The New 'Civic Virtue' of the Internet:

A Complex Systems Model for the Governance of Cyberspace

David R. Johnson
Co-Director, Cyberspace Law Institute

David G. Post
Co-Director, Cyberspace Law Institute
and Associate Professor of Law, Temple University Law School

[Published in:The Emerging Internet (1998 Annual Review of the Institute for Information Studies) (C. Firestone, ed. 1998)

[Back to DPost home page]

Introduction

A great deal of talk these days concerns "wrongful" conduct in cyberspace--rampant copyright infringement, easily accessible online gambling sites, widely disseminated pyramid schemes, and pornographic images available at the click of a mouse to any Net-savvy fourteen-year-old. What to do about all this wrongful conduct is unclear--it is, after all, hard to arrest an electron. Law enforcement (understood in the broadest sense to encompass both public and private law) is unusually problematic in the special environment of cyberspace; not only can it be difficult to locate an anonymous or pseudonymous Internet user within any specific territorial jurisdiction, but the global nature of the Internet decreases the likelihood that the parties to online disputes will be subject to control by the same territorial law enforcement entities.

These are serious enforcement issues, deserving serious attention. But even more important questions must first be considered: Who should set the rules that apply to this new global medium? What polity or polities should function as sources of legitimate and welfare-enhancing rules for conduct on the Internet? Who should become the lawmakers of cyberspace?

One obvious answer is that cyberspace should be controlled by those same territorial sovereigns who set the rules governing conduct off line. In this chapter, we argue that this most obvious answer may well be wrong. We question whether a governance system divided into territories demarcated by physical boundaries can achieve key governmental goals in an environment that decouples the effects of conduct from the physical location in which that conduct occurs.

This is certainly a disorienting and disturbing suggestion, and we do not make it lightly. After several millennia of experimentation, civilization has produced models of governance with which many people are--justifiably--rather pleased. And as the world nears the close of the twentieth century, territorial sovereigns, by and large (and with notable exceptions, to be sure) seem to be embracing democratic values that responsibly serve the interests of their citizens. Substantial discomfort can be expected to accompany any suggestion that existing governance mechanisms will not work well in the new online environment, and that the nature of governance and what we call "civic virtue" themselves may change in fundamental ways (at least as applied to online conduct that principally affects others engaged in online commerce and communities).

But new conceptualizations necessitated by the failure of traditional, geographically based governance mechanisms in the online context may well lead to the emergence of new means of producing a just ordering of human affairs online--means that may in turn lead to new and generalizable insights into what constitutes superior design in collective decision-making processes in other spheres as well. As the authors of this chapter have argued elsewhere,(1)

the establishment of rules specifically designed to apply to online spaces, and the recognition of the legal and practical significance of the electronic boundaries that define, separate, and shield particular online spaces, provide the greatest hope for establishing an effective and legitimate governance system for the Internet.

This chapter, applying recent work involving the study of complex systems, develops our contention that allowing the Internet to evolve laws of its own (in a manner that does not use "representative democracy" of the type we value in the context of geographically defined sovereigns) will systematically produce a better means of finding optimal solutions to "collective-action" problems involving activities online. Representative democracy is grounded in the notion that participation in deliberations on the public good both requires and leads to civic virtue. Men and women (whether acting as voters or representatives) are ennobled by the process of casting aside narrow, selfish, or factional interests and putting themselves in the special frame of mind known as "good citizenship."(2)

Any scheme for governance of the Internet that is not grounded in representative democracy's traditional concepts of citizenship and civic virtue (which have historically been applied within geographically defined polities) must show itself to be superior to governance relying on "high-minded" public dialogue regarding policy issues and/or a system whereby citizens vote for representatives who can themselves engage in such deliberations.

A preview of our conclusions is in order at the outset: Rather than relying upon even the best of our democratic traditions to create a single set of top-down laws to impose on the Internet, would-be regulators of cyberspace should instead foster the emergence of diverse and contending rule sets that "pull and tug" against each other (and that help to recruit or discourage potential participants in particular online spaces) in order to allow an optimal overall combination of rules to arise. Rather than relying on the ability of citizens of the global electronic polity to debate thoughtfully in search of a single shared vision of the common good, would-be architects of online governance systems should look for a form of civic virtue that can tolerate continuous conflict and can reside in the very architecture of a decentralized, diverse, complex adaptive system.

The best available solution to conflicts in individual goals and values regarding online conduct may be found by allowing individuals to join distinct, boundaried communities on the Internet, each with its own divergent set of rules, and by allowing those communities to deal with external pressures by devising their own mechanisms for filtering out unwelcome messages and with internal conflict by easing (or requiring) exit. Democratic debate and traditional legislative action may not, after all, be the best way to make the best public policy for the Internet. If we can preserve individual liberty to make educated and empowering choices among alternative online rule sets, our most thoughtful and high-minded collective-action option may be to abandon the process of elections and deliberations regarding some single best law to be imposed impartially on all from the top down. We may instead find a new form of civic virtue by allowing the governance of online actions to emerge "from the bottom up" as a result of the pull and tug between local online "jurisdictions" that do not attempt to act in a dispassionate or disinterested or "public-spirited" manner.
 

Governance as a System-Design Problem

Imagine designing a new governance system from scratch for a particular population. Central to the task would be the creation of a set of processes and institutions enabling its citizens to seek the collective good. Designers of any system of governance need to define what is and is not "wrongful," create enforceable rules to govern conduct, and set up mechanisms to enforce those rules. They need, in short, to solve the ever-present problem(s) of collective action.

To survive and to produce sound policies, the proposed mechanism of governance would need, substantially, to satisfy two key design constraints. The first constraint, legitimacy, serves to limit a governing authority's right to coerce its citizens--and thereby helps to assure long-term support. Because a major function of governance is to prohibit wrongful self-interested conduct, governance necessarily involves an exercise of coercive authority; it is precisely the prospect of individual "defection" from agreed-upon rules that calls for governance in the first place, and it is to the common good to have cooperative rules enforced against those who would defect from those rules. But to limit the tyrannical exercise of coercive power, and to keep it from becoming oppressive, some principle along the lines of "consent of the governed" must be adopted: Those whose conduct is to be controlled by particular law-making exercises must have some "voice" (directly, or through some form of representation) in determining the substance of the laws.

The second design constraint, internalization, limits a governing authority's ability to make laws whose effects "spill over" onto those outside its "territory." Because a geographic territory's laws may have large-scale effects, some of the people affected by the actions of its citizens may be located outside the boundaries of that territory. For example, to use a common illustration, sovereign X's rule regarding factory emissions may, because of the ability of those emissions to cross geographic boundaries, affect the well-being of individuals living within the jurisdiction of sovereigns Y, Z, etc. Governance mechanisms must limit this problem of "spillover" if they are to produce general welfare-enhancing rules.(3)

This is because the presence of substantial spillovers (or externalities) tends to diminish one's confidence that rules adopted by individual sovereigns will be welfare-enhancing in the aggregate, for it is only to be expected that rules with negative external effects (i.e., rules whose costs are largely borne by outsiders) will be produced at the expense of rules with positive external effects (rules whose benefits are largely realized by outsiders). To protect against substantial negative effects of spillover, then, requires the adoption of a second principle: Those who feel the effects of conduct regulated by particular rules should, to a substantial degree, have a voice in determining the substantive content of the rules.

These two design goals--that both those who are controlled by and those who are affected by particular law-making exercises should have a voice in determining the substantive content of those laws--define two sets of individuals with a "stake" in the law-making process. A well-designed governance system, we assert, as have many before us, should give those two sets of individuals the predominant voice in the design of those laws.

Legitimacy and Internalization in the Real World: The Roles of Physical Clustering and Geographically Defined Citizenship

In the "real world"--the world of atoms--the system of law-making sovereignty is generally defined in geographical terms. That is, both the power and the right to enact and enforce laws are lodged with sovereigns whose domain of authority is circumscribed in territorial terms--applicable within a particular geographic area to persons (real or fictional), events, and conduct located there.(4)

Defining sovereignty in terms of geography means that the sovereign can legitimately exercise coercive power only over those individuals bearing some relationship to the particular geographical territory over which the sovereign has control.

