*317 less comprehensive units of government. Thus, while the governance council is located in the here and now of a particular place with particular problems, it is not by nature either a geographic or functional unit defined as doing certain tasks fixed in a certain place; rather‑‑like the states whose similar motility we describe in more detail below‑‑it goes where its constituents and their problems take it.
The service providers are the link between the government of officials and the local knowledge of citizens. Like their private‑sector homologues, they are responsible for refining and suggesting alternatives to initial design specifications. If necessary, they choose additional suppliers. Alone or with these lower‑tier suppliers, they pursue the agreed‑upon goals, and propose further refinements and alternatives as operations bring difficulties to light. To do all this effectively, they must combine expertise in their respective areas of specialization‑‑education, transportation, policing, and so on‑‑with the ability to collaborate closely with citizen users in the specification of services and the detection of errors in their provision, as well as with other parties who may suffer damages as a side effect of the service activity. As the consumers of the service, citizen users have unique knowledge of those particulars of their own, local circumstances that must be taken into account if even the most apparently routine and impersonal services are to be of value to them; conversely, those exposed to potential side effects are likely to have the sharpest eye for threats to their well‑being. The more directly an effective service must be coproduced by its beneficiaries, and the more complex the service's side effects, the more directly providers must cooperate with citizen users and those with affected interests. Similarly, to generate effective rules, rulemakers must cooperate with those subject to the rules. Consider some schematic examples, beginning with the most routine.
Buses that arrive punctually and frequently at times that are not synchronized with the commuting rhythms of potential riders do not increase the availability of public transportation. Potential riders know more about (changes in) their regular comings and goings than any department of public transportation or private bus operating service. Similarly, current riders are more likely to notice sooner than any fleet operator which bus stops could be relocated to reduce the risk of crime, which routes must be redrawn to permit access to new job opportunities, and which buses are not properly cleaned. Meanwhile, (sedentary) residents whose property abuts a proposed bus stop will be the first to insist on precautions to prevent the new waiting area from becoming a magnet for crime. Periodic community surveys and meetings, together with telephone hot lines and other customer‑service devices, are likely to provide the kind of participation necessary in such circumstances.
The police and public housing form an intermediate category. Here the quality of service provided depends so directly on the contribution of the beneficiaries that their active participation manifestly makes them coproviders. Police on the beat can only identify and avert the situations *318 that lead to disorder and crime‑‑e.g., rival gangs crossing paths on the way to different schools, landlords who tolerate drug dealing, and racist officers‑‑with the help of knowledgeable residents, and those residents can only address these problems of public order with the help of the police and other municipal authorities. Similarly, the way the occupants of public housing use their homes shapes the maintenance, public security, and other services they need from the housing authority, and the services provided by the housing authority plainly shape the conditions of occupancy. [FN120] In these situations, effective participation is likely to be tiered: Beat meetings and project councils provide fora for solving chains of particular problems, while police/civilian review boards, housing authority advisory councils, and ad hoc municipal task forces involving these and other agencies provide the opportunity to address larger questions of institutional architecture.
Perhaps closer and more nearly continuous collaboration between service providers and citizen users will occur in those frequent instances where the very aim of the service is to increase the citizens' ability to furnish it for themselves, and, as a step towards this, to learn to monitor the provider's capacities. This condition generally holds when the purpose of the service is to further learning, and most directly the case when learning is associated with mastery of the new disciplines. Thus, to take only one of many possible examples from current developments in schooling, an increasingly influential school of American pedagogy argues that high school and vocational students learn most when they learn to identify their difficulties in learning. They do so by beginning with questions that arise in familiar environments and uncovering the ways their habitual methods of problem solving obstruct a solution. [FN121] Such methods turn the traditional student‑teacher relation into an ongoing, consultative exchange between more nearly equal partners in the choice and execution of pedagogic projects. [FN122]
A similar redefinition of the learning process occurs when small and medium‑sized firms seek technical services from specialized consultants skilled in the new disciplines. [FN123] The most effective services help firms to *319 benchmark themselves as a means of identifying crucial problems (for instance, whether their difficulties stem from outdated technology or poor organization), and to use error detection methods to attack them. The more knowledgeable firms become, the more outdated they find the initial presentations of the consulting service, and the more they insist on refreshment. Learning becomes an exercise in error detection, and the firm's ability to find and eliminate faults in itself is a measure of the service provider's own utility. In the public sector, this form of learning entails continuous collaboration between service providers and citizen users‑‑through, for instance, industry or sector‑specific advisory councils composed of representatives of former or current clients, and companies and other public‑ sector institutions that work with them. Collaboration becomes a precondition for the service provider's ability to monitor itself, and the ability to organize such collaboration (again taking account of other affected interests and the relevant public bodies) becomes correspondingly more important as a criterion for selection as a service provider.
Hence, the service providers and their subcontractors are accountable, on the one side, to the citizen users, and, on the other, to appointed and elected political officials, who are themselves accountable to the citizenry. Citizens whose interests are likely to be affected for better or for worse by the provision of various services participate in the formulation of the strategy for service providers and help determine why services break down or fall short; they also take part periodically in benchmarking reviews of the providers' performance. These reviews can begin with comparisons of results obtained by various units of all like providers in the local jurisdiction, and extend (with the administrative assistance discussed below) to comparison of results obtained by providers in similar jurisdictions pursuing comparable ends by various means. These same reviews allow the responsible officials on the governance council to determine whether the providers have met the officials' (and the citizen users') expectations and whether, in any case, the providers demonstrate sufficient capacity to learn from their mistakes to improve. Benchmarking of the governance councils' decisionmaking routines‑‑the procedures for selecting service providers, monitoring their performance, and correcting by simultaneous engineering the problems arising from misspecification of the division of labor among them‑‑combined with the service‑specific reviews, allows the elected official to assess the council as a whole.
