey in Human Nature and Conduct, Compstat adopts an action-oriented attitude. The goal is to shift attention from program maintenance to active crime-fighting. For example, McDonald notes the bias against creating standing programs to deal with crime problems:

Too often in the past, both police and government have established “programs” to attack a specific problem. The programs then tended to go on indefinitely, as routines were established, personnel became wedded to new positions, and results were not checked. One impact (p.111) of continuous and ongoing feedback is a new urgency that produces dramatic results. (McDonald 2002, 80)

Compstat helps to overcome a division of labor that keeps top police executives and precinct commanders focused on either very high-level planning or on administrative details and keeps local patrol cops acting in a reactive, incident-oriented way.18

Having broadly described the affinity between Dewey’s analysis of the recursive loop between habit, impulse, and intelligence and Compstat’s continuous reconstruction of crime control strategy, let’s broaden the analysis to consider the more elaborated model of recursiveness. In terms of the organizational chart, the NYPD was, and is, a neatly organized hierarchy. Specialized police units (e.g., investigation, patrol, etc.) are clearly demarcated and nested within a traditional hierarchical structure. However, the Compstat reforms created a new kind of nesting that responds to some of the traditional “stovepiping” problems of police departments. As with many large police agencies, the NYPD suffered from the lack of cooperation between specialized police units. Jack Maple, one of Bratton’s top aides, provides a vivid example:

A borough-wide narcotics commander might assure me his people worked closely with homicide investigators to solve shootings and murders. But if I said, “Then tell me, what is the name of your counterpart in Homicide?” he could draw a blank. Drug trafficking was responsible for 30 to 50 percent of the city’s murders … but narcotics and detectives didn’t get together to compare notes. (Maple 1999, 23)19

The reforms made the geographic patrol unit—the precinct—the fundamental unit of policing, subordinating specialized units to the precinct commander (McDonald 2002 13, 58). Speaking about the failure of detectives and narcotics to cooperate, Maple provides a good example of the reformed structure:

We couldn’t just present this evidence [on drug-related homicides] to the detectives and Narcotics and ask them to play nice together; we forced the issue. Into each borough’s homicide squad, we introduced a “narcotics module,” consisting of a supervisor, six narcotics investigators, and two undercover officers, who could do buy-and-busts into the drug crews targeted by the investigators for questioning about the homicides. (Maple 1999, 104)

Although the actual structure of the NYPD did not need to change drastically, units based on geographical area (precincts) were given a priority over (p.112) specialized units.20 This was symbolized by the Compstat meetings themselves, where precinct commanders made the primary presentations about crime in their precinct while the chiefs and relevant personnel from more specialized waited in the wings (Silverman 1999, 118).

The shift to geographic-based policing led to a corresponding shift from hierarchical (many-to-one) to heterarchical (many-to-many) relations among the units. Operationally, this took the form of small teams assembled for specific crime-fighting projects, which linked specialized units as well as different precincts.21 Henry describes how task forces cut across jurisdictional boundaries:

Small, fluid and flexible task forces were used throughout the NYPD’s transformation, from the reengineering committees to the task forces that developed each crime strategy to the focus groups and policy advisory groups that defined and offered suggestions for the problems facing the agency. In the operational sphere, innumerable small, fluid and committed task forces were established to deal with problems that spanned precinct or patrol borough borders. (Henry 2002, 160)

This heterarchical style operated not only within the NYPD but also between the NYPD and the many other public agencies with which it had to cooperate. Although the NYPD always had relationships with these other agencies, Henry notes that the number of linkages became much more extensive after Compstat, so that “a schematic depiction of all the lines of communication and interaction among these agencies would resemble a web, with multiple interconnecting lines extending from each agency to every other agency” (Henry 2002, 8).

Heterarchy is vividly present in the Compsat meetings themselves. In these meetings:

The Commander can respond directly and publicly to the executive asking the question, and an atmosphere of transparency and fairness is created because everyone in the room is privy to the same information and everyone is held to the same standard. As we’ll see, this arrangement altered the ineffective traditional system of hierarchical communication (that is, communication up and down the established linear channels of a traditional organizational chart), transforming it into a communications model that resembled what Bratton has called a “seamless web” … rather than a pyramid. (Henry 2002, 264)

The Compstat meetings bring executives, precinct commanders, and various units into a small, transparent, and consequently highly energized arena (Henry 2002, 17–18).22 The transparency inherent in these meetings is an (p.113) important energizer of cross-functional cooperation (Henry 2002, 20, 270). As Silverman notes, Compstat brings jurisdictional disputes “out in the open” (Silverman 1999, 111).

Although in many respects the Compstat system appears to operate like a centralized accountability mechanism, it actually works to create a continuous cycle of learning between executives, middle managers, and street-level operatives.23 Precinct commanders learn that the Compstat meetings provide a forum for them to communicate with top executives and make direct requests for resources and cooperation (inverting the traditional top-down relationship). Prior to Compstat, precinct commanders would have had to navigate many bureaucratic layers to gain the audience of top brass (Henry 2002, 109). But with Compstat, they have developed a much more direct relationship. Henry describes the value of Compstat to commanders:

This is the commander’s chance to bring problems and issues (especially those concerning the adequacy of resources and crime patterns that cross precinct boundaries) to the attention of executive staff—in essence to publicly communicate their needs and, in doing so, to place some of the responsibility and some of the accountability on the executives. It is also the commander’s chance to impress his or her peers, and the motivational power of peer pressure is an important dynamic in Compstat…. In other words, commanders feel accountable to their peers as well as to the executive corps. (Henry 2002, 267)

Thus, the increasing transparency of information and decision making is a two-way street.