In the case of designing a government for a "real" world of geographic sovereignty, the legitimacy-related design goal can be satisfied by incorporating a design element we might call geographically defined citizenship: Individuals who bear some relationship to the geographical territory over which the sovereign exercises control should have a voice in the sovereign's exercise of that law-making power. In this way, the set of individuals who have a stake in the content of the sovereign's laws arising from the sovereign's right and ability to exert coercive power over them corresponds to the set of persons who, through the constitution of law-making institutions, control the exercise of that power.

If the effects of conduct taking place at a particular location are felt primarily by those individuals located geographically at or near that location, then the second design goal--internalization--can also be satisfied by defining the "real-world" set of lawmakers in geographic terms. Where this assumption of physical clustering holds true, the set of individuals collectively deciding the rules applicable to particular conduct--the individuals located within sovereign X's boundaries--will substantially overlap the set of those individuals whose well-being is primarily affected by those rules, because those latter individuals are also predominantly located within sovereign X's boundaries.

Geographical clustering of the effects of physical conduct is thus a sufficient condition for territorially defined citizenship simultaneously to serve both of the basic goals of well-designed governance. Of course this is no accident. Both the ability to ascertain the will of a proximate people and the ability to affect the welfare of those people by means of enforceable rules arise traditionally from communications technologies that operate best at close range and on enforcement by means of physical forces that work best within shorter distances. The world in which the impact of governmental action damps down or attenuates over distance is also a world in which individual actions affect most intensely those who are physically closest to the actor.

Where the assumptions of physical clustering and geographically defined citizenship hold, the act of constituting authoritative sets of lawmakers geographically allows reasonable certainty that the two key design goals for good governance will be met. In this setting, the major concerns for enhancing "civic virtue" will relate to the quality of the conversation among those geographically defined citizens, the thoughtfulness of representatives regarding the "public good" as defined with reference to this geographically defined "public," and, perhaps, some thought about the optimal size of geographic territories that can appropriately be brought under a single government's rule.(5)

But what of the goals of legitimacy and internalization as applied to the governance of cyberspace, where the effect of conduct is not contained by physical clustering? Does geographically defined governance make sense for such a realm?

Legitimacy and Internalization in Cyberspace: The Absence of Physical Clustering

Throughout most of human history, human interaction has been governed by laws based on geographically defined citizenship and physical clustering; rules applicable to conduct in, say, a medieval manor or a tribal village had little effect on individuals physically located elsewhere. But it is already a commonplace to observe that this state of affairs has begun to crumble in recent years as a result of the proliferation of cross-border externalities of all kinds. In an increasingly interconnected world, conduct occurring "there" (wherever "there" might be) increasingly affects those located "here"; yet people "here" who are affected by that conduct may have no voice in whether such conduct may lawfully occur, because that decision is left to the citizens of "there."

Although people still live in geographically defined territories, cyberspace has accelerated humanity's transition to a world in which physical clustering virtually disappears, at least with respect to the kinds of activities typical in an online environment.(6)

Cyberspace does not merely weaken the physical clustering assumption (that conduct taking place at a particular geographical location primarily affects persons at or near that physical location); cyberspace renders the assumption virtually unintelligible. Nor does cyberspace merely weaken geographical boundaries, it obliterates them entirely (at least in cyberspace), because geographical location itself is both indeterminate and irrelevant for transactions on the Internet.

Location is indeterminate because no necessary relationship exists between electronic addressing--such as Internet Protocol addresses, domain names, and e-mail addresses--and the location of the addressee (machine or user) in physical space.(7)

It is, perhaps, conceivable that the entire Internet could be redesigned in some way to alter addressing functions so as to retain, or even emphasize, their connection to geographical reality.(8)

But the absence of any connection to geographic location is a fundamental and immutable characteristic of other network interactions; distributed databases (such as Usenet newsgroups), for example, have no single place of existence at all, but are aggregates of information stored on a large (and constantly changing) number of individual machines at any time.(9)
 

Location is also irrelevant in cyberspace in the sense that network servers and online addressees are equally accessible from everywhere. Any Web site in any odd corner of the network can be accessed with essentially equivalent transmission speed and message quality from any other corner of the network, which means that the effects of whatever information is available at a given site are felt simultaneously and equally in all jurisdictions, independent of their "distance" from one another. This is a function of the speed at which information travels (and "information," strings of bits, is all that does travel in cyberspace) and of the digital character of that information, which does not decay as a function of time and distance.

If physical clustering is not thus a valid assumption with regard to activities on the Internet, important questions arise with regard to governance and cyberspace: Is physical clustering a necessary condition for the use of geographically defined territories to achieve the two basic design goals of legitimacy and internalization in governance? If physical clustering is absent, must we abandon geographically defined governance structures in order to serve those key goals? While cyberspace's "inhabitants" have physical bodies that reside in geographically defined territories, the online world itself exists without physical clustering of effects (i.e., it is a world in which the effects of particular actions, as well as the effects of the rules that regulate them, are randomly distributed with respect to geographic location). Lacking the organization of effects provided by physical clustering, can geographically defined citizenship (and rule-making) simultaneously serve the goals of legitimacy and internalization in such a space as cyberspace? We believe that they cannot.

To illustrate, consider a simple hypothetical scenario: Residents of one country--say France--place a World Wide Web page on a server in that jurisdiction. The Web page displays content created by residents of a different jurisdiction--say the United States--in a manner (a) consistent with French copyright law but (b) inconsistent with U.S. copyright law.(10)

Whose law applies to this conduct? Applying French law in this case appears to comport with design goal #1 (legitimacy), inasmuch as the offending (and regulated) conduct takes place within French territory. But it contravenes design goal #2 (internalization) inasmuch as the rule primarily affects the content's authors, who are located in the United States. Conversely, application of U.S. law comports with design goal #2, inasmuch as the authors of the content reside in U.S. territory, but violates design goal #1 because the French actors are then bound by a law they had no voice in making. The same dilemma arises in reference to any action on the Internet that involves conduct viewed differently by differing sovereigns (trademark violations, publication of "offensive" content, defamation, etc.), whenever the originating parties and those affected by their acts reside in different countries. The dilemma is not solved whether sovereigns opt to impose law territorially (and thereby contravene the principle that individuals should have a voice in rules that affect them) or extraterritorially (and thereby contravene the principle that individuals should have a voice in rules that are applied to them).

As a result of an increasingly interconnected world, dilemmas of this sort were already becoming more frequent even before the invention of the Internet. The recent--and ongoing--explosion of the World Wide Web, however, has brought the problem to the forefront: How can the principles of legitimacy and internalization be substantially satisfied in a bit-based world in which the inherent overlap of control and effect accomplished by physical clustering, as in atom-based phenomena, can no longer be assumed?
 

Clustering And Rule-making in Complex Systems

To study the implications of a world in which the effects of conduct extend beyond geographical boundaries ("spilling over" into other jurisdictions) and participation in rule-making is decoupled from the effects of such rules, we have created a computer-based complex system and studied its behavior under various circumstances.(11)

Our computer-based complex system is based on seminal research conducted by Stuart Kauffman and his colleagues at the Santa Fe Institute in New Mexico.(12)

Kauffman's original system can be understood in terms of a collection of individual elements--light bulbs, say--each of which can be "on" or "off" (i.e., each of which may be in one of two possible "states"). Each individual element has a "fitness" rating (we could say "happiness" or "utility") that is partially determined by its own state (on/off); that is, some light bulbs "prefer" to be on, others "prefer" to be off. In addition, however, each light bulb's fitness is also a function of the on/off state of some number of other light bulbs; that is, spillover effects exist by which the state of any particular element affects the fitness of many other elements and the fitness of any particular element is affected by the state of many other elements.

Suppose we are looking for the particular configuration of on/off settings for all light bulbs that produces the highest aggregate fitness for the system as a whole. If we simply try to place each light bulb in its individually "preferred" on/off state, we run into immediate trouble, because the interests of the individual light bulbs conflict; for example, one bulb's preferred state may negatively affect the fitness of large numbers of others. Next we could try, systematically, to examine all possible configurations of the light bulbs (all combinations and permutations of on/off settings for the entire matrix) to find the best configuration (i.e., the one producing the highest aggregate fitness). But this procedure is computationally intractable; with even a small system of, say, 100 light bulbs, we would have to examine 2100 different configurations. (The many different possible combinations of interdependent states create what Kauffman calls a "rugged" fitness landscape--one that produces many peaks, foothills, valleys, and chasms, and one that (1) does not have a reliable or smooth "uphill" direction, and (2) is simply too large to explore fully by trudging one step at a time across the combinations of settings that produce, at any point, its topographic "height.")