The citizens then evaluate the official(s) in elections, using the accumulated benchmarking information to compare the strategic choices and operating results of their governance council with those in similar locales *320 elsewhere. Electoral campaigns publicize the comparisons, underscoring the missed opportunities revealed in better outcomes, criticizing mismanagement at home by the measure of others' accomplishments, or simply provoking discussion of novel possibilities or reconsideration of discarded alternatives by recognizing promising developments in other jurisdictions. Inevitably, elections raise questions about the reorganization of public administration: If some public service providers compare poorly to others, or to private ones, the relevant electorate will want to know what conclusions to draw. By thus situating problems of local government in a broader context, democratic experimentalism suggests an alternative to the familiar idea of politics as concerned ultimately with the public posing of fateful choices based on conflicting ideologies.
Awkwardly balancing the needs of precision and the needs of euphony, we will use the term directly deliberative polyarchy to describe the form of democracy that results when a polity makes public choices by means of tiered governance councils‑‑councils that organize service provision with the collaboration of local citizens, and pool their experience to inform their separate decisions. It is direct because citizens act for themselves in elaborating solutions to problems that affect them, rather than delegating responsibility to representatives. It is deliberative because decisions regarding the provision of services are normally made by means of reason giving through discussion, not (except in cases of deadlock) the counting of votes. It is a polyarchy because the deliberations and performance of each jurisdiction count as considerations in deliberations of those like it. Polyarchy is a general name for a polity in which citizens, grouped in plural jurisdictions, can hold the officials of their jurisdiction sufficiently to account by democratic means to replace them when they perform badly. Jurisdictions in this general view are as free to ignore one another as they are to compete or cooperate. In directly deliberative polyarchy, the underlying pragmatic motives for local deliberation are the same as for pooling information among locales. In this sense, relations among jurisdictions are like relations within jurisdictions: An unanticipated alternative commands as much attention when chanced upon in response to local error as when remarked in distant innovation. Just as benchmarking others' experience illuminates and informs the diversity of disciplinary interests brought to bear in simultaneous engineering‑‑and vice versa‑‑so too deliberative polyarchy complements local problem solving as a means of loosening the hold of routine on public action.
This ability to evaluate routine gives directly deliberative polyarchy efficiency advantages and governance properties that appear as counterintuitive from the point of view of traditional public administration as do the corresponding features of pragmatist firms in comparison to mass producers. Thus, it counted as a truism in public administration that an increase in quality, taken as an effective adjustment to local circumstance, came at the price of a decrease in overall efficiency and, beyond that, *321 public accountability. [FN124] So long as rules are made centrally and are only legitimate when so made, this was almost a bookkeeping proposition. Seen from higher up, the exercise of local discretion required to make adjustment effective in particular cases would render the actual operation of administration opaque, inaccessible to coordinated improvement, and perhaps unlawful as well. Police on the beat were the canonical example of "street level bureaucrats," doing justice by (nearly) breaking the law and operating in a cloudy zone of informality equally inaccessible to organizational reform, judicial oversight, and public scrutiny. [FN125]
But an experimentalist local government that looks to local adjustment for direction in higher level reform makes virtues of these vices. Motivated through open discussion of its purposes and measured with respect to its effects, local experimentalism sufficiently formalizes the informal to give it the character of deliberate and innovative accommodation, rather than furtive caprice. Because its purpose and performance are public, successful local adjustments can lead to broad institutional changes. The same kinds of information that allow scrutiny of motives and outcomes, we will see repeatedly, permit judgments of lawfulness. Thus, in directly deliberative polyarchy, local initiative increases the quality of services while bettering the conditions for their efficient organization and, in any case, augmenting the public accountability of the providers.
Decisions in directly deliberative polyarchy no more rest on deep prior consensus than does collaboration among pragmatist firms. Nor do they depend on the log‑rolling compromises of aggregative democracy, or on the abstraction from the particular interests through which the deliberations of Rousseau and Madison arrive at rules fair to each by regulating only what is common to all. [FN126] Rather, in the pragmatist polity, *322 workable, long‑term collaboration can issue from, and aid the construction of, the institution of problem‑solving deliberation itself. Facing urgent problems that none can solve alone and seeking methods of establishing joint accountability, parties will often prefer to explore a potential solution, even if they are unsure of its outcome, than to do nothing. This collaboration arising from urgency is why direct deliberation, as our examples will show, in fact emerges first where the breakdown of traditional institutions is most conspicuous and its consequences most menacing: family‑support services, policing, and military contracting. Once begun, pragmatic problem solving loosens the hold of interest by fitfully darting, as it were, beyond its reach, thereby discovering solutions bit by bit in the unfamiliar territory beyond the reach of bounded rationality and habitual calculations of advantage. [FN127] Such discoveries beget others: The value to all of the current, partial innovation (measured as improvements in the performance of current problem‑solving institutions) will likely be increased substantially by the next innovation, and (as in the case of learning by monitoring in firms) the continuous exchange of operating information among the collaborators will reduce the risk that any party can use the novel arrangements for self‑dealing. In time, therefore, emerging solutions change what the actors do and how they rely on one another. Their very ideas of what is possible come to reflect these entanglements; "self"‑interest assumes as the starting point for subsequent calculations the surprises of practical deliberation that formerly confounded it. Thus, it is the very practical particularity of this deliberation‑‑above all the novelty that results when diverse standpoints are brought to bear on unfamiliar alternatives‑‑that advances the good of all participants.
The freedom of maneuver accorded local jurisdictions in directly deliberative polyarchy and the obligations of mutual regard that are its precondition both favor exploratory problem solving and become the more effective for it. Above all, an experimentalist regime gives locales substantial latitude in defining problems for themselves. The goals and methods of trash removal might appear to be relatively uncontroversial. [FN128] Not so with schooling: Whether, for example, academic excellence is best realized, or certain values best inculcated, in single‑sex rather than coeducational schools is a matter of dispute. [FN129] Instead of arguing the relevant matters of principle at long range through the institutions of representative democracy, local jurisdictions in directly deliberative polyarchy can in *323 such cases initially act on their own best understanding of ends and means. The possibilities for diversity inherent in this freedom pose problems, of course, for the maintenance of the uniformity of performance measures needed for benchmarking and other comparisons, as well as for the fundamental integrity of constitutional values; we take these up in the discussion of experimentalist administration and courts, respectively. But, problems aside, experimentalist decisionmaking is largely self‑starting precisely because it invites the actors to begin with what, on reflection, they take for granted: who they are and what they want.