Compstat meetings encourage the development of shared ownership of problems across hierarchical levels. After the probing questioning of the Compstat meeting and the direct cooperation in strategy formation, top executives cannot plausibly deny knowledge or responsibility for operational strategies (or constraints imposed by resources or lack of cooperation). The Compstat system therefore empowers precinct-level officers, which they use to claim resources and support unavailable in the traditional NYPD.24 One good example offered by Henry describes a beat cop, Michael Kelley, who was trying to rid a tough neighborhood of drug dealing. Kelley tried arresting individual dealers but discovered that they were replaced as quickly as he arrested them:

Kelley approached his precinct commander, whose position, power, rank, and authority provided access to the kind of personnel and other resources necessary to effectively address the problem in its entirety. The precinct commander then mobilized the resources of other patrol officers, narcotics investigators, and homicide detectives, along with the DA’s office personnel to pursue (p.114) the problem (Henry 2002, 140). This same “inverted” logic can then work its way back up the hierarchy, with precinct commanders making claims on resources and support from higher levels and other units.

It is therefore a mistake to narrowly think of Compstat as a system of top-down accountability or management by results. Clearly, it has strong elements of accountability and performance measurement built into it. But these features can only be seen in proper context by interpreting Compstat as a recursive process of evolutionary learning, as summarized below.

  1. 1. Compstat is built upon a framework of problem-oriented policing. Real-time statistics and mapping technologies are an important part of this framework, but deep “pattern recognition” of crime problems rather than technology is the key to success. A deeply contextual understanding of crime problems must be linked to the development of customized problem-solving strategies (e.g., focused problem-solving).

  2. 2. Problems must be understood in context, but problems are multidimensional and interdependent. Therefore, effective problem-solving requires vertical and horizontal coordination that aligns different perspectives, skills, and resources. Compstat meetings provide the mechanism for aligning different perspectives, skills, and resources around customized problem-solving strategies. They provide a mechanism for voicing the views of precincts, specialist units, and headquarters.

  3. 3. The Compstat meeting is a mechanism for probing the meaning of problems and the usefulness of existing tactics (reflexivity). It is also a mechanism for brainstorming about new problem-solving tactics. Through the joint analysis of problems and problem-solving tactics, precincts develop joint ownership of problems with headquarters and shared responsibility for outcomes. As a mechanism that allows immediate and rapid cycling between different perspectives, direct, face-to-face communication (deliberation) is the key to joint ownership and shared responsibility.

  4. 4. Frequent Compstat meetings provide a mechanism for continuous tracking and assessment of problems and problem-solving strategies. Only continuous tracking and assessment can provide for continual refinement of customized tactics (experimentation).

  5. 5. Compstat was merely the centerpiece of a broader institutional reform that altered both the structure and the process of problem solving. Structurally, the reform decentralized discretion to the precinct level, while creating a more heterarchical project-based structure for problem solving.

Thus, Compstat integrates two interlocking cycles of perspectives. First, there is a cycle of problem solving (problem analysis, development of problem-solving (p.115) strategies, application of these tactics, assessment of tactics, problem analysis, etc.). Second, there is a recursive cycle that aligns the perspectives of precinct commanders and “beat” officers with the perspectives of other specialized units and top police officials at headquarters. The central problem with reform ideas that narrowly emphasize top-down accountability or performance management is that they encourage a distancing of top leadership, middle managers, and the rank-in-file—with leaders higher up the hierarchy standing in judgment of the problem-solving efforts of those lower in the hierarchy. Compstat suggests that whole organizations can only be mobilized around creative problem-solving if leaders are directly engaged in, and share ownership of, problem-solving strategies. Even then, there is nothing magical about Compstat. It can only be effective to the extent that it produces high-quality, innovative, and continuously improved problem-solving strategies.

Finally, it is also worth noting the strong resemblance between Compstat and Mary Parker Follet’s system of coordination and authority, as described in the last chapter. Recall that she stressed the “circular integration” of superordinates and subordinates, whose actions are aligned by the “law of the situation” (problem solving). She also stressed the development of “cumulative responsibility” produced by the “jointness” of activity.

Linking Pin Organization

The whole logic of Compstat eventually got pushed down the NYPD hierarchy. Mini-Compstat processes were created at the precinct level, motivated in part by the need for precinct commanders to develop information and strategy with their own staffs before being grilled in top-level Compstat meetings (Henry 2002, 273). Silverman quotes Inspector Joanne Jaffe, who observed: “The meetings have forced me to try to know everything that I can. If I have to know everything, I will make sure all my bosses and cops know everything, too” (Silverman 1999, 174). Precinct commanders are forced by Compstat to develop a detailed understanding of crime patterns in their precinct (Henry 2002, 257). Mini-Compstat processes, in turn, are pushed down to even the operational level below the precinct (Silverman 1999, 167). This diffusion of the Compstat process suggests that organizations like the NYPD might be better thought of as interconnected networks of recursive processes rather than a single large loop suggested by a Compstat meeting.