We could, in desperation, begin to switch randomly chosen light bulbs from one state to another and observe whether the change increases, or decreases, the overall (aggregate) fitness of the system. But, as Kauffman and others have demonstrated, this approach--this step-by-step "algorithm" for searching out the best "policy" for attaining the fitness of the matrix as a whole--often gets stuck at a low level of overall fitness; whenever the system arrives, randomly, at the top of a low foothill on the fitness landscape, it cannot dislodge itself because every possible one-step change in state leads the overall fitness (temporarily) downward off the hill.

Kauffman's experiments with these models have demonstrated the effectiveness of one particular algorithm for finding high aggregate fitness for complex systems, one that mediates between the local ("selfish") interests of the individual element (e.g., light bulb) and the global ("public") interests of the total system. To use Kauffman's approach, we begin by dividing the entire population of light bulbs into discrete "patches"--groups of neighboring elements. (As we will explain in more detail later in this chapter, the size and character of these patches do have an important impact. In his own experiments, Kauffman selected patch sizes at random and examined the results). Now, as we change the state of a randomly chosen light bulb, we monitor the effect of that switch on the aggregate fitness of that bulb's patch. We then judge any particular "move" by assessing only the resulting well-being of the light bulb's patch--not the impact on the bulb itself and not the impact on the matrix as a whole. That is, if the move causes patch fitness to decline, we return the light bulb to its initial state, even if the move would have benefited either or both the individual and the matrix as a whole; conversely, if the move causes patch fitness to increase, the change of state is allowed to stand even if the move disadvantages the individual and/or the matrix as a whole. As we continue to switch randomly selected bulbs in this way, the overall fitness of the matrix increases--to values higher than those achievable with other methods.

Kauffman has demonstrated that allowing patches to use this method to optimize their internal fitness is a highly effective way to find a high aggregate fitness for the system as a whole (and, of course, for the "average" individual element). The algorithm works because it allows "selfish" decisions by local patches, configurations that can drive the fitness of the overall matrix temporarily downhill. (This effect stems from spillover impacts across patch boundaries.) Over time, however, these selfish patch decisions help drive the system as a whole to positions of higher overall fitness (as all patches continue to seek optimization within the context of other patches' "decisions"). Kauffman's patches work better than algorithms that confirm moves rewarding only the individual or the system as a whole. And they work even though--indeed, because--some of the effects of the patches' decisions fall on "neighbors" who are not "voting" members of the patch. As should be apparent, Kauffman's work thus suggests that--in the context of complex systems with many elements that exhibit many conflicting goals--algorithms that define the problem in terms of finding moves that benefit either individuals or the "public" as a whole may not be optimal. And, as we will discuss at greater length, Kauffman's work shows that some degree of spillover (and its resulting "illegitimacy") may not be a bad thing.

That such a system of decision-making by patches that pull and tug against each other is apparently the most effective way to find the highest aggregate fitness for an entire collection of individuals with conflicting goals demonstrates the value of a primitive kind of decentralized federalism. It suggests that the welfare of all may, on average, be maximized not by allowing each individual to seek its own good, nor by deciding centrally what appears to be best for the whole. Optimal solutions may come (in a complex-systems context) from a particular structure that seems, superficially, to make "wrong" collective-action choices from the perspectives of both the entire polity and some individuals negatively affected by those decisions. This is because complex systems produce rugged fitness landscapes, in which paths to an optimal set of results or policies may lead "downhill" from a suboptimal (but locally highest) foothill.(13)

We believe that, as a highly abstract representation, this model captures certain essential features of the existing international legal order. In that context, the system's elements are not light bulbs but a large number of individuals, each of whom derives some subjective happiness, or utility, as a consequence of (1) being permitted or (0) not being permitted to undertake a particular action.(14)

Individuals' personal utility is a function not only of their own states (act/don't act), but also of the states (actions) of many other individuals; that is, utility is a function both of whether the individual is permitted to take an action that affects others and whether neighboring individuals are permitted to take actions that affect the individual (and others). Kauffman's "patches" are, in effect, territorially local sovereign political groups; each group decides whether conduct by its individual group members will, or will not, be permitted, and each group makes that decision on the basis of the aggregate well-being of its own members, without taking into account the effect of that conduct on persons outside the group.

If we want to apply the lessons from Kauffman's model to cyberspace, how should we organize and implement this decentralized, "federalist" decision-making algorithm? The decision of how to divide the online population into "patches" is critical. Must the patches somehow be defined in terms of geographical "neighborhoods"? If so, how can "patches" of decision-makers be created in a online world in which effects (and involvement in rule-making) don't necessarily cluster geographically? We built a version of Kauffman's model in an attempt to answer the following question: How much does the effectiveness of this patching algorithm in a complex-systems model depend upon the overlap between the set of those whose conduct is regulated by a given "patch's rules" and the set of those who are affected by the impact (spillover effects) of the actions permitted by a patch's rules? We have returned, in other words, to the original question--whether physical clustering is a necessary prerequisite for the use of geographic sovereignty as the basis for governance of online activities--but we can now analyze that question in terms of the behavior of Kauffman's computer-based complex systems, under carefully controlled conditions.

In Kauffman's original model, spillovers and patch membership were both determined by a light bulb's physical location within the overall matrix--i.e., geographically defined "citizenship." Thus, Kauffman's patches inevitably consisted to a substantial degree of groups of light bulbs whose fitness was substantially affected by the state (action) of other patch members. We asked: Suppose we systematically unhook the "spillover sets" (the elements that affect one another's well-being) and what we can call the "political patches" (the elements that collectively decide whether particular conduct is permissible). That is, what if, at one extreme, we allow light bulbs to affect the fitness only of other light bulbs who cannot "vote" on those moves (whose "fitness" is not counted to determine if the "patch" is better off)? And what if, at the other extreme, we only allow light bulbs to affect the fitness of other bulbs that are members of the same political patch, so that all decisions on individual moves are made only by assessing the well-being of the bulb making the move and the other bulbs affected by those moves? We asked: Will the patching algorithm still work to maximize overall system well-being when the "internalization" (spillover) and "legitimacy" dials are twisted to one extreme or the other?

Our results were quite striking. When we set up the system so that bulbs only affect others who can vote to roll back their moves, we got lower overall fitness results. When we set up the system so there was no connection between those who voted and those affected by actions permitted by such votes, we also got lower aggregate fitness results. "Congruence"--the amount of overlap between the effects of individual elements' actions and "political" patch membership--turns out to be a critical determinant of the patching algorithm's success. Systems that are set up with very low congruence (that is, the effects of an individual's change of state are felt primarily by elements outside its decision-making unit) do systematically worse at solving the collective-action problem of finding a high overall fitness score (maximizing average and aggregate "utility," considering the state of all elements). Perhaps even more interesting, however, is our finding that too much "congruence" also makes patching less effective; if an individual's conduct affects only the other members of its decision-making group, then dividing the system into patches does not effectively drive the overall matrix towards a better solution--because there is no way for the patch to push the overall matrix down a hill in the overall "fitness space" and thus no way the matrix can climb up a ridge to a higher hill. Thus, in light of the benefits of "patching," neither perfect "internalization" nor perfect "legitimacy" are optimal for the governance of complex systems. Effects and rule-making authority need to be connected--but only to an optimal degree. (We found the optimal level at approximately eighty to ninety percent overlap, or congruence, in our system, given the particular spillover settings and matrix size we used in our experiments.)

Our computer-based complex system and Kauffman's groundbreaking work thus provide strong evidence that the best way to solve the complex problems of collective action by interdependent groups with conflicting goals is to cut polities up into smaller "patches" that are defined in a way that preserves some optimal degree (neither zero nor 100 percent; approximately eighty percent, in our findings) of congruence, or overlap, between those who are affected by particular rules and those with a voice in setting those rules. A world of geographically defined decision-making patches makes sense as a means of solving the problem of finding optimal rules, but only if the effects of individual conduct are felt primarily within those same geographical boundaries, i.e., only if there is a relatively small (approximately twenty percent) degree of cross-border spillover. The physical clustering assumption thus may not only be a sufficient condition for the effectiveness of geographically defined governance, it appears to be a necessary condition if we want governance to maximize utility in aggregate.(15)

But because physical clustering is a condition that does not hold for most online activities, we need to look to other mechanisms that can, for cyberspace, preserve (or, more accurately, reestablish) an optimal degree of congruence.
 