Alternatively, when such autonomy does not prompt experimentalist exploration, local jurisdictions, or parties within them, have recourse to the apparatus of deliberative polyarchy. Thus, one of the tasks of the agencies responsible for benchmarking is to help the parties advance past blockages in local decisionmaking. They do so, for example, by suggesting how features of apparently irreconcilable alternatives have been combined into new hybrids elsewhere, or by proposing that clashing strategies be investigated separately as independent, concurrent pilot projects, or, more simply still, by chairing meetings by rules that all recognize as fair. Nor must this framework for generalizing results be fully established to be effective. As with workable collaboration, the institutional ligatures of participation can be an outcome of, not a requirement for, initiating change. Although benchmarking is formalized in a fully fledged system of democratic experimentalism, it can begin informally as a rough comparison of competing models and alternative performance measures. So long as the comparisons meet the pragmatist criterion of casting doubt on the assumption that current arrangements cannot be improved and suggest directions for improvement (including refinement of the comparisons), they count as polyarchic benchmarking, and encourage the generalization of forms of deliberation on which their own improvement depends, whether institutionalized or not.
Thus, we do not claim that practical deliberation can never be paralyzed by the clash of interests. It often will be. The argument, rather, is that sometimes it will not. If even one part of a system of democratic experimentalism succeeds in avoiding such blockage, the institutions that frame the regime as a whole can use that success both to help unblock the others and to increase their own ability to do so.
IV. Pieces of the New Polyarchy: Examples of Bootstrapping Reform
If directly deliberative polyarchy as a whole is still unrealized, partial variants of it‑‑with features that appear disruptively novel from the vantage point of current experience‑‑are already widely practiced and spreading rapidly in the United States and other countries. In some of these, the emphasis is on benchmarking and close collaboration between citizen users and service providers; in others, the accent is on simultaneous*324 engineering. All depart fundamentally from the New Deal model of provision of public services by hierarchically organized government bureaucracies with exclusive jurisdiction in their respective activities. These departures from the old model are not, of course, a demonstration of the viability or inevitability of a new one. But they are evidence of the possibility of change in the direction we prospect, and there is no reason to assume that it will be harder to connect these partial innovations into a new order than to have created them amidst the recalcitrance of the old. As coarse‑grained indications of current possibilities for reform, we present highly compressed case studies of pragmatist service provision: the reform of state family‑ support services, community policing, and the design of a submarine‑launched missile system.
The examples are chosen both for the clarity with which they reveal the connection between the details of institutional design and general operating principles and for the way they illuminate the construction of new institutions from the materials provided by the old. The reform of one institution both leads to and depends upon the reform of the institutions with which it collaborates. In this sense, the examples illustrate the bootstrapping nature of reform. [FN130] The examples are, in addition, suggestive because of the sheer improbability of the hostile settings in which they arise. No one was likely to look for the rudiments of new forms of participatory service provision in scandal‑ridden social‑service bureaucracies, crime‑ridden neighborhoods, or military‑industrial complexes. So their existence under inhospitable circumstances may be a harbinger of diffusion in more congenial contexts.
A. Family Support Services [FN131]
The failures of child‑ and family‑protective services, ranging from child welfare and mental health, to school‑based counseling, to juvenile justice and foster care, have become a leading symbol of the failures of government as a service provider and, beyond that, of the limits of public action in the face of urgent social distress. Too often families that might *325 have remained intact with timely help dissolve because agencies that focused on narrowly defined problems could not coordinate an effective, comprehensive response. Too often the dissolution results in the horrifying physical abuse of partners and children, and the foster‑care and other institutions intended to protect the victims inflict new wounds instead. In states as different as Georgia, Idaho, Iowa, Maryland, Missouri, and others besides, the reaction to this has been to attempt a wide‑ranging reform that gives citizen users and local communities a greater say in determining the services they need and how they shall be provided, and to use their decisions in turn to restructure the relevant parts of public administration. [FN132]
The common and defining features of this reform are concurrent and mutually dependent reorganizations of the central (state) and local (county and sometimes neighborhood) levels of service provision. Thus, at the state level, a "governance collaborative" (such is its generic name), consisting at a minimum of representatives of the key child and family services agencies, develops the general outlines of systemic reform and the principles and standards for establishing local collaboratives in consultation with local groups. This state governance collaborative also defines the core results to be achieved as well as measures of progress. Local collaboratives then form under these guidelines. These collaboratives convoke providers and citizen users, and present the state with plans for service delivery in their jurisdictions. As the process continues, the choices and experiences of each level influence the decisions of the others and the overall result is what we are calling the bootstrapping reform of both.
In practice, reflecting the confusion and uncertainty from which the reforms emerge, the composition and distribution of responsibility between state and local governance collaboratives, and the budgetary ties between them, vary sharply from state to state; these can change significantly within any one state in a short time. In Oregon, for example, the governor appoints a Commission on Children and Families of about fifteen members, including representatives of business and civic groups, social services, county commissions, and agencies. The heads of the concerned state departments serve ex officio. Membership on the corresponding county commissions is open, but must include a majority of lay citizens, among whom may be, again, business and civic leaders, or representatives of nonprofit service providers. In addition to administering the delivery of specific programs (Healthy Start, FRCs, Great Start), the county commission may plan the general reorganization of local services *326 for children and families, and draw upon the combined expertise of their own staffs and those of the state commission to pursue their plans.
