Duke Law Journal: Stuart Minor Benjamin, Evaluating E-Rulemaking: Public Participation and Political Institutions, 55 Duke L. J. 893 (2006)

Cited: 55 Duke L. J. 893
[*pg 893]

THIRTY-FIFTH ANNUAL ADMINISTRATIVE LAW ISSUE

Evaluating E-Rulemaking: Public participation and political institutions

STUART MINOR BENJAMIN^†

ABSTRACT

INTRODUCTION

I. WHAT SORTS OF CHANGES AM I TALKING ABOUT?

II. WHAT DIFFERENCE WILL THESE CHANGES MAKE?
   A. Impact on Ordinary Individuals' Participation
   B. Impact on Agencies' Rulemaking
   C. Impact on Congress and the Courts

III. EVALUATING THESE POTENTIAL CHANGES TO THE RULEMAKING PROCESS
   A. If Additional Participation Does Not Result in Changes to the Agency's Decision
   B. If Additional Input Does Affect the Agency's Decision
   C. What Do the Results So Far Reveal?
   D. The Case for Modest, Skeptical Experimentation

CONCLUSION

FOOTNOTES

ABSTRACT

Proponents of electronic rulemaking proposals designed to enhance ordinary citizens' involvement in the rulemaking process have debated with skeptics the question of whether such initiatives will actually increase citizens' involvement. In the debate thus far, however, proponents have largely assumed the desirability of such involvement, and skeptics have usually not challenged that assumption. In addition, proponents and skeptics have focused on the relationship between agencies and individuals, failing to consider the larger administrative law context -- and in particular the role played by Congress and the courts. This Article considers e-rulemaking in a broader institutional context and directly addresses the desirability of the proposed e-rulemaking initiatives. The Article finds that there are good reasons to believe that e-rulemaking initiatives' costs outweigh their benefits, but that it would be premature to settle on that [*pg 894] conclusion. The Article ultimately advocates modest experimentation with e-rulemaking, both to allow for further evaluation of e-rulemaking and to provide additional data about the rulemaking process more generally.

INTRODUCTION

In the last ten to fifteen years, vast numbers of people have become familiar with basic techniques for online collaboration -- e.g., e-mail and files stored on central servers available to multiple users. And in the last few years people have become increasingly familiar with tools that allow for even richer collaboration, such as wikis, communities built on reputational capital, and even multi-player video games. In light of these developments, academic commentators and public officials have discussed new, more interactive modes of collective production of intellectual goods. Thus there has been extensive discussion of open-source or peer-production models for a broad range of products, including software, information directories, data-evaluation, and biomedical research. In the administrative law context, the proposals have focused more specifically on opening up the rulemaking process to richer input from individuals, including changes to the administrative process, not only to encourage such input but also to make it more meaningful.¹ In this Article, I assess the [*pg 895] significance and desirability of these proposed changes to the administrative rulemaking process. In so doing, I connect these questions with larger ones about the role of Congress and the courts, the desirability of public participation in the rulemaking process, and the role of experimentation.

The range of ways that information technologies could change the rulemaking process is endless.² An agency could receive automatic feedback -- for example, by setting up a system of electronic sensors to monitor a pollutant and send information back to the agency. Information technology could help flag foreseeable problems like conflicts with other rules or statutes, or even conflicts within rules, via sophisticated software that would search through all available rules and statutes to identify other provisions with wording that would seem to cover the subject of a given section of a regulation. Within the four corners of the rulemaking process itself, information technologies could make it easier for agencies to give notice to potentially interested people and allow those people to access and search the proposed rule, could allow for more interactive [*pg 896] and collaborative communications from individuals during the rulemaking process, and could make it easier for people and entities to monitor the implementation and enhance the enforcement of rules.

My focus in this Article is on proposals to integrate electronic communications into the rulemaking process as a means of increasing communications to agencies from any and all interested citizens.³ As I noted above, there are many other possible ways that information technologies could influence rulemaking, but the changes that seem to generate the most excitement are those that would allow any citizens who cared about a particular issue to give their input to agencies.⁴ The reason for this excitement, and thus for my choice of focus, is that the rulemaking process is perceived as one that is dominated by interest groups and is largely opaque to the public at large.⁵ Providing for feedback from electronic sensors to agencies, or for easier communications from agencies to individuals, may be a good way to improve agencies' efficiency; but providing for more effective communication from interested citizens to agencies could transform rulemaking more profoundly, or so its proponents suggest. If people can communicate easily with the agency and collaborate easily with each other in formulating their communications to the agency, and if the agency then modifies its decisions in light of those communications, then the rulemaking process itself will be reshaped. Anyone interested in a particular rulemaking initiative could get involved, with a realistic belief that her input could make a difference; and the reasonableness of that belief could lead many others to get involved as well, producing an upward spiral of individual [*pg 897] involvement that would change rulemaking into a truly participatory process.⁶