Reestablishing Congruence in Cyberspace Governance

So far this chapter has examined governance as a design problem substantially constrained by the two key goals of achieving substantial legitimacy and internalization of spillovers. The time-honored design solution provided by the physical clustering of effects in the "real" world of geographically defined citizenship seems inadequate for solving the governance problem in the online world, where the physical clustering of effects can no longer be taken for granted. This has created a seemingly intractable dilemma for territorial governments.

The behavior of computer-based, complex-system models helps us to understand and analyze both the new phenomena presented by cyberspace and the initial reactions of territorial governments in response to these phenomena. The Internet, we observe, dramatically reduces the "congruence" between rule-making and spillover impact sets that otherwise naturally exists in the real, atom-based, world by allowing the actions of an individual located just about anywhere on earth (whose conduct is now regulated by territorially local laws) to affect any number of others located just about anywhere. Now that spillovers no longer fall predominantly on individuals belonging to the same geographically defined patch, rule-making decisions made by these patches--sovereign states--are less likely to find the highest peaks on the global policy fitness landscape. A central question for Internet governance, then, is: How can an optimal level of "congruence" be reestablished in the online world?

Suboptimal Ways to Reestablish Congruence

A number of remedies for repairing a loss of congruence within the Internet are already being advanced. Each of them, however, has significant limitations or negative consequences:

1. Treat the entire system as a single patch. To eliminate low patch/spillover overlap in any system, all external spillover effects could be internalized by treating all individuals as members of a single, system-wide patch. This "single patch" strategy is reflected in increasingly frequent calls for a global governance mechanism for the Internet to achieve a "harmonization" of substantive rules regarding particular activities in cyberspace (and thereby take into account the geographical-boundary­destroying character of the Internet).(16)

Global harmonization may indeed remove some of the negative effects of spillovers. But this strategy also eliminates the benefits that the decentralized patching algorithm might otherwise provide. A single-patch (harmonized) system will tend to get stuck on low local optima because decisions by subsidiary patches cannot override the central consensus, and cannot thereby drive the matrix towards a better solution. A single system of global governance for the Internet might be legitimate (if all participants had a say) and would necessarily internalize all costs and benefits within a single decision-making entity. But it would still not be the best algorithm for collective action, because it could not make use of the benefits of federalism--the ability of local experiments to cause reactions and adjustments by adversely affected "neighbors," creating a pull and tug among conflicting rule sets that ultimately causes the overall matrix to achieve a better solution.

2. Recouple spillover sets and patch membership by filtering spillovers so that they fall more clearly within geographic patch boundaries. If "political patch" boundaries are regarded as geographically fixed, a recoupling of spillover effects and patch boundaries might be achieved by controlling spillover effects at those geographic boundaries. Filtering spillovers so that they fall more clearly within geographic borders--that is, interposing existing geopolitical patch boundaries on the flow of information across the network so as to eliminate unwelcome effects spilling across those boundaries--is a strategy used by many territorial governments. The government of Singapore, for example, has at times attempted to control World Wide Web content accessible to its citizens by means of a "proxy server" system that permits access from within Singapore only to previously downloaded copies of "approved" Web pages.(17)

German authorities try to disable access by German residents to certain global Usenet newsgroups deemed pornographic.(18)
 

Implementation of geographically based spillover-filtering systems in cyberspace is likely to be costly in the extreme and quite possibly futile, at least under cyberspace's current network architecture, which allows the routing of messages without reference to locational information.(19)

More importantly--the ultimate futility of such measures aside--this strategy is counterproductive, as it has as its aim the elimination of interpatch spillover effects that are, from other perspectives, the new medium's most obvious strengths. Converting the global marketplace into a series of geographically defined local flea markets solves the "excess spillover" problem, but only by denying everyone the chance to participate in global electronic commerce, which inherently consists of widely scattered spillover sets. In addition, this "solution" tends, like the harmonization approach, to create isolated, totally congruent patches, thereby reducing system search efficiency (losing the benefits of patching) and, in effect, destroying the Internet in order to save it.

3. Recouple spillover sets and patch membership by redrawing geographic political patch boundaries so that they encompass foreign sources of spillovers. If political law-making and law-enforcing boundaries are not viewed as fixed, then patch boundaries themselves--in the form of the legally operative borders of nation-states--might be adjusted to track spillover effects. As in our earlier example of a hypothetical U.S./France dispute over copyright law, we could extend the power of local governments to cover any participant in an online space who has any impact on the local territory (whether or not "targeted"). But this would not repair congruence fully or symmetrically. Allowing Germany, for example, to set rules applicable to everyone whose actions in an online forum affect any German citizen checks the "spillover" onto German interests, but disserves the goal of legitimacy because "foreign" participants in such online spaces do not have a say in establishing German law. Extending the reach of local jurisdictions can check spillovers, but, absent more radical changes regarding participation in local law-making, the natural two-sided "congruence" found in geographically based governance of the tangible world is not repaired. In addition, conflicts among the rules of local jurisdictions inevitably subject individual actors to multiple, inconsistent, purportedly equally binding rules of law.

This third category presents an option all too popular with local governments and is most obviously illustrated by attempts by existing sovereigns to extend their authority to encompass activities--wherever located--whose effects spill into their jurisdictions by asserting a "jurisdiction to prescribe" in respect to outside-the-boundary activities. An example is the attempt by Minnesota to regulate gambling that occurs on a foreign Web page because it can be accessed and "brought into" the state by a local resident.(20)

In this case, a single "patch" is attempting to govern the actions of those outside of its borders when those actions have, in the view of the local patch members, a detrimental effect on the fitness/happiness of the patch's citizenry. The local authority uses local effects as the basis for asserting its "long-arm" jurisdiction and applying its law.

While redefining patch boundaries by expanding a geographic patch's "jurisdiction to prescribe" does achieve a degree of recoupling, it does so only asymmetrically. Patching performs two separate functions: (1) patches are loci of decision-making processes, defining the members of the set whose preferences count in regard to determining the desirability of particular moves, and (2) patches are enforcement devices, empowered to "roll back" those actions/changes of which patch members collectively disapprove. Extending a geographical patch's jurisdiction to prescribe adjusts that patch's boundary only for the latter (enforcement) purpose; patch boundaries are extended only for purposes of exercising "vetoes" on actions by individuals outside the patch boundary, not for granting those individuals membership in the patch itself (i.e., those individuals are not permitted to participate in the determination of the acceptability of particular changes). In such a case, the patch boundary is thus made selectively permeable, a kind of one-way valve that grants only partial patch membership to those whose activities spill into its jurisdiction. This asymmetry can be expected to encourage strategic behavior on the part of the patch members that systematically disserves the liberty interests of external actors. Extraterritorial application of patch decisions not only threatens core "legitimacy" goals, it gives individual patches an incentive to attempt to impose regulatory burdens on those who cannot affect decision-making regarding the scope of those burdens--to export regulatory costs to individuals lacking a voice.(21)

A Better Way to Reestablish Congruence

We are aware of only one way to repair the needed level of congruence in the online world without eliminating the virtues of federalism implicit in patching or otherwise harming legitimacy and internalization goals. That way is to allocate decision-making regarding the rules applicable to particular areas of the online world to those people who are most affected by such activities--the system operator and the users who actually "inhabit" the online spaces in question. These "netizens" are the people for whom the rules regarding action in this online space have the greatest significance--for whom the rules are, in the most meaningful (and a non-geographical) sense, a "local" ordinance. Actions in this space may have some "spillover" effects on other real or virtual places. But, with rare exceptions, such spillovers are less important than the impact of the rules on those who choose to participate directly in such online space(s).

Insofar as online activities involve communications of information, and insofar as the areas in which such communications take place are separated by meaningful boundaries--login codes, habits of users who visit some spaces and not others, filters that screen out incoming messages, etc.--the online world is building a non-physical type of "clustering" of effects and controls.(22)

What is said in a Usenet newsgroup I never visit is less likely to have an impact on me--though it may occasionally "spill over" in various ways--than what is done in the online spaces I do choose to frequent. Rules designed to govern "conduct" on the Internet most directly affect those who take part actively in the online spaces where the rules apply. Thus, congruence between those whose actions are regulated, and those who are affected by those actions, can be "naturally" reestablished for cyberspace by respecting the natural electronic boundaries (filters and other separations of online spaces) that accomplish, for this new cyberspace, what physical separation and the attenuation of impacts and controls over distance have always accomplished for the geographical world.