In Georgia, the governor, lieutenant governor, and speaker of the state house together appoint about twelve members to a Family Policy Council that groups together business and religious leaders as well as state legislators. Here, too, state service department heads are members ex officio. The county counterparts include local elected officials, representatives of business, public agencies, and community and civic groups. The mandate of these entities is extremely broad: to achieve results with the Policy Council through comprehensive service plans financed by funds created by pooling the allocations of separate state agencies normally available to each area.
Maryland, to take a final variant, has created a sub‑Cabinet for Children, Youth, and Families, to which the governor appoints about nine members, including the secretaries of the pertinent state departments. This sub‑Cabinet collaborates with county‑level Local Management Boards (LMBs), including the county commissioner or manager and local representatives of providers of education, social, juvenile, health, and mental health services. The emphasis is on the reduction of out‑of‑home or foster placements, with LMBs allowed, in effect, to reinvest savings achieved in this area as they see fit. [FN133]
These reforms are still too new to permit any overall assessment of their effectiveness, or even to say which methods of composition and collaboration encourage effective monitoring and dissemination of best practices. Nonetheless, observers note two features of the reform that suggest that it is politically viable enough to survive to be judged on its substantive successes and failures. We call attention to them to suggest the kinds of mechanisms that may contribute to the bootstrapping diffusion of experimentalist methods more generally.
The first is that the very process of reciprocating consultation between central and local levels, and horizontally within each, necessary to formulate and adjust reform plans, is frequently judged by proponents of and participants in reform to be politically expedient as well. Interests that may oppose change find it hard to pursue a strategy of pure obstructionism in settings where many others are attempting sincerely to collaborate; once the potential obstructionists do begin to participate, they typically must counter proposals with other proposals, not vetoes. In this way, they become accomplices to the outcome, and subsequent criticism, if any, is blunted by this complicity. We take this finding as a fragmentary and preliminary indication of the disruptive effects of deliberative problem solving on settled interests. At the very least, it demonstrates in a context directly connected to the larger frame of reform under discussion *327 that bootstrapping incrementalism can be self‑ reinforcing rather than self‑blocking.
The second politically relevant feature of the reform movement in social services concerns the diffusion of information. Although neither systems of evaluation nor performance measures have yet been consolidated, it appears that the reforms reduce the possibilities for information hoarding‑‑and particularly the hoarding of unfavorable results. Performance failures, accounting anomalies, or abusive exercises of discretion are very likely to come to light because the constant consultation inherent in the articulation and redefinition of standards places many persons with expert knowledge in the know about current projects. The same forms of publicity that make it unlikely that the reforms will become hostage to particular interests make it similarly unlikely that they will become hostage to scandals and the manipulations that go with them.
Whether jurisdictions that wish to learn about one another's successes and failures will be able to find each other and usefully pool their results remains to be seen; thus far, innovators and bureaucrats coexist in a complicated patchwork. Nonetheless, inconclusive as it is, the record of the reforms suggests that they will not fail for a utopian disregard for the political conditions required to test them. [FN134]
B. Community Policing in Chicago [FN135]
The movement for community or problem‑oriented policing in large American cities such as Chicago, Houston, and New York closely resembles reform of child‑protective services. As in the latter, formal change begins with the creation by central (here municipal) authorities of local (neighborhood) service‑provision units whose substantial autonomy is to be exercised in collaboration with citizen users. And also as with child‑and family‑protective services, the central policing authorities find it easier to catalyze decentralization than to pool usefully the resulting local experience.
The differences concern the sheer difficulties of the task. However difficult reform of child‑protective services proves to be, reform of big‑city police departments appears in prospect harder still. First, there is the extraordinary centralization of police organizations. Large police departments were professionalized first on the military model during a wave of Progressive reform in the first half of the century, then professionalized again in the image of the mass‑production corporation during a second reform wave in the 1960s. [FN136] With their use of rank nomenclature, barracks, military‑ style roll calls at the start of each watch, minute direction *328 of the rank and file by voluminous written orders and direct daily commands, and designation of outsiders‑‑citizens and employees of other public agencies alike‑‑as "civilians," police departments seem a pastiche of the most centralized and hierarchical features of military and mass‑production corporate organizations.
Then there are the apparent incapacities of the citizen users most in need of better police services. Crime concentrates in distressed neighborhoods whose residents, impoverished and without much formal education or skill, would seem unqualified for, as well as disinclined to take part in, voting and the other routine forms of participation in representative democracy. [FN137] How are these residents to meet the manifestly greater demands of directly deliberative problem solving under favorable conditions, let alone in collaboration with organizations as rigid, hierarchical, and separated from the broader public by rifts in understanding and culture as the police? That there is demonstrably effective citizen participation in community policing in cities such as Chicago suggests that experimentalist institutions can serve two complementary functions. On the one hand, they act as organizational emollients, making hierarchies more fluid internally and more open to the outside. On the other, they validate local knowledge as a form of expertise by demonstrating its utility, thus fostering participation (and the learning it occasions) by those usually held incapable of it.