Rensis Likert’s (1967) concept of a “linking pin organization” suggests a way of thinking about how an organization can be structurally conceived as an interconnected network of recursive processes. Likert developed the concept of a “linking pin organization” to describe a structural pattern in which a (p.116) supervisor is a member of the team she supervises and a member of other management groups where she represents her team. The supervisor is a linking pin because she links different parts of the organization together. Ackoff (1987) developed the model further, emphasizing the circularity inherent in the linking pin model. He suggested the creation of “boards” at each managing level that included representatives from higher and lower levels in the organization. This structure, he argued, can create circular responsibility across levels.

Romme (1996, 1997) points out that the Likert and Ackoff models are two of the most prominent attempts to integrate the values of hierarchies and teams.25 The linking pin model can be said to embody at least two structural principles: interpenetration of groups (via overlapping membership) and short circuits of interaction across levels. Romme suggests a “double linking” between vertically ordered teams in which those with local information join the higher-level team (producing an upward flow of critical information) and a manager from the higher-level team joins the local team (to provide guidance on higher-order objectives). Romme and Witteloostuijn (1999) argue that double linking increases double- and triple-loop learning. The heterarchical model discussed above suggests that the organization needs to be linked horizontally (linking groups at the same hierarchical level) and vertically (linking groups across hierarchical levels).

Although Compstat is often understood to be a strategy for achieving accountability, this analysis suggests that it is better understood as having strong affinities with managerial models of learning and continuous process improvement.26 The linking pin model suggests that organizations are not a single loop of evolutionary learning but instead a network of overlapping loops.

One of the implications of the problem-solving model developed in the previous chapter is that the problem-solving strategy is continuously emergent and simultaneously engineered. In organization theory, the concept “emergent strategy” has two meanings. The first meaning is that strategy emerges from interaction among many participants and is not simply imposed hierarchically (Mintzberg and Waters 1985); the second meaning is that strategy is continuously being adapted to changing environmental conditions. Carr, Durant, and Downs write that “emergent strategy is undertaken by an organization that analyzes its environment and implements its strategy simultaneously” (2004, 80).27 Compstat offers examples of both meanings: it emerges vertically through the interaction of precincts and headquarters and horizontally through the interaction of multiple divisions; and it adapts in real time to changing patterns of crime. Simultaneous engineering means that the process of organizational adaptation is a nonlinear one. An organizational unit must continuously adapt its local action to the outputs of other units, which must in turn adapt their own local actions. In other words, simultaneous engineering (p.117) is a recursive process of mutual adaptation. To avoid the drift and reactive decision-making that may plague emergent and simultaneously engineered problem-solving strategies, both vertical and horizontal problem-solving activities must be aligned. The linking pin model provides a structural basis for this alignment.

An Alternative Model of Hierarchy

The model of recursiveness and the example of Compstat suggest an alternative model of hierarchy, which complements ideas about institutions and organization developed in chapters 3 and 4. Much of the current “post-Weberian” discourse identifies “hierarchy” as the problem. We are told that we have a choice between markets, hierarchies, or networks. Yet the problem is not exactly hierarchy, but rather a certain aspect of the conception of hierarchy: the sharp separation of planning from execution combined with a tight coupling of these antipodes through vertical chains of command and work by rule. The pathologies of this kind of hierarchy are typically seen from the perspective of street-level bureaucrats, who face cumbersome and dysfunctional rules that they must ignore in practice because they are so out of step with actual task imperatives. But it is less often appreciated that top executives also suffer serious constraints in this kind of hierarchy. After all, aren’t they in charge? Don’t they get to do the planning? Don’t they get to take the broad view? From their perspective, however, they are constantly fighting fires. For their top-level control system to work, they must continuously monitor the effect of their rules on their subordinates. Otherwise, they confront resistance from street-level bureaucrats whose “workarounds” make a mockery of their rules. But in devoting their attention to the detailed operational implementation of their rules, executives become embroiled in administrative minutiae. They are accused of micromanaging details, though this micromanagement appears necessary to achieve vertical control over the organization. From the top executives’ point of view, the cost is to the breadth of their perspective. Rather than focusing on the big picture, they become mired in administrative detail.

At some level, organization theory long ago noticed this dilemma and proposed alternatives, which were discovered in practice. For example, in the structural contingency tradition, Chandler described the multidivisional (M-form) organization, Stinchcombe described the craft organization, and Burns and Stalker described the organic organization. Perhaps Chandler’s description of an M-form as it developed under Dupont (as opposed to GM) is most telling.28 Rather than creating a tight coupling of planning and operations, the Dupont family gradually separated a high-level planning body from its semi-independent divisions. This allowed the planning body to truly plan (p.118) and allowed the divisions to focus on how to best achieve their goals. In many respects, the move to the M-form is the original version of “management by results”—the shift from control by inputs (operational rules) to control by outputs (production results). Yet what sometimes gets lost in this contemporary discussion—especially as interpreted by New Public Management and by a principal-agent perspective—is something that Chandler noticed: this separation allowed the Dupont board to focus on high-quality planning.

Pragmatism leads toward some of the same conclusions as the principal-agent model and New Public Management, but like Chandler it notices different features of the resulting system. Chandler’s M-form model, however, also tends to miss some of the important features of a Pragmatist-inspired approach. For example, if a high-level planning board develops general policy, and semi-independent divisions focus on the details of running their operations, then how do the two levels get aligned? How does the planning level, for example, know what it wants in the form of “results” without a detailed understanding of what is happening on the ground? And if they don’t understand the context in which their policy will be implemented, don’t they run into the same kinds of mismatch problems that their Weberian predecessors encountered? Pragmatism points to the value of an intermediary third level—a problem-focused level—that can intermediate between the top and the bottom.