That, at any rate, is the hope of the proponents of e-rulemaking (the term I will use in this Article as convenient shorthand).⁷ The purpose of this Article is to discuss the likelihood and the desirability of that transformation. Other commentators have considered these issues (often briefly), but their descriptive efforts have usually focused on two institutional players -- agencies and individuals -- and have not focused on the role of Congress and the courts.⁸ The latter in particular are major players; significant rulemakings are frequently challenged in federal court, and judicial review thus affects rulemaking to a significant degree. In this Article, I attempt to take fuller account of all the possible effects of the implementation of e-rulemaking initiatives on the rulemaking process. I find that there is a strong basis to doubt that e-rulemaking will have much of a positive impact on rulemaking. I then turn to the question of evaluation with the broad range of costs and benefits in mind. The question of evaluation raises one of the most basic questions in administrative [*pg 898] law -- do we want greater public participation in the rulemaking process? In my view, that age-old question takes on a slightly different character when significant collaboration among individuals is relatively easy, as is arguably the case with some electronic communications tools. Even in light of these new tools, however, there are still good reasons to be skeptical about the impact and desirability of e-rulemaking. The problem is that the experience with e-rulemaking has been too modest, and thus the level of experimentation has been too small, to reach a firm conclusion that the costs of e-rulemaking outweigh the benefits. The combination of justified skepticism and the lack of empirical confirmation leads me to the ultimate conclusion that the government should self-consciously, rigorously, and narrowly experiment with e-rulemaking, treating agencies as laboratories of democracy. Such experimentation would allow researchers both to evaluate e-rulemaking proposals specifically and to gain information about the rulemaking process more generally. States may or may not serve as good laboratories, but in this instance I think that agencies have the potential to be excellent ones.

Part I briefly summarizes proposals to enhance public participation in the rulemaking process. Part II then turns to the question of impact, asking how those proposals might affect the various players in the rulemaking process, including individuals, agencies, Congress, and the courts. Finally, Part III considers those impacts in evaluating the desirability of the e-rulemaking proposals. My ultimate conclusion is that the uncertainties about the impact and desirability of e-rulemaking are sufficiently great that experimenting with e-rulemaking should proceed on a trial basis, in an attempt to gain greater empirical grounding before the government plunges into any particular set of changes to the rulemaking process.

I. WHAT SORTS OF CHANGES AM I TALKING ABOUT?

The integration of information technologies in general, and the Internet in particular, into the rulemaking process could take myriad forms. Even among technologies designed to enhance participation by any and all interested citizens, there are many possible variations. I will briefly summarize the main possibilities.

Starting with the most straightforward and modest changes, integrating the Internet into rulemaking can make it easier for ordinary citizens to comment on agency rulemaking proposals. [*pg 899] Agencies began taking steps in this direction in the 1990s.⁹ Agencies have allowed people to comment via electronic submissions (including via e-mail), making it easier for them to present their views to the agency. Moreover, some agencies have digitized their processes, so that all comments are available online.¹⁰

More recently, the federal government has moved to centralize and unify e-rulemaking procedures through the Federal Docket Management System (FDMS) and the Regulations.gov website. That website (which has been running since 2003) has links for documents open for comment and for lists of regulations by topic and date, as well as search tools for regulations.¹¹ More ambitiously, the aim of the docket management system is to store, and allow for retrieval of, all agency documents related to rulemaking.¹²

The centralizing tendency reflected in the federal docket management system and Regulations.gov has potential benefits in the form of economies of scale and making it easier for citizens to track and comment on any pending regulations, but it also reduces the chances that agencies will experiment with e-rulemaking initiatives on their own. As I discuss in Part III.D, I believe that is a significant cost.

In any event, these measures put comments online, but they do not create possibilities beyond those already available on paper. One can easily imagine broader changes, though. Indeed, a number of commentators have proposed them.¹³ Individuals could post a [*pg 900] response to a given communication, or set of communications, producing threaded comments resembling a dialogue that would allow other individuals (and ultimately the agency) to see a string of communications on a given topic. Either in addition to or instead of such procedures, individuals could review electronic communications to the agency and electronically join or endorse one or more with which they agreed.¹⁴ Similarly, individuals could rate comments on a scale from one to ten, which would allow for greater differentiation in levels of endorsement (or non-endorsement) of existing comments. Such a system would permit individuals and the agency to see which contributions others deemed valuable.

Indeed, the rating system could be set up so that the raters themselves were rated.¹⁵ Many websites use this model, allowing users to rate communications and allowing them also to rate other raters.¹⁶ Perhaps the most successful use of such a system is Slashdot,¹⁷ a website that consists of technology-related news stories and comments on those stories.¹⁸ Slashdot begins with a filter -- anyone can submit a story, but editors filter out the ones that they do not want to include. Once stories survive that filter, user-based ratings come to the fore: users submit comments on stories (often hundreds of comments), other users rate those comments, and some users -- designated as "moderators," or meta-raters -- rate other people's ratings. The result is that a visitor to the Slashdot website can see at a glance what stories have been highly rated by people whose views are considered valuable by people who themselves have been judged to be good raters.

[*pg 901]

Slashdot users do not produce a final product or decision -- Slashdot just enables their comments and allows for ratings of them.¹⁹ But individuals could work together toward a jointly produced submission to an agency. Agencies (or private entities) could allow individuals to post their concerns and solicit interest from others who might want to draft a shared comment. This would be akin to a bulletin board where a person could express a concern and seek co-drafters of a comment. Entities could also foster collaboration by setting up the online collaborations themselves. Some existing computer programs allow for individual involvement, although most of those programs entail a relatively small group of individuals acting as "policy juries."²⁰ But there is a well-known existing model that allows any and all interested individuals to participate in a shared project -- wikis. A wiki is an online collaboration to which anyone can contribute. The basic characteristic of wikis is that anyone can add or edit existing content.²¹ The best known, and probably the most successful wiki is Wikipedia.org, an open encyclopedia available on the Web. The idea behind wikis is the content equivalent of the programmer's notion that "Given enough eyeballs, all bugs are shallow."²² In the words of Wikipedia: "[O]ne of the great advantages of the Wiki system is that incomplete or poorly written first drafts of articles can evolve into polished, presentable masterpieces through the process of collaborative editing. This gives our approach an advantage over other ways of producing similar end-products."²³

[*pg 902]

II. WHAT DIFFERENCE WILL THESE CHANGES MAKE?

So there are lots of proposals designed to use information technologies to increase citizen participation in rulemaking. What, exactly, would implementation of these proposals change? The most obvious set of changes, and the one on which commentators have focused, centers on the actions of individuals in the rulemaking process and agencies' responses to them. As I shall discuss, that ignores other important actors, such as Congress and the courts. But I begin with the behavior of individuals, and the effect of that behavior on agencies.