The backdrop to the current reforms was the extended attempt in the 1970s and 1980s to make the best of police isolation from local communities and the growing realization in the last decade of the profound limits of that accommodation. [FN138] At the most abstract level, the responsibility of the police is to ensure the greatest possible social order with full respect for the rights of citizens. Yet, in isolation from the communities in which they operate, the police have no information with which to render that responsibility more precise. For want of an alternative, they typically came to define their task as responding as effectively as possible to indisputable signs of disorder, such as 911 calls. But, naked deterrence aside, rushing to scenes of disorder cannot make local society more orderly. Thus, even as response times were reduced [FN139] and the prison population*329 swelled enormously, [FN140] police officials and academic observers increasingly came to realize that more effective policing required anticipating and interrupting the causes of disorder‑‑not just reacting decisively to it. [FN141] But for such anticipatory intervention the police would need to determine what to count as disorder, what the sources of disorder might be, how to address them, and how to determine the effectiveness of the resulting measures. Learning all this would require, in turn, the organization of discussion between police and citizens, as experts in the life of their communities and as direct users of police services, and perhaps even criminals, as particularly knowledgeable "users" of the criminal justice system, and then further discussion of what to do next and what to do better given the outcome. From these considerations has emerged the broad movement for community policing. Within that movement, certain wide‑ranging institutional reforms, such as the Chicago Alternative Policing Strategy (CAPS), unmistakably exhibit the features of democratic experimentalism. [FN142]
The Chicago Police Department (CPD) introduced pilot versions of CAPS into five of the city's twenty‑five districts in 1993 and expanded the program to the entire department in 1994. [FN143] The reform has been guided by three principles from the beginning. First, the business of rank‑and‑file policing is defined as the identification and resolution of particular, persistent problems of crime and disorder at the street or community level, rather than the execution of tasks defined by higher orders of command. Second, for purposes of this problem solving, "beat teams" are formed of seven or eight line officers, jointly responsible for a specific neighborhood (Chicago is divided into 279 beats) and assigned there for substantial periods. [FN144] The closer and more durable relation between the officers in the team and the neighborhood that is meant to result is called "beat integrity." Finally, in a sharp reversal of the historical effort to insulate *330 the force from the surrounding community, beat teams cooperate extensively with local residents, primarily in monthly community problem‑ solving sessions called "beat meetings," at which priorities are assigned to crime problems and strategies are developed to address the most pressing ones.
The beat meetings at the heart of the reform are typically attended by five or six patrol officers, of whom one may be a sergeant or other supervisor, and by anywhere from five to thirty beat residents, some representing community groups and others simply citizens concerned by crime in their neighborhood. As beat meetings have grown into primary centers of community and neighborhood problem solving, officials from aldermanic offices, city service departments, and the Mayor's office have begun to appear with increasing regularity. The usual practice is for the participants to locate a pressing problem‑‑such as an apartment building being used as a center for selling and using crack cocaine‑‑ to specify a linked series of measures in response, and to assign responsibility for accomplishing them. Thus, in the case of a crack house, one subgroup might determine whether the landlord is especially permissive with drug dealers or neglectful of the property and, if so, begin informal negotiations with the landlord or consider countermeasures to be pursued in housing court. A second might contact tenants or neighbors to enlist their aid in gathering evidence against the dealers or landlord. Yet a third might consult with specialized drug‑detail units within the police force about the possibilities of increased enforcement or with municipal maintenance services about the possibilities for repairing street lights and other public amenities. A map of the beat, annotated to reflect pooled knowledge of the problem, and a list of pending tasks frequently records the results. Thus, the first phase of CAPS created a novel administrative unit‑‑beat teams and resident‑participants convoked in beat meetings‑‑capable, within broad limits, of setting its own goals and choosing the means for reaching them, even when these strategic choices breached formal barriers within the police department and between it and other city agencies. By inviting direct resident participation in the identification of problems, formulation of strategies, and evaluation of results, the plain intent of the reform was to ensure that the choice of experimental projects and their evaluation would immediately reflect the judgment and preferences of those who, as the primary consumers of police services, must bear the brunt of police failure.
A recent study of CAPS indicates the breadth of participation and the counterintuitive make‑up of the participants. In 1995, some 5000 residents were attending community police meetings in their neighborhoods each month. The counterintuitive finding is that beat‑meeting turnout was greatest in those neighborhoods where crime levels were highest. This might seem a straightforward expression of need, but for the association between crime and poverty and the inverse association between poverty and political participation, noted above. Thus, in flagrant *331 exception to the de facto exclusion of the poor and the uneducated from the politics of mass democracies, the high‑crime Chicago neighborhoods with lower aggregate levels of education and wealth had higher participation rates than those that are better off. [FN145] Whatever the precise explanation for this outcome, moreover, the realization by beat‑meeting participants that their own participation mattered surely played a central part. A survey found that some sixty‑four percent of problems brought up at these meetings were being addressed and that full or partial solutions had been implemented in some sixty percent of the cases. [FN146]
The initial CAPS emphasis on beat integrity and the reform associated with it went hand in hand, however, with inattention to the problems of corresponding reforms at the center of the CPD to coordinate to best advantage the results of the local learning. Indeed, as first implemented, CAPS reforms seemed to presume that decentralized and informal problem solving would proceed with virtually no support from the center at all, let alone coordinating direction. Only when many beats floundered in their problem‑solving endeavors and others began to demand help from the CPD in implementing solutions, did managers at central headquarters respond by formalizing the core of the new procedures in a set of General Orders. These require documentation of problem‑solving efforts, evaluations, and justifications of priorities and strategies. [FN147] These reporting requirements not only assure line‑level accountability to management, but provide data that can be used to benchmark problem‑solving teams and efforts against one another. The Research and Development group at CPD is investigating implementation of electronic collection and distribution of this information.
All these efforts have been complemented, meanwhile, by attempts to benchmark and pool information initiated from below. In the course of the new problem solving, patrol officers and community activists have realized that decentralization can bring a kind of isolation: Though they may face similar problems with open‑air drug markets, abandoned buildings, problem parks, poor city services, or low beat‑meeting participation levels, neighborhood residents proceed separately, and therefore without benefit of the experiences of others. As a result, emergent leaders in community policing in some areas‑‑ sometimes police officers and sometimes residents‑‑dispatch successful problem solvers in one beat to assist less successful counterparts nearby by making available their expertise and, less often, additional resources.
Viewed through the lens of democratic experimentalism, the very recent project of Chicago community policing reforms is therefore promising*332 but incomplete. Though the fundamental elements of experimentalism are in place, the essential institutional machinery of benchmarking discipline has yet to be installed. Whether the reform will proceed along this path is an open question whose answer will in large measure depend upon whether reformers inside and outside the CPD at the center can join forces with those active in the beats to solve the difficult organizational problem of capturing, pooling, making sense of, and then diffusing the vast range of experience that these problem‑solving efforts instigate and reveal.