Ironically, perhaps, we find an interesting and concrete example of this tripartite system in the development of the Tennessee Valley Authority (TVA), which Selznick critically studied. The TVA was run by a three-member board that operated like a high-level planning body. Below this board was a series of semi-independent operational divisions. As described in detail by Hargrove (1994), the key institution that mediated between the board and the semi-independent divisions was the general manager system. The general manager was a two-way communication conduit between the board and the divisions, often running shuttle diplomacy between the two and ultimately developing a measure of independent power because of this powerful brokerage role.

Compstat is a similar type of intermediary institution between high-level planning and street-level operations. Even more than a general manager system, Compstat provides the interface between policy planning and street-level operations. This interface is built on the concept of a strong role for precinct commanders, those archetypal “middle managers.”29 New Public Management, by contrast, often reduces the role of these middle managers, who are understood to be cogs in an input-oriented bureaucratic rule system. For Compstat—as for the TVA—these middle managers are the critical link in communicating between planning and operations. The Compstat interface itself allows top-level planners to understand the street-level context in which they work. It allows middle managers to directly communicate street-level conditions.

(p.119) A Constitutional Model of Hierarchy

Based on this introduction, we can further elaborate a theoretical model of Pragmatist hierarchy. This model is based on the idea, just introduced, that by freeing up top executives from detailed operational control, they can not only more fruitfully engage in strategic planning but also focus on giving the organization a clear and consistent policy direction. They do not do this by engaging in planning from a lofty and isolated position, freed from the cares of actual operations. Rather they engage in their role of shaping strategy and policy through closer engagement with the concrete problem-solving dilemmas of street-level bureaucrats.

The organizational levels of this hierarchy are loosely coupled and semi-autonomous. The top level of this hierarchy can be thought of as the “constitutional level.”30 This level focuses on setting the overall policy direction of the organization and defining the central purposes of the organization. In other words, this top-level body is consistent with Selznick’s notion of leadership as defining the mission and purpose of the organization. It is also compatible with the idea in the governance literature of steering or what Sørenson and Torfing (2007) call “meta-governance.” It also parallels the discussion in chapter 3 of institutional learning as a constitutional process.

As in the M-form, this organization is also composed of a level of semiautonomous subunits engaged in operations. These subunits, however, are also engaged in planning to the extent that they must plan for their own operations. In fact, there is no sharp separation of planning and operations in this organization. Subunits must plan their own operations. There is a distinction between operational planning and policy planning. But, as described in the Compstat system, the top level must be intimately familiar with operations and operational planning. A middle-level interface between top-level executives and the subunits is necessary. Rather than being simply a command system from top to bottom, high-level goals and operational practice are aligned through this interface. Here is where we see a major distinction between the Pragmatist hierarchy and a New Public Management (NPM)/principal-agent hierarchy. In shifting to a “results-orientation,” NPM suggests that the top and bottom can be aligned through performance measures, which will allow bureaucracy to escape from the pathologies of “input-oriented” (rule-driven) control. In fact, the problems of input-oriented control—the mismatch between general rules and action situations—persist for performance measurement. Performance measurement only works when the measures are meaningful measures of productive work—for example, “a test worth teaching to.” Otherwise, all sorts of pathologies develop. In the constitutional hierarchy proposed here, managers do not merely shift from inputs to outputs. They must remain (p.120) engaged in operations and input-oriented details, just as they must remain concerned about how to measure outputs.

Another point needs to be made here about the relationship between rule and action. Part of the criticism of Weberian hierarchy is that it leads to a stifling set of bureaucratic rules, ultimately leading to workarounds. The “results-oriented” New Public Management seeks to dispense with coordination-by-rule and replace it with coordination-by-incentives. The middle ground that Pragmatism adopts is that rules ought to be closely aligned to action. Thus, by implication, we should expect fewer rules, and rules that constrain and guide, but do not prevent judgment; and in return, we should expect these rules to be meaningfully expressed in concrete action.31 In other words, organizations should strive to continuously align rules with action. Mathilde Bourrier (1996) has described maintenance operations in two U.S. nuclear power plants. She expected to find that operators would consistently work around rules. However, she found that operators in both plants consistently applied and respected rules because those rules were regarded as important safety measures.

Continuous alignment of rule to action (this is Dewey’s continuous readaptation of habit again!) means that institutions ought to be restrained in the promulgation of rules. And they need mechanisms to continuously adjust and delete unused or inapplicable rules (see Bourrier 1996 for examples). Of course, every organization has such mechanisms in place to some degree. The constitutional model suggests that such institutional mechanisms for revising rules to align them with action ought to be enhanced. Rule alignment and revision is, however, difficult in organizations where top executives must use daily operational rules to gain vertical control over detailed actions of subordinates.

It is worth noting a changed meaning of centralization and decentralization in this model. These terms have the most meaning in a Weberian hierarchy. In a centralized bureaucracy, for instance, decision making takes place at the top of the organization and then flows vertically down the bureaucracy through a system of command and control. Thus, the pathologies of such a system are seen as problems of centralized control. Within this system, decentralization is seen as a useful corrective to these problems, which means relocating decision making to lower-level units. The Pragmatist hierarchy described here weakens this dichotomy between centralization and decentralization. Like the Forest Service organization described by Kaufman (1967), organizations can be simultaneously centralized and decentralized. This is possible because planning and operations are not sharply delineated.