A. Impact on Ordinary Individuals' Participation

The proposed changes could have an enormous effect on individuals' participation in agency rulemaking. The goal of the proposals is to make the rulemaking process more interactive, and it is entirely possible that ordinary individuals will respond enthusiastically to a more interactive process. Ordinary individuals might get actively involved in the rulemaking process -- proposing language, responding to agency proposals, and responding to proposals from other commenters. Moreover, if the early adopters are excited about their involvement, they might encourage others to get involved, with the result that involvement in rulemaking would become a truly mass undertaking.

Of course, it is possible that public participation will not increase significantly, even if all the proposals for changing rulemaking are implemented. The proposals lower the costs of participation, but it may be that relatively few people are sufficiently interested for this lowering of the barriers to their involvement to make a difference. I discuss the existing data in Part III.C; so far, the Internet has not significantly increased individual participation in the rulemaking process. Although this is disappointing for e-rulemaking advocates, the data do not rule out the possibility that the e-rulemaking initiatives will produce some increase in public participation. Indeed, it seems likely that, for some individuals, the decrease in the costs of their participation, combined with the promise of greater effectiveness, will make the difference between participating and not participating.

New rulemakings are initiated every day, making this a multiple-round interaction. As a result (and importantly), if the e-rulemaking initiatives result in a political groundswell among individuals with [*pg 903] respect to a given rule, and the government's ultimate action seems to be positively affected by that groundswell, then one might expect both that the individuals will feel good about their participation and that the success of those individuals will persuade yet more individuals to participate.²⁴ Put a bit differently, if the individuals gain pleasure from their participation in a successful movement (as seems quite plausible), then the prospect of that pleasure becomes an additional benefit to potential participants that might make the difference for those on the margin. There would be a virtuous circle of participation.²⁵

For better or worse, the converse is true: if e-rulemaking produces a groundswell that seems to make no difference in the ultimate action the government takes, that would probably tend to dampen enthusiasm, both among the new participants and among the potential participants at the margin, who now would not have an additional reason to join. Indeed, the frustration entailed in spending time on communications that have no impact on the agency might create a reason not to join. And the more this happened, the greater one would expect the dampening to be. The circle would be vicious, instead of virtuous, as fewer and fewer individuals saw much point in participating.

This discussion highlights the importance of the other players in this process -- agencies, Congress, and the courts. I now turn to that discussion.

B. Impact on Agencies' Rulemaking

The e-rulemaking proposals could affect agencies in a number of ways. First, creating more ambitious mechanisms for meaningful interaction will consume agency resources. All but the most basic proposals will require that each agency reconfigure its website and alter its existing rulemaking procedures to some degree. This probably will not constitute a very great cost to the agency, but (unlike the costs I discuss below) it will exist whether or not participation increases. So if it turns out that implementation of these e-rulemaking proposals does not increase individuals' participation, [*pg 904] the agency will have incurred costs without any compensating benefits.

Assuming, though, that the proposals do increase public participation to at least some degree, there will be other impacts on the agency. Those impacts will depend on the nature of the additional communications that individuals send.

1. Increases in Quantity Only.
It may be that implementing these proposals simply increases the quantity of messages without adding any new ideas. One way this could occur is if the entirety of the increase takes the form of individuals endorsing pre-existing messages (e.g., adding their names to a list of endorsers, or transmitting verbatim a form letter that was sent to them). If so, the cost to the agency would likely be small. As long as a computer program could pick up the fact that the new messages were identical to others that were received, the program could log those additional messages without any human having to read each of the duplicate messages. So, with any luck, the additional expenditure of resources would be modest.

The impact on the agency would be greater, however, if messages were substantively identical but were formulated differently enough that the agency's computer program did not recognize them as identical. If, for example, individuals did not send a form e-mail but instead composed their own messages making the same point (e.g., saying that raising the permissible level of soot would make it harder for people to breathe versus saying that an increase would adversely affect individuals' respiration), then humans would have to read each message in order to determine if it presented any new ideas. In fact, "[i]nterest groups have been known to encourage their members to take steps to make it hard for an agency to treat a mass of comments as if they were just X number of form letters."²⁶ This need not occupy the time of any senior agency officials -- indeed, agencies often farm out such reading to other entities. But if even one percent of the U.S. adult population commented on a given rulemaking, producing, say, a few hundred thousand comments that were sufficiently different to [*pg 905] require a human reader, the costs of plowing through those comments would be considerable.²⁷

The government might be able to reduce those costs to some degree insofar as programmers could increase the sophistication of the computer programs that read comments (although such programs would entail their own expenses) and could have confidence in the programs' ability to make ever more discerning determinations as to which comments really were duplicative.²⁸ But it is hard to imagine that any program could accurately find the vast majority of substantive overlaps (much less all of them), given that the words used might be so different. Once people started using their own idiosyncratic word formulations, it would be extraordinarily unlikely that any program could find all the overlaps. So the costs of having human readers will remain, even if they are somewhat reduced.