We are mindful that some online activities have serious impact on those who are not participants in the online spaces whose rules permit the activities to occur. But there are long-distance impact exceptions that prove the rule (and justify self-defensive measures by a remote target) in the real world as well. In the general case of online spillovers, such as negative effects from offending messages, the ability to construct software filters that eliminate messages from specific sources should go far towards allowing any particular online group to avoid noxious spillover from another group that has differing views about what should be permitted online. Moreover, we must remember that the optimal level of spillover is not zero. We will learn a great deal, collectively, from the pull and tug among differing rule sets that contend for favor online--in part as a result of actions taken by differing groups to defend themselves against the negative externalities created by others with differing utility functions and/or political views.

Some Examples

Consider the problem posed by "spam"--defined generally as unsolicited commercial e-mail sent in bulk to many unwilling recipients. This practice poses two different types of challenges for Internet governance: (1) how to decide which spam is wrongful, and (2) how to enforce any such rule. If geographically based governments attempt to pass laws against spam (and the U.S. Congress is considering several bills right now), they may succeed only in driving prohibited practices offshore. If they attempt to extend their jurisdictions to cover any source of an unwelcome message received locally, the end result may be doctrines that significantly constrain the free flow of information on the Internet, that create substantial conflicts between local jurisdictions, and that fail to embody the symmetry (with regard to participation by the regulated in rule-making) implicit in the notion of legitimacy. Moreover, given the time it takes to pass laws and the difficulties in formulating authoritative texts to deal with fluid technical situations, any rules produced may quickly become outmoded as e-mail adds audio and video capability, as new filtering tools become available, and as messages begin to be exchanged among physical devices as well as human beings.

An alternative solution is to allow each online space to develop a set of rules of its own regarding both what messages to allow into local mailboxes and what conduct (in sending mail to others inside or outside the system) is permissible. Some online venues composed of those who hate getting spam will quickly adopt a rule prohibiting such conduct, on pain of banishment. (And they may, in some cases, expand the definition of the prohibited conduct to include non-commercial messages.) The sysops of those spaces will also try to establish effective filters against unwanted messages from outside, saving their members the time and effort involved in constructing personal barriers. In contrast, more "liberal" spaces may not bother to construct filters or, indeed, may tolerate spamming by members. A "spam haven" may evolve to expressly cater to those who send unrequested commercial e-mail for profit. But, from the perspective of remote victims, such havens may be easier to detect and filter out than a population of spammers randomly distributed across cyberspace domains. The spam havens may, to be sure, develop technology and practices designed to avoid the filters. But those who want to eliminate spam may also escalate their technological weapons and, in an extreme case, might band together to eliminate all routing of messages from the offending domain to their own areas of the Internet.(23)
 

This "pull and tug" of conflicting policies may lead to a complex de facto compromise based on the ability to attach labels to allow each side to deal only with others who follow policies acceptable to it. It will, in any event, surely allow the development of policies that (1) closely reflect the wishes of the users most affected (those who continue to frequent any particular online space because they endorse its local rules), (2) are capable of evolving rapidly in response to changes in underlying circumstances, (3) do not inherently (or de jure) unfairly impose one population's views of sound policy on another population whose actions do not impact the first, and (4) even allow a single individual to achieve a subtle mix of policies by means of spending differing portions of time in differing online spaces with differing policies. The result will not only achieve internalization goals (because it minimizes spillovers on those who find such spillover most noxious), achieve legitimacy goals (because it gives most say to those whose conduct is regulated by particular rules), and approach optimization (because its algorithm allows the kind of experimentation that leads more quickly to a high peak in policy space, even when perturbed by changing underlying circumstances), but it will also be personally empowering (because it allows individuals to decide whether and how to subject themselves to multiple differing rule sets.)

Spamming is a type of conduct that a disapproving "jurisdiction" can substantially reduce by means of filters. Can this decentralized model be extended to cover other types of actions whose impacts cannot so easily be filtered out? Consider Internet copyright rules, which affect authors by determining what other users of an online space may do with copies of the materials originating in that space, or defamation rules, which affect the reputations of those who may not frequent the online space where the defamatory comments are published. Even in these cases, diverse inconsistent rule sets could develop in a way that better serves the collective interest than traditional "top-down" democratic governance.

In the copyright case, areas liberally defining "fair use" (or, indeed, eliminating any prohibition at all against copying and subsequent distribution of an online author's works) would attract those who want to be free to reuse materials they find online and who would be willing to allow their own postings to be subject to such "liberal" copyright rules. (We are not here considering unauthorized "leakage" across boundaries but, instead, the various rules that might be applicable to texts originally posted with authorization in a particular online space with any particular set of rules regarding subsequent copying of such postings.) What we could expect, ultimately, is a sorting out of authors and readers and works into online spaces with rules that fit the combined preferences of specific groups of participants. Restrictive regimes will drive away (or expel, for rule violations) "liberally" minded readers. Liberal regimes will drive away authors who favor restrictive policies. And any effort on the part of either type of online "patch" to export costs onto the other will be met by retaliation in some form or other (such as rejection of visitors originating from, or downloads destined to, areas known to support disapproved policies). We end up with something that looks more like "foreign policy"--the familiar "pull and tug" among discrete patches--than representative democracy. But it might well work better, for all concerned, because it offers a diversity of differing rule sets as it preserves (1) incentives for those authors who need it (because they will practice their art only in areas that protect copyrights and prevent leakage) and (2) the free flow of information for those readers and authors who expect to prosper under more liberal rules.

The problem of online defamation, too, might be dealt with more successfully by a decentralized, emergent model of law-making. Assume that various people have thinner or thicker skins and worry to a greater or lesser degree about the free flow of robust discussion. Some will favor online areas that strictly limit apparently unreliable and derogatory comments. Some will favor areas that make no attempt to regulate false and damaging comments. Of course, a participant in the "defamation haven" might speak ill (and falsely) of an individual who never visits such dens of iniquity. Let's suppose the victim heard about it and that the comment had an actual substantial impact on the victim's reputation. If that occurred, and if the rule against defamation was taken seriously by the victim's own online venue (or local geographical sovereign), then the victim's sysop (or state) might be expected to seek to punish the offender--to prohibit messages or visits from the offending party or to seek to extract a remedy from the offending party by some means. An ugly online "war" might develop--but (assuming the victim's protective jurisdiction lacks physical control over the perpetrator and has no extradition power) its only casualties would be electronic messages. Perhaps a peace treaty among sysops might eventually provide that any victim of allegedly false and defamatory speech must be given an opportunity promptly to respond within the online context of the original message. Or perhaps some "defamation havens" would simply lose their power to inflict harms on reputation--because (1) most readers wouldn't choose to go there, and (2) the very lack of any ability to bring a defamation action against speech originating in such fora would be known to attract just those persons interested in making false and unreliable statements, thereby automatically reducing the perceived reliability of messages originating there and thus lessening their negative impact. The overall, complex result is likely to be a state of affairs in which the personal preferences of actors and victims, regarding the vigor with which reliability is policed and reputational harm is remedied, more closely match the rules of the spaces where they spend their time.

Of course, autonomous online decision-making could allow some online spaces to develop rules that allow small groups to take actions that threaten the vital interests of "foreigners." The virtual or geographically based governments of those affected foreigners may then feel the need to protect their own "local" populations from such adverse spillover effects. If an online forum is used to plot violence against a particular country, that country might be justified in declaring war against the online upstart, wherever it happens to be based. Real-world governments and online governors will, of course, always need to protect their local populations against threats that emanate from "foreign" sources--and they will use whatever "force" they have at their disposal to do so in appropriate cases. But online areas that adopt rules allowing actions that primarily affect only willing participants, that reduce external spillovers to modest levels, and that are populated by users whose decisions set rules appropriate for such online spaces, are entitled to a certain amount of "comity"--the use of the contract enforcement mechanisms of real-world (and virtual) governments to assure respect for and the security of their virtual boundaries. Most importantly, geographically based governments can and should begin to pay attention to the possibility that the tendency of the Internet to break down geographic borders and eliminate physical clustering of effects requires a whole new perspective on how to measure the virtues of any particular governance mechanism applicable to activities on the Net.
 