This constitutional model of hierarchy supports recursive alignments among strategic planning, rule making, and operations. It also supports cross-functional problem-solving, which is essential for focused but comprehensive (p.121) solutions to problems (e.g., the Boston Gun Project described in the previous chapter) as well as more systematic problem-solving. With its emphasis on creating efficiency through specialization and its strong vertical emphasis on tight chains of command, the Weberian or “mechanistic” hierarchy tends to create vertical “silos” that constrain lateral coordination. With its more loosely coupled levels, by contrast, the constitutional hierarchy allows for greater lateral coordination, which is essential to Pragmatist problem-solving.

If silo relationships are created by vertical and segmental tendencies in organizations, relaxing the tight coupling between levels should create new opportunities for cross-functional integration or communication. This argument is already implicit in the description that has been so far given of the Pragmatist hierarchy. For example, the constitutional-policy level of the organization is freed from the administrative minutiae of controlling vertical chains of command. As this task is relaxed, top executives can focus on the broader issues facing the organization. When administrative boundary issues are no longer the focus of so much daily attention, it will be easier to see patterns of relevance that exist across functional areas. The same is true for lower-level units. As they become less focused on meeting the myriad segmental reporting and administrative requirements coming from above—and the more that they focus on solving concrete problems—the more they can be freed to develop lateral coordination with other units. One of the advantages of problem-focused action is that cross-functional coordination can be achieved at the point of need.


This chapter has argued that a continuous and interlocking cycle of perspectives—recursiveness—can help to overcome or manage some of the institutional tensions that exist in large-scale organizations and that discourage effective problem-solving. Building on Hofstadter’s understanding of recursiveness as “strange loops” that occur under heterarchical or fuzzily nested conditions, recursiveness is useful for understanding a pattern of iterative exchange across hierarchical levels. In a strange loop, the subordinate can temporarily become the superordinate, inverting normal status or scale orderings. This model of inversion helps us to imagine how recursiveness processes in hierarchical organizations might create conditions for more fruitful coordination across organizational levels and scales.

The Compstat case provides an example of recursiveness in practice. By creating a mechanism that brings precinct commanders into direct face-to-face contact with headquarters brass, a Compstat meeting creates the conditions (p.122) for continuous exchange and mutual adaptation between local problem-solving efforts and citywide strategy. Compstat meetings create top-down and peer-based pressures to engage in vigorous local problem-solving. But they also allow precinct commanders to demand resources from top brass and to develop shared responsibility for problem-solving strategies. The creation of “mini-Compstats” at the precinct level suggests that organizations can be thought of as many linked recursive cycles rather than a single loop. Likert’s linking pin organization provides a model for how these recursive cycles might be structurally linked together.

Drawing together this recursive model of organization with the discussion of “constitutional processes” in chapter 3 and the Selznick-Follet model of organization developed in chapter 4, this chapter concludes by describing a “constitutional” model of hierarchy. Higher levels of organization set broad strategies, guiding principles, and organizational values while granting considerable operational discretion to lower-level units. While the levels of the constitutional hierarchy are relatively autonomous from one another, the levels must be in continuous direct contact through the mediation of middle managers. However, rather than replacing “input-oriented” controls (bureaucratic rules) with “output-oriented” controls (New Public Management’s results), the constitutional hierarchy requires recursive alignment and improvement of problem-solving strategies, guiding principles, and organizational values.

Excursus: A Brief Analysis of Dewey’s Human Nature and Conduct

In a series of remarkable books—notably, Experience and Nature, Human Nature and Conduct, Art and Experience, and Logic: The Theory of Inquiry—John Dewey sought to overcome the philosophical dualisms that he saw poisoning modern thought and life: individual and society, subject and object, mind and body, thought and action, and reason and emotion, among others. Read as a strategy to overcome philosophical dualism, these books are not just a series of treatises on important philosophical concepts but rather a cumulative “reconstruction” of how we think about the human condition. Building on the work of other Pragmatist philosophers—Charles Peirce, William James, and George Herbert Mead—Dewey’s basic strategy was to reconstruct the basic concepts of philosophy so that the two sides of the dualism are continuous rather than dichotomous. But how did he do this? Did he have a particular technique? If so, an understanding of this technique might help us to think through other related problems. Yet Dewey never fully articulated the overarching technique he used to overcome dualism, though his magnum (p.123) opus on logic certainly provides a general description of how he thought we ought to engage in inquiry. This excursus examines Dewey’s overarching technique as a continuous looping “recursiveness.”

The technique of recursiveness is implicit in the model of iterative and organic growth that the Pragmatists, in their different ways, advocated (e.g., Peirce’s abductive method, James’s meliorism, Mead’s model of socialization, Dewey’s educational philosophy, etc.).32 This excursus analyzes one important Pragmatist work: John Dewey’s classic work on social psychology, Human Nature and Conduct (1957; hereafter, HNC), to provide one systematic illustration of the recursive nature of Pragmatist argumentation.33 This choice is relevant from an organization-theoretic perspective because, along with James’s Principles of Psychology, HNC was an important influence on Herbert Simon’s important analysis of organizational routine (Cohen 2007).