What impact would an increase in quantity have on a given rulemaking? One way of looking at the question is to ask, what information does the agency receive that it would not already have? By hypothesis, it is hearing no new arguments or ideas. All it is learning is that X number of people felt strongly enough to participate. That has some significance if (a) the agency is unsure of public sentiment on the issue, (b) it concludes that the number of messages is a good indicator of public sentiment, and © it decides that public sentiment is relevant to its regulatory outcome. I doubt that all three of those conditions will obtain very often. If an agency wants to count noses, public opinion polls with randomized samples will give a more accurate count.²⁹ That said, strength of preference has [*pg 906] some significance, and the number of messages will indicate something about strength of preferences -- those who take the time to communicate are likely to have stronger feelings on the issue than the average person does. But there is a distinct possibility (maybe a probability) of a skewed sample of those who really care about the issue.³⁰ For example, the average online commenters will tend to have higher than average wealth (as Web users more generally have higher than average wealth). In other ways, too, e-rulemaking participants may be a skewed sample. Maybe the rulemaking was publicized on websites frequented by people who are on one side of the issue and not publicized on websites frequented by people of the opposite view. Or maybe people with the opposite view tend not to frequent websites, or more simply tend not to use computers and thereby avail themselves of the e-rulemaking process. To pick an obvious example, a proposed regulation of online indecency is likely to attract comments from a population that is disproportionately opposed to such regulation, simply because it seems likely that opposition to such regulation would be positively correlated with use of the Web.³¹ And of course there is the potential for manipulation by interest groups³² -- which, as I discuss in Part III, has already occurred with respect to e-mail comments on proposed rulemakings.³³

Finally, even if the agency concludes that the number of messages is a good indicator of public sentiment, it may also conclude that such sentiment is irrelevant to its mission. Many statutes leave no [*pg 907] room for an agency to consider public sentiment. When Congress tells an agency to regulate based on scientific findings,³⁴ or based on its conclusions about the existence of harms to competition,³⁵ or based on what a given set of companies would need in order to compete effectively in the marketplace,³⁶ it does not leave a role for public opinion. In such situations, if an agency did consider public sentiment a reviewing court would likely find that it had acted unlawfully.³⁷

I am aware of no statute that actually directs an agency to consider the public's views. Some statutes have language that might seem to permit the agency to consider the public's views. Most notably, a number of agencies created in the New Deal era are directed to regulate certain areas consistent with "the public interest, convenience, or necessity."³⁸ But this language was used in statutes when the prevailing vision for agency heads was that they would be politically insulated experts who would act in what they concluded was the public interest and not in response to public sentiment.³⁹ The [*pg 908] history of the implementation of that language has been that agencies have reached their own conclusions, and have not invoked public sentiment in support of their positions. I cannot exclude, of course, the possibility that the agencies were persuaded by public sentiment but declined to reveal this fact, but the agencies have not so indicated. One striking example was the Federal Communications Commission's 2003 rulemaking on media ownership rules.⁴⁰ The Commission received more than one million comments, 99.9 percent of which were opposed to its proposed rules.⁴¹ Nonetheless it promulgated those rules in largely the form it had proposed, by the exact 3-2 vote everyone expected.⁴² The overwhelming sentiment against the rules in the comments appears to have had no effect.⁴³

[*pg 909]

2. Increases in Quality as Well as Quantity.
The larger goal of e-rulemaking advocates is not merely more quantity, but also more quality. Ideally, as I noted above, individuals will put forward points that otherwise would not have been presented to the agency, or at least they will sharpen points that might otherwise have been lost in the mass of comments.⁴⁴

If that happens, the agency will obviously have to devote considerable human resources to sifting through the messages. By hypothesis, individuals are putting forward new points, or variations on points that others have made. To determine exactly what information these new participants are putting forth, the agency will need to have someone (or, more likely, a phalanx of people) reading the messages with some care. And as the sheer number of messages rises, the burdens of reading them rise. In this way, increasing the number of messages is at cross-purposes with the goal of sharpening points that would otherwise get lost in the shuffle: the larger the number of messages the agency must review, the greater the danger a powerful argument will be overlooked. The only way to avoid that problem is to devote substantial resources to the careful reading of each message -- quite an undertaking if there are hundreds of thousands of nonduplicative messages.

Significantly, these costs will arise whenever there are messages that a computer program cannot match as being identical -- even if the nonduplicative messages in fact make identical points and thus add quantity without adding any new information. The agency cannot know a priori whether the messages that the program fails to match are merely using different words to make the same point, or instead are using different words to make a slightly different point. So the agency must carefully read all the messages that the program cannot match as duplicative, in order to ensure that it does not miss the new or different point. Accordingly, the costs outlined in the previous section are actually higher than they might first appear.⁴⁵

[*pg 910]

Note that increased costs per regulation mean that unless Congress increases the spending for agencies to cover these increased costs, agencies will be able to issue fewer regulations.⁴⁶ They may pursue fewer regulations from the outset (if for each regulation an agency adds additional resources sufficient to cover the costs), or they may pursue the same number of regulations but then spend a longer time or more money promulgating each one, but the net result at the end of the day (or, more to the point, the end of an administration's four-year term) will be fewer regulations.

What difference will this new input make? Assuming it is relevant to the agency's statutory mandate, it might make a big difference.⁴⁷ If the additional messages put forward points that the agency would not have considered in the absence of those messages then they could help to shape the ultimate regulation. Two basic conditions are necessary for this to occur: (a) the points must be ones that the agency would not have received from other sources, or thought of on its own; and (b) they must actually persuade the agency -- or at least persuade a court that the agency should have taken them more seriously (I take up this latter possibility in the next section).

[*pg 911]

As for the first point, there is reason to doubt. First, for any major agency action, there are dozens of interested parties who are well-paid to find every point that can help their side -- and virtually every side is represented. That said, most existing commenters represent interest groups, and it may be that even groups styling themselves as representing ordinary people or consumers' interests in fact fail to do so.⁴⁸ In particular, the existing commenters may tend to underproduce both public goods themselves (e.g., data that are in the public's interest) and arguments supporting the creation of public goods. But interest groups that oppose a regulation know that they, like all commenters, should couch their comments to the agency in terms of the public interest, as both the statutory mandate and the political realities call for such arguments. So the question is whether an interest group that strongly opposes a proposed regulation for self-regarding reasons will put forward all the public-regarding data and arguments that a new individual commenter might want to bring forward. To put the proposition differently, it may be that the additional participants enabled by e-rulemaking would find that their "new" public-regarding points had in fact already been made to the agency by other commenters, even though those other commenters might have acted in their own private interests.