C. Simultaneous Design: Military Procurement
The disciplined decentralizations of the kind associated here with Japanese‑ inspired production methods have been discovered and rediscovered independently many times in this century by various armed services and their civilian suppliers. Typically this occurs when complex projects involving the integration of discrete, highly sophisticated subsystems must be developed and produced quickly, and yet‑‑given the menace of catastrophic failure‑‑with extraordinary reliability. [FN148] In the context of sophisticated and demanding military procurement, secrecy attends matters of national security, tradition favors hierarchy, complex technical debates appear opaque, and vast profits are to be made. All of these factors would seem to favor management by closed, even conspiratorial oligarchy. That decentralized methods have been widely used in this setting suggests as strongly as in the case of community policing both the power of such methods and their amenability to adoption across a wide range of government institutions.
The history of the design by the Special Projects Office of the United States Navy of the Polaris submarine‑launched missile system‑‑the largest and most successful naval project of the early Cold War‑‑illustrates one of the many paths to the new class of institutions. [FN149] Until 1955, when the project began, naval procurements were managed by collaboration between the Chief of Naval Operations‑‑who, as commander of the fleets, determined weapons requirements‑‑and a group of technical departments‑‑specializing in areas such as aeronautics, ordnance, or ships‑‑which selected a prime or general contractor to produce the material demanded in their respective jurisdictions. The prime contractor, in turn, parceled out specialized tasks to subcontractors and assumed responsibility for integrating the weapons system itself. But missile development‑‑and particularly the development of missiles requiring complex shipboard fire‑control mechanisms‑‑straddled the boundaries between *333 jurisdictions: Ordinance and aeronautics both insisted on control. [FN150] Moreover, as strategic ballistic missiles were then an innovation for the Navy, commanders and their representative in the Chief of Naval Operations could not use experience as a guide to formulating operational needs. Nor, finally, was there a pool of prime contractors who might have supplied at least part of the coordination that the Navy could not. Previous developments, principally with the Chrysler Corporation, had been with liquid‑ fuel missiles, whereas for submersed launchings the Navy needed a solid‑fuel missile with correspondingly different guidance and fire‑control mechanisms. [FN151] Lockheed's Missile and Space Division, which did have solid‑fuel competence, had none in other important aspects of the project. [FN152]
In response to these difficulties, the Navy created the Special Projects Office as a kind of federation of the key military participants and specialist private‑industry contractors. A Steering Task Group composed of senior representatives of these two groups fixed the performance characteristics of the new missile system; a Projects and Programs Branch of the Special Projects Office was assigned the task of supervising realization of the goals. [FN153]
This federation might have easily degenerated into a cartel of established interests, with each technical bureau defending its own bailiwick and those of its favorite contractors by insisting on particular versions of the subsystem within its competence‑‑thus disregarding the effects of its preferences on the performance of the system as a whole. This danger was all the greater as the Special Projects Office formally had responsibility for submitting a unified budget for the program, but no direct authority over the participating technical bureaus. [FN154]
To avoid this combination of conflict and collusion, the senior naval officers in the Office adroitly played on the original jurisdictional ambiguities both to create and moderate competition among technical bureaus, among contractors, and between contractors as a group and their supervising bureaus. Almost inadvertently, they created a system for proposing and comparing design variants with many affinities to simultaneous engineering. Thus, bureaus with competence in adjacent areas were encouraged to propose solutions to particular problems so that the relevant panel of the Steering Task Group or section of the Projects and Programs Branch could choose between them. The bureaus, in turn, responded to the pressure to perform by encouraging competition among their specialist contractors, and choosing the superior design. To ensure that neither historic loyalties nor efforts to gain an advantage in the jurisdictional jostling would lead to collusion between bureaus and particular *334 subcontractors, or more generally to the suppression of local initiative, contractors were entitled to appeal decisions against them to the Office in the early stages of the program, and later allowed to negotiate minor design changes with subunits of the Office located at their own facilities. [FN155] The overall result, as in simultaneous engineering and its related disciplines, was to create an environment that shifted incentives from information hoarding (to avoid the possibility of threatening comparisons) to information pooling (to increase the likelihood of surviving them). [FN156]
The design of the Atlas/Titan series of intercontinental ballistic missiles, begun formally in 1954‑‑a year before the start of the Polaris program‑‑was informed by closely kindred principles that came to be called "system engineering." [FN157] Here too the novel demands of the project were at odds with military jurisdictions and orientations, as well as with industrial specializations. An Air Force hierarchy dominated by experienced pilots had trouble even imagining the utility of unmanned vehicles operating outside the atmosphere, let alone contemplating the optimal specifications for such a craft. [FN158] Even innovative airframe manufacturers, at home with the idea of missiles, had trouble conceiving of such missiles as mere platforms for electronic guidance systems and warheads, rather than engines of flight with an integrity of their own. [FN159] Here too the solution was to create a design team federating specialists and generalists from the military and private sector, together able to challenge and correct one another and the additional specialized subcontractors they would recruit to the project. On the private‑sector side, the key figure was the newly formed firm (Ramo‑ Woolridge Corporation) that was both a specialist in aerospace electronics and a pioneer of the emerging systems approach. The systems approach was understood, according to one of the company's founders, as an inherently "interdisciplinary" form of engineering whose "function is to integrate the specialized separate pieces of a complex of apparatus and people‑‑the system‑‑ into a harmonious *335 ensemble that optimally achieves the desired end." [FN160] The Air Force in turn assembled a team of technically expert officers, each responsible for monitoring progress in one of the subsystems of the overall project. [FN161]
Design under this structure was, in the jargon of systems engineering that later shaped the vocabulary of simultaneous engineering, concurrent and parallel. Concurrence meant that parts, and even machines to make those parts, were developed simultaneously, in advance of final definition of the whole. Parallel design meant that related but distinct alternative systems were developed side by side so that failures in the parts or subsystems of one could be corrected by substituting the corresponding component of the other. Thus, the launch facilities at Cape Canaveral (now Cape Kennedy) were under construction before there was a missile to launch.