Dewey’s classic treatise on social psychology reads like an analytical dissection of three foundational concepts of human behavior: habit, impulse (or what we might today call “instinct”), and intelligence. Part I of the book is devoted to habit, part II focuses on impulse, and part III analyzes intelligence. Yet taken together, Dewey describes a recursive relationship between these three aspects of social psychology. In its simplest form, his argument is that habit, as organized impulse, shapes intelligence, while intelligence, via reflection, reconstructs habit. Development and creativity call for a continuous recursive cycle between habit, impulse, and intelligence.

What is the mechanism that drives this continuous loop between habit, impulse, and intelligence? It is this meta-analytic description of recursiveness that remains largely implicit in Dewey’s work, though he uses an equivalent strategy in other works (particularly, Experience and Nature and Art and Experience). To create a looping action, Dewey adopts at least six basic strategies. These are presented below with selected quotes from HNC to illustrate the point:

  1. 1. Dewey and the other Pragmatists overcome dualisms by introducing a third mediating term. In HNC, that third term is “habit,” which mediates between impulse (instinct) and intelligence: “Like most opposite extremes in philosophic theories, the two theories suffer from a common mistake. Both of them ignore the projective force of habit and the implications of habits in one another. Hence they separate the unified deed into two disjoined parts, an inner called motive and outer called act” (Dewey 1957, 43).

  2. 2. For each term, Dewey stresses its direct, immediate orientation toward action or activity: “Over and over again, one point has recurred for criticism;—the subordination of activity to a result outside itself…. Memory of the past, observation of the present, foresight of the future are indispensable. (p.124) But they are indispensable to a present liberation, an enriching growth of action” (Dewey 1957, 265). This direct, immediate orientation to action is what propels the recursive cycle forward. The Pragmatism orientation toward action is well known and is captured by Morton White’s (1949) description of Pragmatism as a “revolt against formalism.”

  3. 3. Although each term is oriented toward direct, immediate action, it also mediates between the other two terms. He makes this point in reference to overcoming the dualism of ends and means: “Means are means; they are intermediaries, middle terms. To grasp this fact is to have done with the ordinary dualism of means and ends. The ‘end’ is merely a series of acts viewed at a remote stage; and a means is merely the series viewed at an earlier one. The distinction between means and ends arises in surveying the course of a proposed line of action, a connected series in time” (Dewey 1957, 34). The treatment of a term as simultaneously “immediate” and “intermediate” is characteristic of Pragmatist attempts to overcome dualisms.

  4. 4. Each term has its own advantages and its own limitations. Value creation and creativity comes from harnessing them together, such that the advantages of one term overcome the limitations of the next term.

  5. 5. The loop between habit, impulse, and intelligence is continuous and the focus is on the process rather than a final destination. This recursive activity builds upon itself, leading to an expansion of capacity: “Yet the choice is not between throwing away rules previously developed and sticking obstinately by them. The intelligent alternative is to revise, adapt, expand and alter them. The problem is one of continuous vital readaptation” (Dewey 1957, 239–240).

  6. 6. The activity-oriented quality of Pragmatist thinking regards the present as a point of synthesis. Dewey writes that “‘Present’ activity is not a sharp knife-blade in time. The present is complex, containing within itself a multitude of habits and impulses” (Dewey 1957, 281).

Dewey does not place priority status on habit, impulse, or intelligence. But habit has a special role in his analysis because he sees it as the missing term in a dualism that pits the primacy of instinct against the primacy of reason. As Dewey emphasizes in the quote above, he sees habit as “projective,” by which he means oriented toward action. Habits are “active means that project themselves, energetic and dominating ways of acting” (HNC, 25). At the same time, habits are also intermediate terms that mediate the expression and execution of thought. Our ideas and our purposes work through the “refracting medium of bodily and moral habits” (Dewey 1957, 32).

Habits, however, run into problems. They may run up against environmental constraints or they may conflict with each other. And it is when they run (p.125) into problems—when their action-oriented style is thwarted—that they create opportunities for impulse. For Dewey, impulses are our native instincts. They come from within us rather than from our interaction with the environment and they can be in tension with habit. Impulses are also action-oriented. They demand expression and become pathological when suppressed. The action-orientation of impulse, however, is directed at habit (and by extension, at custom and institutions). Thus, Dewey argues that they are potentially creative: “Impulses are the pivots upon which the re-organization of activities turn, they are agencies of deviation, for giving new directions to old habits and changing their quality” (Dewey 1957, 93). Rather than suppress impulse, Dewey argues that it ought to be used intelligently to reorganize habit, custom, and institutions (Dewey 1957, 130).

Ultimately, impulses mediate the relationship between habit and thought. “The position of impulse in conduct is intermediary,” he writes, and goes on to say: “Thought is born as the twin of impulse in every moment of impeded habit” (Dewey 1957, 169, 171). For Dewey, thought means the capacity to reflect upon your own behavior. This capacity to reflect is a form of imagination that Dewey calls “dramatic rehearsal.” It arises from conflict or tension between habit and impulse, but it is also action-oriented:

We begin with the summary assertion that deliberation is a dramatic rehearsal (in imagination) of various competing possible lines of action. It starts from the blocking of efficient overt action, due to that conflict of prior habit and newly released impulse to which reference has been made. Then each habit, each impulse, involved in the temporary suspense of overt action takes its turn in being tried out. (Dewey 1957, 190)

Dramatic rehearsal (reflection) allows us to imagine action without actually acting. Like habit and impulse, however, reflection is also a mediating term. Reflection (or what he sometimes calls “deliberation”) uses ideas, goals, and principles to reconstruct habits, impulses, and their relationship. This argument comes out in Dewey’s experimentalist view. From this perspective, ideas and principles are hypotheses that reshape our basic premises of action. They complete the cycle that began with habit, but they also begin a new cycle.