Second, the agency itself may have thought of the points that the additional individual participants would make. After all, the agency is generally supposed to act in the public interest, and presumably the reason the agency proposed the regulation in the first place is that it recognized the public interest considerations in favor of its plan.⁴⁹ Third, if neither existing commenters nor the agency would have [*pg 912] come up with a particular point on its own, the question remains whether the additional increment of participants enabled by e-rulemaking would do so. One's answer about the likelihood of new input ultimately depends, then, on one's sense of the degree to which interest groups and agencies really do put forward all the relevant points, and, if they do not, the realistic chances that the new participants enabled by e-rulemaking will, in fact, add relevant points. If the existing participants fail to represent public-regarding views but new participants would represent them, then the first condition would be satisfied.

Note that the points in the preceding paragraph about participants and agencies focus on public-regarding arguments and data. That does not exhaust the additional information that the new participants under e-rulemaking might offer. They may submit additional input that is not public regarding, but instead advances their own private interests or their own idiosyncratic views. If the additional increment of messages is so composed, this raises questions about the normative value of such input and the chances that such input will affect the agency in desirable ways.

Turning to the second question noted above (whether the new information persuades the agency), the answer depends on how agencies actually operate. How often are agencies convinced by arguments and data put forward in formal messages (as opposed to informal communications that occur outside the agencies' routinized processes, or direction from political actors)? Note that many commentators have argued that the influential communications to agencies occur outside the rulemaking process delineated by the Administrative Procedure Act. In Don Elliott's words, "Notice-and-comment rulemaking is to public participation as Japanese Kabuki theater is to human passions -- a highly stylized process for displaying in a formal way the essence of something which in real life takes place in other venues."⁵⁰ But also note that part of the goal of e-rulemaking is to change that dynamic -- to make individual participation more useful and relevant to agencies. The question is whether the additional increment of messages that e-rulemaking generates will in fact create that change.

[*pg 913]

This inquiry about the degree to which agencies are persuaded by comments is part of a larger inquiry that is central to the question of the effectiveness of the new participation enabled by e-rulemaking: are agencies so influenced by powerful political and/or industry forces that the public-regarding arguments made by the new participants will not persuade the agency? If agencies are captured by powerful interest groups, or controlled by political forces, then the additional messages will just be flotsam for the agency to disregard.

Putting the first question (will individuals add new information?) with the second (will agencies be persuaded?) highlights the prediction of many theories of agency behavior that the likelihood of a significant impact from the additional messages is fairly low. If the agency is captured, then the additional messages will not make a difference. And if the agency is acting in a public-regarding manner and is reasonably competent, then one should expect it to come up with the good public-regarding arguments and data on its own. One can tell a story in which the new messages might make a difference -- maybe the agency is so understaffed that no one creating the rule has the time to think of the public-regarding considerations (but, if so, on what basis did they propose the regulation in the first place?), or maybe (less implausibly) the agency staffers are not smart enough to come up with the good arguments and data on their own but are smart enough to recognize them when they are made by others. Simply put, many theories of agency behavior would suggest that the additional participation enabled by e-rulemaking will not, in the end, change agency behavior.

C. Impact on Congress and the Courts

As I noted above, commentators have largely focused on the potential impact of e-rulemaking initiatives on the citizenry and on agencies' behavior in response. This focus is understandable -- after all, the agencies promulgate the rules. But the matter often does not end there. In some situations, Congress countermands the agency's action. Admittedly, such congressional intervention is rare. Much more common, of course, is a challenge to the legality of a rule by an aggrieved party, which entails judicial review of the agency action.

1. Congress.
Depending on what actually motivates members of Congress, it is possible that an enhanced role for the Internet in rulemaking will lead to more frequent congressional rejections of agency actions. If (a) agencies make rulemaking more open to [*pg 914] citizens' electronic participation, with the result that in at least some cases citizens mobilize more than they would have in the absence of such electronic participation; and (b) if that greater mobilization leads those citizens to lobby Congress about that rule more effectively than they would have in the absence of their electronic participation; and © if such increased mobilization can be the difference between congressional action and congressional inaction on the rule, then e-rulemaking may indeed produce congressional actions that reject or modify agency rulemakings after they are issued.

The idea behind e-rulemaking is that citizens will become more involved in the process. Such a result is not obvious, but it is certainly possible.⁵¹ Thus the first element of this causal chain seems reasonable. The second element is reasonable, too, in that it requires only that citizens be a bit more effective in lobbying Congress. The biggest problem is in the third element. It is one thing for e-rulemaking to lead to somewhat more effective lobbying by Congress, and quite another for that increase to be the difference between congressional action and inaction. This is particularly so in light of the rarity of post-issuance congressional actions that countermand or modify agency rulemakings.⁵² So even if e-rulemaking does have some effect on the margin of congressional behavior after rules are issued, such an effect might not lead to any more congressional rejections or modifications of agency rulemakings. That is, the paucity of successful [*pg 915] cases may also indicate that there are precious few marginal cases, such that e-rulemaking will in fact play out little differently from ordinary rulemaking in Congress.