Congruence, Cyberspace, and Civic Virtue

We think that the most obvious path to reestablishing an optimal amount of legitimacy and internalization in cyberspace, while taking advantage of the virtues of federalism and of limited spillovers from local political patches, is to free cyberspace altogether from the notion that rules must be made (for online spaces) by institutions based on geographical sovereignty. To the contrary, the best "collective-action" results for cyberspace will be achieved by treating each separate online location as a distinct, largely self-governing place, with its own rules governing actions that primarily affect its own participants. (Those rules may be made in the first instance by system operators, but they will be ratified, in effect, by individual users' decisions to frequent the online spaces they find empowering.)

If rule-making is allocated to online groups in this way, other basic concepts will also need to be rethought, concepts that underlie notions of "civic virtue" in the context of traditional representative democracy and responsive geographic governance. Citizens of territorial democracies, for example, are accustomed to the idea that selection of representatives from a large territory can reduce the risks of factions, and that the number of representatives drawn from particular geographic regions must be set so as to allow constructive conversation in a real legislative chamber.(24)

This may not work well in the context of online governance, because there can be no neat sorting of one-citizen, one-vote in cyberspace and because the "territory" that needs to be represented in the rule formulation process may be very large. Indeed, the search for ways to reestablish "congruence" (and thereby optimal results) of governance in geographically decoupled cyberspace may lead to a whole new set of conceptions regarding online citizenship.

Current "real-world" concepts of citizenship may not fit well within a world whose "citizens"--known only by login IDs or e-mail addresses--may concurrently participate in many online polities. Fortunately, the very ease with which users can choose to exit an online space (in notable contrast to the high cost of leaving a geographically defined politically sovereign territory) provides an alternative mechanism for preserving a "voice" for "netizens" in the context of Internet self-governance. In essence, users will be directly subjected to the rules established by sysops only for those online places that they choose to visit and revisit. If users don't find local rules for an online space congenial, they can leave that space and, indeed--cheaply enough--create their own. They can also embed their time and attention in multiple online spaces, thereby finding just that mix of rule sets that best suits their needs.

Rules on the Internet will thus be subjected to popular will by means of "voting with one's modem" rather than by means of traditional balloting or the election of representatives. In the online context, the check against sysop tyranny is not "one person, one vote" but, rather, ease of exit. And there is reason to believe that the combination of decentralized rule-making by means of (1) the unilateral actions of sysops to define online spaces and (2) the unilateral decisions by users to join or leave such spaces will arrive at a good solution to the collective-action problem. This is because the establishment of rules for online spaces by those who participate in those spaces will both (1) provide the "federalism" benefits of the "patching" algorithm, and (2) insofar as most effects of action in online fora are predominantly felt "locally" within those fora, reconnect the "congruence" between spillover sets and political decision-making bodies.

Additional good reasons exist for believing that this decentralized solution to the Net governance problem will prove optimal: Individuals are in a better position than territorially elected representatives to know how actions taken within particular online venues affect their own personal utility. Therefore, if we reduce the governance role of territorial sovereigns, those individuals who are most intensely affected by, and who care most about, the effects of online actions can be relied upon to influence optimally the relevant decision-making processes.(25)

Using their newly granted ability to "enter" and "exit" political venues at will, the individuals most directly affected will have the most say in choosing the rules applicable to actions that cause such effects.

What emerges from this analysis is little short of a new definition of "civic virtue" in the online context. Based on the lessons from complex systems suggesting that multiple and competing rule sets may pull and tug each other down ridges and up higher hills in policy fitness space, we may not need to ask online citizens to adopt a frame of mind that sets aside personal or "local" interests in order to deliberate, directly or via representatives, regarding some globally applicable laws. Nor will we judge the effectiveness of Internet governance by its ability to foster rational and thoughtful deliberation among elected delegates who seek to define and serve some overall "public good." Rather, Internet governance can be judged on a systems design level by whether, and to what degree, it has adopted algorithms that effectively and substantially (but not totally) couple spillover effects with political decision-making, thus preserving congruence, and optimal legitimacy and internalization, while still allowing an adequate pull and tug among contending "local" constituencies.

Internet governance should be so constituted that the overall matrix of individuals with conflicting goals makes steady progress towards a higher and higher aggregate level of utility (that is, succeeds in its collective pursuit of happiness). Individuals will participate in this decentralized, complex, emergent form of law-making by taking many different roles as participants in many different online spaces with different rules. Our "collective conversation" about what sets of rules to adopt will continue in this new context, but it may now take the form of a complex interaction among discrete online spaces rather than that of a rational debate among a small group of elected representatives authorized to impose uniform rules from on high.

This new form of governance will also play havoc with our notions of the virtues of equality. Rather than judging Internet governance with reference to the well-being of a conceptualized standard "citizen" whose vote is entitled to equal weight in selecting representatives who then act on behalf of the populace as a whole, each Internet user will, directly, have differing degrees of influence in setting, encouraging, abiding by, and escaping from differing sets of rules for differing online spaces--depending on where each individual decides to embed time and attention in the online world. This direct involvement by "netizens" in establishing the rules that govern online life will not have the principal vice of "direct democracy"--because no tyrannical or selfish majority will be able to impose its will on an unwilling minority. (Those opposing the rules of a particular space will simply leave that virtual town and filter out disagreeable packets coming from that location.) Nor will rules established by such decentralized means be unenforceable. (It will be relatively easy for the cybersheriff of a given online town to banish wrongdoers.) Nor will rules necessarily be "unfairly" or arbitrarily applied. (Users can demand, by their usual means of threatening desertion, any particular degree of "due process" they may find desirable for particular online venues.)

Nevertheless, this new form of governance will require us to rethink what it means to be a good "citizen" in the online context--and what degree of "equal protection" is or is not assured by such status. The collective-action problem--the problem of finding an optimal solution to the complex and conflicting needs and desires of individuals--cannot be solved fully without paying attention to the impact on such individuals of the very actions they take in addressing these problems. Those who see themselves as citizens of a geographically defined sovereignty derive some of their own conceptions of well-being from that vision. The ability of the Internet to introduce positive and negative effects across those geographic borders can thus be seen either as a threat or as an opportunity to redefine the geographical or political group that defines what is permitted to us. Those online users who reconceptualize themselves as participating in groups that could become substantially, collectively self-governing without regard to geographic boundaries (using desertion, banishment, and software-code boundaries to limit unwanted actions and spillovers), will begin to reconceptualize themselves in terms of their multiple contingent participations in such groups. In that context, the highest form of "civic virtue" for the Internet may become not dispassionate concern for a uniformly defined public good but, rather, the grant and enjoyment of a freedom to travel--virtually--among online groups without restriction. And, from that perspective, the highest goal of geographically based governments concerned with the well-being of their own local citizens may become provision of the tools (including both access and education) necessary to facilitate such empowered freedom of movement.

Perhaps the most troubling aspect of any suggestion that we turn to a new form of civic virtue for governance of the Net is that none of the characteristics of this new mode of collective action appears necessarily to respect the liberty and unique value of individual elements. We're not sure, on first reflection, that we want to be treated as if we were light bulbs! Is the method we propose for finding the highest "fitness" level for the overall matrix flawed by the standard vices of utilitarianism: the lack of protection for individuals? What is to prevent the group, in its search for the highest peak on the policy fitness landscape, from throwing the interests of a contrarian individual off the nearest cliff?

We think the answer to this important question is that a diverse set of rule spaces, coupled with real freedom of movement, structurally respects individual liberty (and minority opinions about values) to the greatest extent possible, even as compared with democratic top-down rule. Unlike a standard utilitarian calculus, the patching algorithm does not try to find and impose a single answer, based on weighting each vote equally and then allowing the majority to rule. Instead, it allows all individuals selectively and partially to embed (and create) their own identities in any mix of online rule spaces and activities they find most empowering. The patching algorithm may enable particular online spaces to expel dissenters, but it also enables minorities to create their own more congenial online homes. If optimal congruence is preserved, the architecture of patching keeps the members of the overall online community sufficiently engaged with one another that they must continuously participate in a new form of online "conversation" with those who do not share their values. Even in those cases in which particular value systems are so at odds that total banishment or filtering removes some online spaces from any social intercourse with one another, those decisions must be made deliberately--and can now be made by all individuals as well as powerful officials speaking for sovereigns. In short, the complex world of diverse, decentralized, "federalist" rule-making for the Net will not perfectly preserve equality, nor respect every individual's views in every context, but it may well best serve liberty and personal empowerment interests just as it serves the goal of finding rules that best satisfy the needs of groups to seek their collective vision of the highest public good.