(1) . Mead’s analysis of recursiveness can be seen as a cycle between three perspectives: I, me, and generalized other. Peirce’s “unlimited semiosis” is a continuous cycle between sign, object, and interpretant. James suggests that we understand experience as a continuous shift in perspective from “flights” (process) to “perchings” (structure).

(2) . Note that the term “recursiveness” may have different meanings in different fields. For example, the term has distinctive meanings in mathematics, computer science, and linguistics.

(3) . Bateson thinks recursiveness is triggered by difference (by something making a difference), which is similar to Peirce’s view of the “fixation of belief” (meaning as built on what difference something is making). Bateson was influenced by Maturan and Varela’s “autopoietic” concept of recursion. Clark writes: “Recursiveness means the socially accomplished reproduction of sequences of activity and action because actors involved possess a negotiated sense that one template from their repertoire will address a new situation. The sequences are activated and unfold in accordance with the socially constructed durational time based on events within processes…. Recursivness is always (p.220) improvised, even when the new cycle seems to replicate the old cycle and even when those seeking to impose reproduction are in agreement…. Recursiveness in social systems tends towards the autopoietic and is a form of self-organising complexity subject to dislocation” (1999, 67).

(4) . This tension between nesting and heterarchy is implied by Hofstadter’s treatment but not clearly stated.

(5) . Hofstadter also uses Russian dolls as an example of nesting (1979, 127).

(6) . Koestler (1967) suggested the concept “holon” to describe nested entities that are both wholes and parts.

(7) . In Pragmatism, there is no single privileged perspective from which to develop knowledge or make truth claims. Knowledge develops according to one’s perspective. Yet Pragmatism also argues that different perspectives are actually quite useful for the collective generation of knowledge. In fact, it is precisely this tension between competing perspectives through which progress is achieved.

(8) . Inversion is similar to switching back and forth between figure and ground. See Hofstadter (1979), chap. 3.

(9) . Karmen (2000) develops a skeptical analysis of the more vaunted claims of Compstat advocates. He argues that crime in New York “crashed” because of a “fortuitous confluence” of factors, many of them unrelated to the NYPD. However, he acknowledges that Compstat was one of these fortuitous factors. When comparing the reduction of homicides in New York to similar reductions in other cities, Rosenfeld, Fornango, and Baumer (2005) do not find that Compstat reduced homicides. Berk (2005), however, criticizes their statistical analysis.

(10) . Precinct commanders were given much more discretion over the assignment and deployment of precinct personnel (Silverman, 1999, 211–212). “In keeping with the theme of empowerment, local commanders were given broad discretion in how they investigated the complaints and what kind of discipline they imposed and they were not required to report back to IAB with the results of their investigation” (Silverman 1999, 215). Silverman describes this greater discretion as follows: “By increasing the authority and responsibility of precinct commanders, the NYPD freed them from having to forward information along the chain of command simply to receive high-level confirmation or assurance” (1999, 183).

(11) . Maple argues against seeing Compstat as a technological fix: “Outsiders often assume technology drove many of the changes in policing that we modeled in the NYPD. What a joke. A faith in timeless ideas like mapping came first, and we worked our way around or over the technological hurdles because we knew the results would be worth the trouble” (1999, 104–105).

(12) . See Silverman’s discussion of the “myths” about Compstat, most of which describe Compstat from a single point of view (statistics, technology, strategy meeting, accountability, etc.) (1999, 189–192). Scholars have often had a hard time classifying Compstat as a model. Weisburd et al. (2003) argue that one of the reasons Compstat has diffused rapidly is that it reinforces a hierarchical style of policing that was under pressure from community policing advocates. Some see it as a variant of the “rational legal-bureaucratic model” and in opposition to a “community policing model” (see Walsh and Vito 2004). Magers (2004) argues that Compstat combines the rational-bureaucratic and community policing models. Walsh and Vito distinguish three policing models: the rational-legal bureaucratic model, the community policing model, and the strategic management model. They argue that Compstat is an attempt to “synthesize the elements of the rational-legal bureaucratic and community problem-solving paradigms with strategic management concepts taken from the business world” (2004, 66).

(13) . Firman argues that Compstat helps to overcome the tension between creativity and control in police work by “making sure things are ‘under control’ while maintaining support for local (district) initiatives to meet demands” (2003, 458).

(p.221) (14) . Weisburd et al. argue that Compstat has six key elements: “mission clarification, internal accountability, geographic organization of command, organizational flexibility, data-driven problem identification and assessment, and innovative problem solving” (2003, 427).

(15) . As Maple notes: “In the NYPD of early 1994, not many detectives had access to all nineteen data systems, due in part to concerns that corrupt cops might sell information back to criminals, but mostly because catching crooks was not the department’s top priority” (1999, 88).

(16) . Silverman provides a good description of the evolution of Compstat (1999, chap. 5).

(17) . According to Silverman: “Traditionally, NYPD headquarters was perceived as the nerve center of the department’s decision-making apparatus. Changes in operational police tactics were conceived, formulated, and issued from headquarters, primarily on a citywide basis, and often with very little input from field commands. The post-1993 restructuring and crime-control strategies provided field commanders with far more leverage over their own troops” (1999, 183).