Congress directs agencies ex ante with much greater frequency. Might e-rulemaking have an effect on congressional action that shapes the action the agency will ultimately take? The short answer is that there is little reason to suppose so. The idea behind e-rulemaking is that citizens will become more engaged in the agency's rulemaking on a given issue. One could imagine that some of these citizens will become interested in the agency's workings more generally and thus will pressure their members of Congress to direct the agency's future actions in a particular way, but that is a much more tenuous connection. The big change promised by e-rulemaking proponents is that it dramatically lowers the costs of participation in rulemaking for citizens. But it doesn't dramatically lower the costs of citizens educating themselves about the business of the agency. Online availability of agency regulations, and the possibilities for e-mail communication among concerned citizens, may lower the costs of citizens educating themselves, but those technologies are already in place.

2. Courts.
The branch that regularly responds to agency rulemakings is of course the judiciary, not the legislative branch. Agency rulemakings frequently do not end at the agency, but instead are challenged by an aggrieved party and reviewed by federal courts.⁵³ [*pg 916] If e-rulemaking develops as its proponents suggest, it could have a significant impact on this judicial review.

In the previous section I noted that increases in the quality of messages must either persuade the agency or persuade a court that the agency should have taken them more seriously. In that section I discussed persuading the agency, and now I turn to persuading a court.

A central element in judicial review of agency actions is the question of whether the agency adequately considered relevant materials. This arises to some degree in courts' review of agencies' factual findings. Courts will reject agency fact-finding under the relevant standard ("arbitrary [or] capricious"⁵⁴ in the context of informal rulemakings and adjudications, "substantial evidence"⁵⁵ for formal rulemakings and adjudications) if the agency ignores evidence that undercuts the agency's position. But rejections of rulemakings based on infirm factual findings are relatively rare. The more common basis for rejecting a rulemaking is that an agency's policy decision was "arbitrary [or] capricious" -- a standard that is implemented via "hard look" review.⁵⁶ Under such review, a court will invalidate an agency action if the court determines that the agency failed to take a hard look at the significant considerations against its position.⁵⁷ If an agency fails to offer an adequate explanation for its [*pg 917] rejection of such countervailing considerations, or promulgates a regulation that fails to take into account relevant factors, a court will invalidate the action.⁵⁸

The changes to rulemaking discussed in the previous sections could have a significant impact upon hard look review, and thus upon the likelihood of a court's rejecting an agency's action. The significance of adding more people to the process has two ramifications for judicial review of agencies. First, it produces more sheer data. More communications to the agency mean more material for the agency to digest.⁵⁹ Second, those additional data mean more ideas that a court might find to be alternatives at which the agency should have taken a hard look.

As I have noted, it is the hope and expectation of those who are excited about the possibilities for e-rulemaking that people will not simply check a pre-existing box, or e-mail the same form letter that every other individual is sending, but instead will contribute new ideas. If that happens, agencies will have new points to consider.⁶⁰

What this means for aggrieved parties is that, if they discover a comment that they believe is significant but that the agency did not carefully consider, they will likely bring this to the court's attention. More substantive comments thus present more opportunities for litigants to find something the agency overlooked. They also provide more opportunities to find an area of disagreement (as to the significance of an argument against the rule) between the agency and the reviewing court.

The point about overlooking a comment is straightforward: if the agency receives a hundred thousand comments, it may simply miss a good argument presented in one of them. As to potential [*pg 918] disagreements, courts and agencies do not always concur as to whether a given argument is significant enough or a given alternative is sufficiently viable to require the agency to take a hard look. An increase in new substantive communications translates into an increase in the likelihood that the agency and the court will disagree about the necessity of the agency's considering one point that a participant puts forward. Just one such failure can be fatal to a regulation. When an agency declines to take a hard look at a participant's proposal but the court decides that the agency should have done so, the court will find the regulation unlawful.⁶¹

Interest groups are well aware of this, of course. In fact, interest groups are often confident that their input will not affect the agency's action, but submit comments in an attempt to slow the agency down and, more importantly, to create a record that will lead to judicial invalidation of the agency's action. Indeed, this apparently was the case in the rulemaking that prodded more comments from individuals than any other -- the FCC's proposed rule relaxing some media ownership limits. More than a million individuals filed comments on the proposed rulemaking (99.9 percent in opposition to it).⁶² The leaders of the opposition, however, had no expectation that these comments would affect the FCC. Instead, they saw them as an opportunity to place in the record arguments and data that they could invoke in the inevitable judicial challenge to the FCC's actions, as well as an opportunity to engage in political mobilization that might influence Congress to reverse the media ownership rules after the [*pg 919] FCC had issued them.⁶³ The intended audience for their comments, in other words, was the other two branches of government (and in particular the courts).

This has enormous consequences for agencies. Either an agency will have to devote substantially more resources to its rulemaking process, or there will be a much greater likelihood of a court invalidating the regulation. As to the resources, I noted above that having lots of extra messages that a computer program cannot identify as duplicative will consume agency resources, because someone will have to read them. It might seem tempting to say that the extra costs to the agency will not be that great, because it can give those extra messages pretty short shrift (which, of course, undercuts the value of having them in the first place). But the significance of hard look review is that giving them short shrift increases the chances that a court will invalidate the regulation. In an era when significant rulemakings take about three years from start to finish, the costs of invalidation are pretty high. So the looming possibility of judicial review means that agencies really will need to read all those extra messages, to avoid overlooking important arguments and data.