What we think we have shown--and that cyberspace is showing us every day--is that geographically based sovereignty, even as exercised by representative democracies, will not best solve collective-action problems in cyberspace. We think it's also clear there are serious reasons to worry about any global "harmonization" that loses the clearly beneficial effects of diversity, experimentation, and local self-determination--the ability of a "patching" algorithm to force the overall system downhill, off a local optimum, towards higher ground. Instead, we think that those concerned with the governance of cyberspace may find it better to rely on decentralized decision-making by the groups most directly affected by their own online activities, leaving it up to the unpredictable pull and tug between contending rule sets to find the combination of policies that is optimal in aggregate.
 

Conclusion

Americans have been fond of saying that representative democracy is the worst form of government except for all the other alternatives. In the special context of a need to govern the online world, however, there may indeed be a better alternative. Participants in online spaces cannot readily make use of the basic ideas of geographically defined representative democracy. There is no way to meaningfully provide equal votes to online "citizens," who are not for this purpose whole people entirely located in particular territorially defined "places." Law-making for diverse online spaces should not be centralized in any single global government, even a democratic one, much less one created by the slow, incomplete, and undemocratic treaty process. Legitimacy cannot be preserved by allowing existing geographical sovereigns to assert an inconsistent and asymmetrical control over all online activities that might have any impact on their citizens. But the task of setting rules for online spaces can be cut into workable pieces. And the collective-action and liberty problems can be solved simultaneously for online spaces by relying on the decisions of participants to join (or leave) those areas with rules they find empowering (or oppressive).

The new science of complex systems gives reason to hope that an overall system of governance of the Internet that reconnects rule-making for online spaces with those most affected by those rules--and that also allows online groups to make decentralized decisions that have some impact on others, and that therefore elicit disparate responsive strategies--will create a new form of civic virtue. The old hope that rational debate among wise elected representatives will result in the overall public good may be replaced, online at least, by a new certainty that dispersed and complex interactions among groups of individuals taking unilateral actions to serve "local" goals will be best for everyone, overall, over time.

ENDNOTES

The authors thank Jeff Williams for his superb assistance in developing and implementing the computer-based complex systems discussed herein. Thanks also to Larry Lessig, Dawn Nunziato, Mark Lemley, Warren Schwartz, David Skeel, Peter Swire, Steve Salop, and Avery Katz for both their assistance in formulating many of the ideas expressed herein and their comments on earlier drafts. Helpful comments were also provided by Jack Goldsmith, Peter Menell, and other participants at the Olin Law and Economics Symposium on International Economic Regulation at Georgetown University Law Center, April 5, 1997, and participants at the Aspen Institute's "Internet as Platform" conference at Wye River, June 19­20, 1997. We also gratefully acknowledge our obvious debt to the work of Stuart Kauffman. We welcome comments; you may e-mail us at david.johnson@counsel.com and postd@erols.com.

1.. David R. Johnson and David G. Post, Law and Borders: The Rise of Law in Cyberspace, 48 Stanford L Rev, 1367 (1996).

2. See Jeffrey Abramson's contribution to this volume for a discussion of citizenship and community on the Internet.

3. Consider, for example, a set of rules regarding copyright in creative expression. If sovereign X adopts a strong property rule, that may inure to the benefit of authors from other jurisdictions (at least to the extent that their work is subject to protection in X). Conversely, a weak rule may produce negative externalities, diminishing the welfare of individuals in those other jurisdictions.

4. We recognize that sovereigns do also on occasion govern actions that occur outside their territories insofar as such actions are held to have an impact on the sovereigns' citizens. That special case does not, however, detract from the general rule that sovereigns control those they can readily reach by means of physical force.

5. See, for example, James Madison's discussion of optimal "extent of territory" in The Federalist Papers, number 10.

6. Note that we are not asserting that activities in cyberspace do not have offline effects; our claim, rather, is that whatever those effects may be, they are most often distributed more-or-less randomly with respect to geographical location.

7. See Dan Burk, Federalism in Cyberspace, 28 Conn L Rev, 1095 (1996). Because the Internet Protocol calls for "geographically extended sharing of scattered resources," Burk writes, users may therefore be completely unaware where the resource being accessed is, in fact, physically located. So insensitive is the network to geography, that it is frequently impossible to determine the physical location of a resource or user [because] such information is unimportant to the network's function [and] the network's design thus makes little provision for geographic discernment. (1098­99)

See also Joel Reidenberg, "Governing Networks and Rule-Making in Cyberspace," in Borders in Cyberspace, ed. Brian Kahin and Charles Nesson (Cambridge: MIT Press, 1996), 84, 85­87, which describes the destruction of territorial and substantive borders in cyberspace.

The system of machine addressing on the Internet is discussed in Dan Burk, Trademarks Along the Infobahn: A First Look at the Emerging Law of Cybermarks, Richmond J L & Tech (1995); Robert Shaw, Internet Domain Names: Whose Domain is This? available online only, at http://www.itu.int/intreg/dns.html; Jon Postel, Domain Name System Structure and Delegation, RFC 1591 (March 1994).

8. It is, we note, not entirely clear how such a network redesign could achieve a congruence between machine addressing and physical location sufficient to identify a communicating machine's physical location from its address alone. There are, of course, optional country-specific location designators in use under the current Internet domain-naming conventions (e.g., "www.abc.fr," "www.xyz.jp," etc.), where top-level domains (*.fr and *.jp, in this example) have a geographic referent (France and Japan, respectively). But currently, there are no countries that require, as a matter of domestic law, that machines within their geographic boundaries use such a country designator; no registries that can prevent individuals holding such a country designator from assigning that designator to machines in other countries; and no means to limit the use of technical protocols such as Telnet that allow users to access the Internet from other Internet access points (with other addresses). Nor, perhaps most importantly, is the use of non-country-specific designators (e.g., *.com, *.org, etc.) restricted in any geographical way.

And we note further that the link between physical location and machine addressing appears to be substantially diminishing in recent years; the use of non-localizable global domains (e.g., *.org, *.com) is increasingly supplanting Internet domain names that include country designators such as *.fr and *.jp. See, e.g., R. Bush, B. Carpenter, and J. Postel, "Delegation of International Top Level Domains (iTLDs)" (January 1996); and Jon Postel, "New Registries and the Delegation of International Top Level Domains."

A recent proposal by the Internet Ad Hoc Coalition, a group including the Internet Society and the Internet Engineering Task Force, has proposed modifying the domain name system to, inter alia, increase the number of global non-localizable domain names. See "Final Report of the International Ad Hoc Committee: Recommendations for Administration and Management of gTLDs."

9. Johnson and Post, Law and Borders, 1375 and 1375, n. 23. For a wealth of information about the architecture of the Usenet network, see "What is Usenet?" available online at http://www.smartpages.com/bngfaqs/news/announce/newusers/top.html, and http://www.cam.org/~intsci/newsmail.html.

10. Suppose, for example, that the two jurisdictions have importantly differing rules about what constitutes permissible (not actionably infringing) "fair use."

11. Our findings are presented in more detail in our paper "Borders, Spillovers, and Complexity: Rule-Making Processes in Cyberspace (and Elsewhere)," presented at the Olin Law and Economics Symposium on International Economic Regulation at Georgetown University Law Center, Washington, D.C., April 1997. Copies of the paper are available from the authors.

12. The best introductions to Stuart Kauffman's work can be found in his The Origins of Order: Self Organization and Selection in Evolution (New York: Oxford University Press, 1993) and At Home in the Universe (New York: Oxford University Press, 1995).

13. We recognize that arguments based on the aggregate results of the patching algorithm will, in increasing overall utility, raise concerns about distributional equity and liberty. As more fully discussed near the end of this chapter, we think those concerns are best answered by recognition that, in the context of diverse online rule sets, individuals are free to embed their time and attention in multiple "patches"--and can thus choose the mix of online rules that they find most empowering. While both education and differences in access and real-world economics will constrain any effort to achieve "equality" among individuals, there is no reason, in the context of such governance mechanisms for the online world, to think that we should not be striving to increase the "utility" of the aggregate and of the average individual participant.