(18) . Willis, Mastrofski, and Weisburd (2004) studied the implementation of Compstat in the Lowell, Massachusetts, police department and found that it did not increase innovative problem-solving. They found that it did not encourage innovative problem-solving because Compstat was largely used to reinforce hierarchy and accountability. Willis, Mastrofski, and Weisburd (2003) report similar findings for the use of Compstat in Minneapolis and Newark.

(19) . Later, Maple writes: “If you asked the detectives if they talked to Narcotics about these cases, they’d stroke you by saying, ‘We work very, very closely together.’ In reality, they only work together on what they perceive to be ‘big cases,’ and not many of their ‘big cases’ involved small players in the narcotics business” (1999, 103).

(20) . Moore observes that Compstat was a “shift in responsibility and status from those who led special functional units to those who led the geographically defined, patrol-dominated precincts” (Moore 2003, 472).

(21) . Silverman notes that Compstat personnel “developed the idea of a pattern identification module (PIM), or unit, for each borough. A PIM is composed of representatives of the housing, transit, patrol, detective, and organized crime control bureaus, and the robbery squad” (1999, 119).

(22) . The Compstat system has many of the features Sabel (1994) describes as “learning-by-monitoring.”

(23) . As Moore observes: “If the system operates as a strict liability system, never pays much attention to what the managers are trying to do to deal with crime, or rewards them only when they do the conventional things to deal with particular crime problems, then the system will not do much to encourage organizational learning…. If, on the other hand, the system is interested in how managers are thinking about dealing with particular crime problems, and exposes that issue in collaborative discussion about what could be tried to reverse a problematic crime trend, then COMPSTAT can very well serve as a device for increasing the innovativeness and learning of the organization in dealing with selected crime issues” (2003, 485–486).

(24) . Weisburd et al. (2003), however, argue that police departments who have adopted Compstat have emphasized control over empowerment.

(25) . Another model that tries to combine the values of hierarchy and teams is Nonaka’s “hypertext” model of organizational knowledge creation: “The core feature of the hypertext organization is the ability to switch between the various ‘contexts’ of knowledge creation to accommodate changing requirements from situations both inside and outside the organization” (Nonaka 1994, 32). His hypertext model combines heterarchy and hierarchy: “Non-hierarchical, or ‘heterarchical’ self-organizing activities of teams are indispensable to generate new knowledge as well as to acquire ‘deep’ knowledge through intensive, focused search. On the other hand, a hierarchical division of (p.222) labor is more efficient and effective for implementation, exploitation, and accumulation of new knowledge as well as acquisition of information through extensive, unfocused search” (Nonaka 1994, 32).

(26) . Cavaleri (2008) argues that the roots of the idea of a learning organization are derived from Pragmatism; and Elkjaer (2004) suggests a model of organizational learning inspired by Dewey’s model of experience. Cavaleri also argues that the intellectual genealogy of Deming’s Total Quality Management (the forerunner of contemporary continuous process improvement models) can be traced back to the Pragmatist philosopher C. I. Lewis’s Mind and the World Order. Emison (2004) has also argued that Pragmatism provides a philosophical framework for Total Quality Management, and Silwa and Wilcox (2008) show that pragmatism inspired the work of W. A. Shewhart, one of the founders of quality management (which was one of the foundations for Deming’s work).

(27) . Following Peirce, they see emergent strategy as an abductive process, which is built from the “middle levels of the organization” (2004, 84).

(28) . There is now a considerable literature on “post-bureaucracy.” Part of the controversy around this literature is its epochal prediction—its argument that the era of bureaucracy is coming to an end and the era of “post-bureaucracy” is beginning (McSweeney 2006). McSweeney, by contrast, argues that in the United Kingdom there has been an intensification of bureaucracy. The post-bureaucracy model paralleled a similar discussion about the development of postindustrialism (Emery and Trist 1972).

(29) . Nonaka proposes a “middle-up-down management” that he contrasts with a “top-down” or “bottom-up” model (1994).

(30) . Sabel (1999) describes a similar “constitutional” organization.

(31) . Adler and Borys argue that enabling rules can be designed to allow discretion, capture lessons from experience, and promote integration of perspectives across the organization. “In an enabling approach to procedure design … employees are provided with a wide range of contextual information designed to help them interact creatively with the broader organization and environment. Procedures are therefore designed to afford them an understanding of how their own tasks fit into the whole” (1996, 73). They also argue that enabling procedures must be open to flexible readjustment (Adler and Borys point to Blau’s 1955 discussion of “adjustive development,” where practices develop in the course of working out operational problems.)

(32) . Maines argues that Blumer developed a recursive theory of social structure and social action that built on Mead and Dewey’s view of the “inextricably interrelated” of the knowing and the known (Maines 2001, 62–63). Trueit (2004) points out that Peirce’s mode of inquiry is “ongoing and recursive.” She links this idea of inquiry to Churchland’s inquiring systems and to contemporary models of learning organization. On “recursive constructivisim” in cognitive psychology, see Bickhart (2005), who links this approach back to Pragmatism and to the idea of scaffolding.

(33) . This reading of Human Nature and Conduct is indebted to a small conference, “Rereading Dewey: Implications for Organizational Studies,” organized by Michael Cohen, Karl Weick, and myself at the University of Michigan. It was Michael Cohen’s inspired idea to have all the participants of the conference read HNC. My analysis here has been influenced by Cohen’s argument that Dewey’s analysis of habit offers a way to reinvigorate organization theory’s concept of routine (Cohen 2007).