The Titan rocket was produced as an almost accidental byproduct of parallel development. First, the Air Force commissioned construction of a second missile engine "for reasons of competition and the undesirability of entrusting such an important part of the program to one company." [FN162] The same logic suggested paralleling the other subsystem; doing this created all the components of the Titan. [FN163] By the early 1960s, the Air Force's success with parallel design attracted the attention of economists, who constructed sophisticated models of the savings in money and time that it made possible. [FN164]
Much more generally, the application of at least the rudiments of simultaneous engineering principles stands out in a recent survey of government sponsored research and development in the United States in the post‑War period as a central determinant of success. Consider the experience of sectors such as semiconductors, computer hardware and software, and biotechnology. In these areas, government policy (including strict enforcement of anti‑trust provisions and patenting policies favoring new entrants) encouraged early exploration of diverse alternatives and competition among public and/or private entities performing research or developing technology in connection with particular projects. The aggregate outcomes were extraordinary successes. In contrast, when, as in the case of the supersonic transport project or the effort to build a liquid metal fast breeder nuclear reactor, the government made an early and *336 sustained commitment to a particular design and a group of firms associated with it, the results were clamorous failures. [FN165]
There is, to be sure, irony in the observation that the military‑industrial complex‑‑symbol to many of government as an instrument of self‑dealing, and to others of a suspect connection between official power and violence‑‑may well have been a pioneer in the use of methods that we would associate with a new form of democracy. But it is irony of a familiar and instructive kind. The means of production and the means of destruction mirror and shape one another as breakthroughs in civilian life are passed to the military, and vice versa; and every contrast among civilian forms of organization has its counterpart in differences of military style. Thus, trench warfare solidified hierarchical forms of command, while tank warfare favored flexible and decentralized decisionmaking. [FN166] (Or, to choose an example with a closer connection to political participation, the armed services were desegregated by Executive Order before the Supreme Court ruled Jim Crow unconstitutional. [FN167]) From this perspective, the surprising aspect of the successes of simultaneous engineering in recent military projects is not, perhaps, their occurrence, but rather our tardiness in drawing implications about the plasticity and transferability of government from it.
D. Limitations of Piecemeal Efforts
A few hints to the contrary aside, the foregoing examples suggest that directly deliberative polyarchy, because it does not presume consensus and because the information‑pooling devices on which it ultimately depends may begin informally, can spread piecemeal, by a process of bootstrapping akin to the one noted in the diffusion of learning‑by‑monitoring institutions in firms. However, without national coordinating mechanisms, those interested in innovating will have difficulty finding one another and pooling their efforts to overcome the predictable resistance of vested interests. To see why such mechanisms are necessary, we sketch the possibilities and limits of uncoordinated reform.
The possibilities for bootstrapping reform stem from the urgent need to respond to, and the freedom of maneuver afforded by, breakdowns in current arrangements. Central (state or national) governments, for example, may admit defeat in efforts to solve certain problems by programs under their direct control and then redistribute the funds dedicated to the failed efforts to lower level authorities, perhaps on condition *337 that these give some public account for the use they make of their new discretion. Local authorities may have the freedom of maneuver under current arrangements to engage in some form of experimentalism, or attribute themselves this freedom, and the central authorities are so preoccupied by other matters (or so sympathetic, surreptitiously, with the initiatives) that they do not protest. Success by any one of these approaches will often be self‑reinforcing and will encourage attempts by the others. Thus, a school district relaxes certain rules to permit decentralized experimentation in curriculum development (itself informed by new ideas of learning through participation in problem‑solving groups), perhaps in response to certain semiofficial demonstration projects undertaken by teachers with the help of some parents and administrators. So authorized, the school affiliates itself with one or more associations or networks of similar institutions pursuing related forms of decentralization. Through contact with these, the school launches additional pilot projects that, together with the remonstrations of other schools in the same jurisdiction pursuing slightly different projects in association with other networks, result in requests for further rule changes at the district level. Meanwhile, the federal government begins to consolidate separate programs for linguistic minorities, students with learning disabilities, and many others into block grants to be used substantially at the discretion of local school districts. But a condition for the use of the funds is the articulation of a proposal for the reorganization of the old services on new lines‑‑particularly their integration with one another and classroom activities. Efforts at curricular experimentation now have to be coordinated with experimental delivery of supplementary services, occasioning discussion of further rule changes, comparisons with responses in other districts to this kind of integration, and so on. Processes of this kind explain why examples of the new system, however fragmentary, abound.
But to say that directly deliberative polyarchy, because of the forms of accountability and consensus it establishes, can begin to take shape spontaneously in response to current difficulties is not to say that its diffusion is automatic and assured, nor even that the new system of local government as described is self‑sufficient and requires no national complement or preconditions. Diffusion is not assured because the change to directly deliberative polyarchy, like the introduction of learning‑by‑monitoring methods in firms, creates great uncertainty about who will win and lose by the change. This uncertainty produces a vast reserve of potential opponents, easily activated by any event that confirms their fears. Think of public‑sector employees and unions menaced by competition from private service providers or from public‑sector service providers from other jurisdictions; of central‑ office administrators menaced by the increasing autonomy of local officials; of vulnerable economic or ethnic groups that prefer the threadbare protection they now enjoy from national programs to the gamble that they will do better with some version of decentralization. Whether such opposition will prevail is indeterminate. Looking at *338 the forces deployed against change, close students of the capture of regulatory agencies by special interests did not anticipate the deregulatory movements of the 1970s and 1980s that loosened the hold of those interests; [FN168] nor, for similar reasons, did close students of U.S. corporations anticipate the vast changes in business organization beginning in the late 1980s. [FN169] In such turbulent circumstances, the only reliable prediction is that opposition can be overcome, if at all, only after a great struggle.