But reading the comments is not sufficient, and seriously considering all the input that a reviewing court will later find to be significant is even more costly. It is one thing to hire more comment-readers, and another entirely to require the agency officials who are crafting the regulation to show that they have taken a hard look at the raft of arguments and data. The usual demonstration of the agency's hard look appears in the agency's explanation of its regulation -- the agency explains why X option is not attractive, or why Y argument is not persuasive. In the absence of such an explicit discussion in its explanation of its rule, the agency would need to be able to prove to the court that it had, indeed, given serious consideration to these other options -- a demonstration that would probably occupy even more resources than would a discussion in its justification for its rule. The bottom line is that if lots of new ideas come forward, the agency will find that its officials in charge of the rulemaking will have to [*pg 920] devote substantially more energy to responding to the ideas it has rejected.⁶⁴

But maybe that is a good thing. The previous paragraphs focus on the costs to the agency, but maybe those costs are benefits to the general public. After all, perhaps the new input will persuade the agency (as the previous section discussed), or should have persuaded it (in which case judicial invalidation of the rule would be welcome). More generally, maybe policymakers should embrace the changes that e-rulemaking would bring about, despite the costs.

III. EVALUATING THESE POTENTIAL CHANGES TO THE RULEMAKING PROCESS

The modest literature on e-rulemaking has devoted most of its energy to the two issues discussed above -- how might interactive use of the Internet be integrated into the rulemaking process, and what changes would adoption of these proposals likely produce -- while devoting considerably less energy to the more fundamental question of whether the benefits of those changes outweigh the costs.⁶⁵ How should one evaluate the desirability of implementing these proposals? I now turn to this latter, crucial question.

A. If Additional Participation Does Not Result in Changes to the Agency's Decision

If the additional input enabled by the e-rulemaking initiatives would not change the substantive decisions that agencies make, should policymakers nonetheless embrace those initiatives? At first blush, the answer would be no: As I noted in Part II.B, combing through additional messages imposes some costs, and by hypothesis the messages would have no effect on the agency's actions.

[*pg 921]

This calculus does not take into account the potential benefits to participants from their participation. Some commentators have suggested that face-to-face deliberation has benefits for participants that might also apply to e-rulemaking -- making participants feel that their views have been heard, helping them to understand others' positions and perhaps changing their own views (or changing the views of others), and, maybe most important, making them into more engaged citizens.⁶⁶ On this last point, theorists have for years decried Americans' relative lack of involvement in policy debates. E-rulemaking could change this as to specific rulemakings: communications to an agency on a particular issue are one form of citizen engagement. But e-rulemaking also could work on a larger level. One goal of e-rulemaking proponents (and advocates of greater deliberation more generally) is to energize citizens into becoming involved in debates beyond those at issue in a given rulemaking -- to become active, engaged citizens more generally. This would be valuable not only to those particular citizens but also to society as a whole: this greater involvement would be a public good.

Recall, though, that I am positing that the public's participation has no impact on an agency's decision. It is, of course, possible that the individuals who participate will gain pleasure from their participation, even if it makes no difference to the outcome. But it seems at least as likely that (as I noted in Part II.B), once they realize their comments are not having any impact, they will become frustrated. More specifically, whatever benefits they receive from feeling that their views have been heard will be mitigated by their awareness that the hearing of their views did not translate into any change in the agency's position. By contrast, the possibilities of individuals understanding others' views, and changing their own minds or others', exist independently of the impact of citizens' communications on an agency. As long as individuals have a meaningful back-and-forth among themselves, such understanding and persuasion is possible. But this means that the likelihood of this understanding and persuasion depends on the degree to which the e-rulemaking initiatives allow for meaningful interchanges among [*pg 922] participants and the degree to which people use them. I will discuss the former in the next section, so I set that aside here. As to the degree to which people would use such interactive e-rulemaking tools, that would likely depend in significant part on whether participants became frustrated once they realized that agencies were not responding to their input. This point also applies to the last, and biggest, benefit to individuals and society from individuals' participation (i.e., producing engaged citizens). If participants are demoralized rather than energized, then the optimistic vision of an engaged citizenry will not materialize. My own guess is that there might be a short-term increase in engagement (and benefits to society, in the form of energized citizens and understanding and persuasion of others), but that this increase would dissipate after the ineffectuality of the individuals' attempts to persuade the agency became apparent.

It is interesting to contrast e-rulemaking with negotiated rulemaking. In negotiated rulemaking, agencies begin a rulemaking by establishing a committee comprising representatives from regulated firms, trade associations, citizen groups, and other affected organizations, as well as members of the agency staff. The idea is that, with the stakeholders meeting together around a table, they might be able to reach a consensus, resulting in a rulemaking process that would be much speedier and much less likely to produce litigation (because the stakeholders would have already agreed to the rule).

Negotiated rulemaking turned out not to produce those results, and more generally not to be as successful as its advocates had hoped. Rules generated through negotiated rulemaking take just as long to promulgate as rules generated without it, and are no less likely to be litigated.⁶⁷ There was a small study (comparing the reactions of [*pg 923] participants in six conventional rulemakings with participants in eight negotiated rulemakings) that found that participants in the negotiated rulemakings had higher levels of satisfaction with the process, found the process more instructive, and believed that they had generated better quality rules.⁶⁸ The significance of this study was mitigated by its small scale and some significant methodological questions about it, but it does suggest some benefits to negotiated rulemaking.⁶⁹ In any event, many commentators have concluded that the benefits of negotiated rulemaking do not outweigh the costs (given that negotiated rulemaking does not speed up rulemakings or reduce the likelihood of litigation).⁷⁰

Here, by hypothesis, e-rulemaking does not generate better rules -- or even change the rules.⁷¹ It thus seems quite unlikely that e-[*pg 924] rulemaking would generate even the modest benefits associated with negotiated rulemaking (absent an effect on the agency's rules). This is not to say that participation for the sake of participation has no value. Rather, its value is likely to be low over time -- especially once participants realize that their participation is not making a difference.⁷²

And, as I have discussed, the costs for the agency are likely to be considerable. The bottom line is that, unless an agency's position changes as a result of e-rulemaking, it seems hard to justify.