14. Or, shifting the level of granularity, each element might be an individual country, the state of whose copyright laws may affect "utility"--the condition of authors, publishers, and consumers of information--in numerous other jurisdictions.

15. See discussion at the end of this chapter regarding the distributional and liberty effects of the alternative governance system we propose.

16. Because of the obvious significance of intellectual property laws in this new medium, the single patch/global harmonization strategy for cyberspace has perhaps been most forcefully pressed in this context. See National Information Infrastructure Task Force, "Intellectual Property and the National Information Infrastructure" (Washington, D.C.: Government Printing Office, 1994), 132, for a statement of Clinton administration positions in favor of "efforts to work toward a new level of international harmonization" of international copyright law. See also Raymond Nimmer, Licensing on the Global Information Infrastructure: Disharmony in Cyberspace, 16 J Intl L Bus, 224, 246­47 (1995). As Nimmer notes about the Global Information Infrastructure (GII):

The multinational character of GII-related property and contract law creates potentially huge problems for the development of commercial relationships relating to information and other intangible property moved through and around the GII. More so here than in any prior commercial/economic context, an enhanced degree of harmonization and simplification is needed to enable the transactions made possible by the technology to occur. (246­47; emphasis added)

See Charles Goldring, "Netting the Cybershark: Consumer Protection, Cyberspace, the Nation-State, and Democracy," in Kahin and Nesson, eds., Borders in Cyberspace, 322, 340­44, which argues that because cyberspace activities know no national boundaries, harmonization of intellectual property laws is necessary. See also Charles McManis, Taking Trips on the Information Superhighway: International Intellectual Property Protection and Emerging Computer Technology, 41 Vill L Rev, 207 (1995), which discusses efforts to harmonize international trademark and unfair competition laws. The recently concluded meetings called by the World Intellectual Property Organization to consider proposals to modify the Berne Convention for Literary and Artistic Property is a prime example of an attempt to implement this strategy.

In regard to other legal regimes, see Joseph Grundfest, Internationalization of the World's Securities Markets: Economic Causes and Regulatory Consequences, 4 J Fin Svcs Res, 349 (1990), which discusses the drive to increase global harmonization of securities laws in response to the increasing frequency of cross-border transactions); and Louis Solomon and Louise Corso, The Impact of Technology on the Trading of Securities: The Emerging Global Market and the Implications for Regulation, 24 John Marshall L Rev, 299, 330 (1991), which notes the difficult choice of legal problems posed by cyberspace securities markets, and suggests that "adoption of a uniform international law by all nations would surely be an ideal solution to the problems of internationalization." See also Michael Rustad and Lori Eisenschmidt, The Commercial Law of Internet Security, 10 High Tech L J, 213, 300 (1995), which notes the demand among commentators for a "uniform commercial law of the Internet."

17. U.S. export administration act regulations, which declare encryption technology to constitute a "munitions" technology subject to licensing by the Department of Commerce prior to export, are discussed in National Research Council, "Cryptography's Role in Securing the Information Society" (June 1996). Also see, generally, A. Michael Froomkin, The Metaphor is the Key: Cryptography, the Clipper Chip, and the Constitution, 143 U Pa L Rev, 709 (1995).

18. See Johnson and Post, Law and Borders, which describes German attempts to suppress Usenet activity. See also Karen Kaplan, "Germany Forces Online Service to Censor Internet," Los Angeles Times, 29 December 1995, A1; and Ruth Walker, "Why Free-Wheeling Internet Hit Teutonic Wall over Porn," Christian Science Monitor, 4 January 1996, 1. Kara Swisher's "Cyberporn Debate Goes International; Germany Pulls the Shade On CompuServe, Internet," Washington Post, 1 January 1996, F13, describes efforts by a local Bavarian police force that had the effect of requiring CompuServe to temporarily cut off the availability of news groups to its entire audience (at least until a way to prevent delivery of specified groups to the German audience could be developed). Although CompuServe was initially compliant, it subsequently rescinded the ban on most of the files, sending parents a new program to choose for themselves what items to restrict. Also see "CompuServe Ends Access Suspension: It Reopens All But Five Adult-Oriented Newsgroups. Parents Can Now Block Offensive Material," Los Angeles Times, 14 February 1996, D1.

19. See Johnson and Post, Law and Borders. See also A. Michael Froomkin, "The Internet as a Source of Regulatory Arbitrage," in Kahin and Nesson, eds., Borders in Cyberspace, 140­42, which describes the difficulties inherent in any attempt by the government of "Ruritania" to prevent communication between its citizens and those in other countries.

20. The Minnesota Attorney General's Office distributed a "Warning to All Internet Users and Providers" which stated that "Persons outside of Minnesota who transmit information via the Internet knowing that information will be disseminated in Minnesota are subject to jurisdiction in Minnesota courts for violations of state criminal and civil laws" (emphasis omitted). The conclusion rested on the Minnesota general criminal jurisdiction statute, which provides that "a person may be convicted and sentenced under the law of this State if the person . . . (3) Being without the state, intentionally causes a result within the state prohibited by the criminal laws of this State." Minn Stat Ann § 609.025 (West 1987). Minnesota also began civil proceedings against Wagernet, a Nevada gambling business that posted an Internet advertisement for online gambling services. See Complaint, Minnesota v Granite Gate Resorts, Inc. (1995; No. 9507227). A Minnesota trial court judge recently upheld personal jurisdiction over the defendant in Minnesota. See State of Minnesota v Granite Gate Resorts, Inc. Also see, generally, Mark Eckenwiler, "States Get Entangled in the Web," Legal Times, 22 January 1996, S35.

21. See Dan Burk, "The Market for Digital Piracy," in Kahin and Nesson, eds., Borders in Cyberspace, 220­221, which describes the "race to externalize" given the spillover effects of the Internet on the market for intellectual property law products; see also Burk, Federalism in Cyberspace, 28 U Conn L Rev, 1095 (1996). As Roberta Romano's The Genius of American Corporate Law (Washington, D.C.: American Enterprise Press, 1993), notes:

A state will not want to pay for benefits experienced by nonresidents, for example [and may wish to] export the cost of providing goods and services for [its] residents to nonresidents, for instance by adopting taxes that are more likely to be paid by out-of-state than in-state individuals or firms. (5)

See also Clark Gillette, In Partial Praise of Dillon's Rule, or, Can Public Choice Theory Justify Local Government Law? 67 Chi-Kent L Rev, 959 (1991), which argues that outside the idealized world of no inter-jurisdictional externalities, some constraints on local power are justified "to prevent strategic local behavior"; Frank Easterbrook and Daniel Fischel, Mandatory Disclosure and the Protection of Investors, 70 Va L Rev, 669 (1984), which observes that the multistate nature of securities markets creates strategic opportunities for states to attempt to exploit investors who live elsewhere. 

22. Lawrence Lessig has written insightfully about the ways in which software code effectively enforces the virtual boundaries between online spaces. See Lessig, Constitution and Code, 27 Cumb L Rev, 1 (1997).

23. A recent decision by some collaborating sysops to send "cancel" messages for all UUNet traffic directed towards Usenet newsgroups, and by many sysops receiving these messages to implement them, represents just this sort of collective-action filtering. The "Usenet Death Penalty" is controversial, because it is aimed at an entire group of users, only a few of whom are originating "wrongful" messages. Its proponents argue that they have no alternative other than to act against the Internet Service Provider (UUNet) that "harbors" this form of wrongdoing. The technical and social pull and tug among spammers and their enemies can be expected to continue.

24. See Madison, The Federalist Papers, number 10.

25. Not only are individuals more likely to be in possession of the relevant information regarding the effect of cross-border and inter-jurisdictional spillover on their own lives, they can also act more quickly on that information than can agents at a higher level of the organizational hierarchy (i.e., their elected representatives) to whom that information must be redirected and through whom it must be reprocessed before "official" boundary realignment can occur. In other words, putting boundary definition in the hands of individuals allows a faster and more flexible response to rapidly changing spillover patterns. See Friedrich Hayek, "Notes on the Evolution of Systems of Rules of Conduct," in his Studies in Philosophy, Politics, and Economics (Chicago: University of Chicago Press, 1967), 66­81, which suggests that a single-stage, polycentric order without a "directing center" need not be inferior to a "hierarchic order" because it dispenses with the necessity of communicating to a common center all the information on which its several elements act.