Amidst that predictable fight, directly deliberative polyarchy will only progress if benchmarking and the complex of institutions for its monitoring are established on a large scale and ultimately attain national scope. The need for such institutions was anticipated in the earlier discussion of learning by monitoring in firms: Benchmarking requires a survey of possible comparisons, evaluation of possible metrics for ranking the comparables, and revision, when necessary, of initial choices of both. Furthermore, the effectiveness of such surveys, evaluations, and revisions depends on the willingness of all participants to disclose information in view of the investigations of the others. That is why, we noted, firms often turn to third parties to organize benchmarking, and the organizers secure participation by various combinations of inducements and threats: disclosing, for example, the pooled information only to those who add their own results to the pool, or expelling members who refuse to cooperate in information gathering. Likewise in government, after easily accessible, informal opportunities have been exhausted, most jurisdictions willing to benchmark will nonetheless be unable to organize the activity themselves, and some jurisdictions‑‑or at least their leaders‑‑will be unwilling to exchange information for fear of showing poorly in comparison. A crucial role of the national institutions in an experimentalist regime is to assist the willing and oblige the unwilling to provide the information citizens require for direct deliberation.
There is a straightforward connection between progress in the construction of a national system of comparisons and the outcome of the fights to extend the reach of directly deliberative polyarchy. The more extensive and accurate the national information pooling, the more likely that innovative solutions will be widely known, and the greater weight the novel alliances formed through these solutions will have in local and national debates about adopting experimentalist measures. Put another way, national information pooling reduces the vulnerability of the vulnerable, and, to the extent that this is so, the more effective the system, the more support there will be for augmenting it.
The second, indirect connection between creation of a national framework and the prospects of experimentalism points to the limits of the analogy between constitutional reform and economic reorganization *339 and underscores the distinctive and fundamental importance of higher‑order, national institutions to the former. For employees and owners alike, the possibility that successful, day‑to‑day operation of current arrangements may generate fundamental errors that those arrangements cannot detect is a business risk worth running if current returns are high and the chances of timely improvements in error detection good. Indeed, if a single firm adopts the new methods while its competitors do not, the firm stands to benefit from the comparative advantage conferred. But constitutions are constitutions in the sense of foundational principles of a political order only if they consistently establish the institutions of normal political decisionmaking and the institutions that repair breakdowns in such decisionmaking. By establishing these higher‑order institutions, constitutions entrench the conceptions of political justice, thereby constituting the polity as well as the government. To pursue a strategy of decentralized design and production in anticipation of corresponding governance mechanisms may be prudent or rash. To adopt cognate forms in the polity‑‑subverting what is left of representative democracy without providing some coherent picture of the emergent regime‑‑gambles that crucial part of our identity that derives from citizenship in a democracy in reckless ignorance of the stakes. Hence, it should not be surprising that the little evidence there is suggests that, even in those isolated cases where experimentalist provision of public services has been broadly and successfully introduced under conditions of near emergency, practical success does not long forestall insistent questions regarding the relation between the new forms of participation and the old. [FN170]
Thus, either from the point of view of calculations of advantage that tip allegiance in struggles over the desirability of new methods or from the vantage point of large ideas that frame, through citizenship, our political and collective being, the project of democratic experimentalism is incomplete and fragile without preliminary indications of the national ends towards which reform should be directed. To be sure, change may proceed by bootstrapping, as local incrementalism leads to complementary reforms at higher levels. However, even if elements of traditional representative democracy long coexist with elements of a new participatory one, it is essential to present at least a tentative account of the envisioned national framework of experimentalism as a whole. To that task we now turn.
V. The National Framework
Under current conditions, the national lawmaking apparatus performs the Penelope labor of undoing with one hand what it does with the other. To assure democratic deliberation for the common good, an elected Congress has authority to make decisions of national importance; but to take account of the legislature's bounded rationality, Congress delegates *340 its power to administrative agencies or shares its power with the states.
In an experimentalist democracy, in contrast, decisionmaking is from the first presumptively decentralized, hence adjusted to local circumstance, and fragmented, for rules originate in the deliberations of distinct local governments. The principal role of the national government in domestic affairs, accordingly, is to encourage and coordinate this decentralized decisionmaking, and to protect citizens against abuses of power‑‑especially, and paradoxically, those abuses that may result from or be exacerbated by the pulverization of central authority itself. The task of the legislature is to authorize these deliberations and finance the ensuing experiments where local resources are insufficient to do so. The task of the administrative agencies is to provide the infrastructure of coordination, again where local resources are insufficient. The courts in this system are charged with the familiar tasks of policing government and safeguarding rights.
As experimentalist government is government by direct deliberation, these judicial activities are now more conspicuously than ever in the service of the common end of increasing citizen participation in political decisions, and especially in making forms of participation that produce effective results in one jurisdiction available to others in which they are applicable. But policing and monitoring are here also more directly connected than under the traditional approach. The judiciary draws on the experimentalist capacities of the system it superintends and itself uses experimentalist methods in interpreting the law. Information furnished by the system of decentralized learning organized by the administrative agencies informs the judiciary's initial findings; it treats these findings as a framework within which local governments and other rule makers elaborate solutions that meet their needs, provided they explain how local variants respect the framing principles and demonstrate by performance benchmarking that the solutions work as they should in principle. Thus, constitutional considerations are instinct in political deliberation, and the judiciary‑‑like the agencies and the legislature‑‑guides deliberative democracy from within.
This Part presents our familiar national institutions as parts of a system that complements and completes local experimentalism. The Parts that follow explore them further from the vantage point of an experimentalist judiciary recasting and solving in novel ways the dilemmas that give rise to the traditional problems of federalism, separation of powers, and judicial protection of individual rights. Our description of the transformations that democratic experimentalism would require of American national institutions does not minimize the scope of change we envision. At the same time, however, the transformation envisioned has deep resonance with the values of American democracy and is partially anticipated by countercurrents within the tradition of American public administration.