B. If Additional Input Does Affect the Agency's Decision

Let's now consider the possibility that the additional input does affect the agency's action. This is the scenario that advocates of e-rulemaking hope for -- individuals participate in the rulemaking process more significantly than they have in the past, and this additional involvement influences the substance of agency regulations. The advocates often assume that such public influence would be beneficial, but that is far from clear. More important, it implicates one of the central debates in administrative law: should agencies be influenced by, or insulated from, input from the public?

Many key figures in administrative law have argued that agencies should not respond to outside forces but should be guided instead solely by their expertise and independent judgment. They contend that the questions before the agencies should, and do, turn on technical expertise, not political considerations. In their view, influence by individuals is to be avoided, not encouraged.⁷³ Others take a quite different view, arguing that agencies should respond to concerns from the public, and should not try to reach their own [*pg 925] expert judgments in a vacuum.⁷⁴ Thus the possibility that agency heads will receive valuable guidance from public-spirited citizens is matched against the possibility that those agency heads will receive biased or ill-conceived guidance from players who are motivated by a self-interest that is not in the public interest (e.g., special interest groups) or who are woefully ill-informed (e.g., citizens who are responding to scare tactics). Adding hundreds of thousands of new citizen participants to the rulemaking process would be particularly troubling to those who believe that citizens' input would be misguided: they would fear that agencies would indeed respond to these new participants. Such participation may correlate to political mobilization, so an agency might bend to the participants' wishes even if it did not believe that they represented the median American, as long as the agency believed that they represented a politically potent force.

This description of the dichotomy is of course vastly oversimplified. There are many permutations in the setup of this debate -- public choice theorists would suggest that everyone acts, to at least some degree, in her private interest, so that there is no truly "public-spirited" input, civic republican theorists would have a more sanguine view of the input of individuals, etc. Accordingly, the debate has taken many different forms over the years.⁷⁵ Does it play out differently in the e-rulemaking context? Perhaps.

As I noted in Part I, some of the proposals for e-rulemaking would allow for collaboration and/or for individuals rating others' contributions. Working models of such enterprises include, most notably, Wikipedia and Slashdot.

The idea behind Wikipedia is fairly straightforward: two heads are better than one, and one million heads are far better than one. The editors of Wikipedia do not claim that each addition will be an improvement, but they do claim that the movement will be toward a better product. New people will help to refine what is written, so continued adding/editing will improve it.

[*pg 926]

There is some dispute about the reliability of Wikipedia entries in comparison to conventional encyclopedias. Wiki skeptics suggest that adding people will lead not to continual improvement, but instead to regression to the mean. Robert McHenry, a former editor in chief of the Encyclopedia Britannica (and thus not a neutral observer) wrote a criticism of Wikipedia making exactly this argument. He focused on an entry on a subject about which he knew quite a bit (Alexander Hamilton) and found not only that it was riddled with errors and bad prose but also that the entry became worse as more people contributed to it.⁷⁶ In his words, "the earlier versions of the article are better written overall, with fewer murky passages and sophomoric summaries. Contrary to the faith, the article has, in fact, been edited into mediocrity."⁷⁷ The journal Nature, however, undertook a more systematic study entailing peer review of forty-two scientific entries in Wikipedia and the Encyclopedia Britannica.⁷⁸ It found that Wikipedia's entries had more errors, but "the difference in accuracy was not particularly great: the average science entry in Wikipedia contained around four inaccuracies; Britannica, about three."⁷⁹

There are, however, two reasons why the relatively positive experience underlying the Nature study would not extend to collaborative commenting on agency rulemaking. First, Wikipedia has a fairly sophisticated set of rules to deal with people whose contributions do more harm than good. Wikipedia's founders recognized the danger posed by "cranks," "trolls and flamers," "partisans," et cetera,⁸⁰ and they created several mechanisms for [*pg 927] policing the site to eliminate "patent nonsense."⁸¹ Most notably, other users can edit such nonsense or, if they deem the entire entry to be nonsense, can list it on a page devoted to nominations for deleting entries.⁸² If the user is more confident of an entry's worthlessness, she can make it a "candidate for speedy deletion."⁸³ Beyond that, there are a few hundred "administrators" of Wikipedia, each of whom has the authority to delete any page.⁸⁴ There is even a dispute resolution process.⁸⁵ And in 2005 (in response to a notorious episode in which an anonymously created Wikipedia entry falsely suggested that John Seigenthaler Sr. may have been involved in the Kennedy assassinations), the founder of Wikipedia announced that Wikipedia would stop accepting new pages from anonymous creators.⁸⁶

Creating a mechanism for crafting communications to agencies on the Wikipedia model would thus entail a substantial investment of time and energy for whichever private entity chose to do it -- and it would have to be a private entity, as it is hard to imagine a government entity playing the editing role described above.⁸⁷ It is [*pg 928] certainly possible that some person or group of people would go to the trouble to set up a wiki for communications to an agency, but it is not at all clear that individuals would come in significant numbers to such a wiki, as their contributions would be included in the communication to the agency only if the wiki's controllers so agreed.

Isn't the same thing true for Wikipedia? Yes, but this highlights the second difference between Wikipedia and communications to an agency: such comments will heavily involve policy preferences, and thus the chances for disagreement are very high. One aspect of the Wikipedia model is that its entries are compilations of facts. And insofar as there are disputes about facts (e.g., whether surface temperatures on earth have risen over the last one hundred years), then the Wikipedia entry can compile the points made by either side. Communications to agencies involve facts, too, but policy considerations are almost always central. If, say, the EPA were to reconsider whether it should create a cap-and-trade system for emissions of carbon dioxide, various facts (including the rise in surface temperatures over the last one hundred years) would be relevant, but so, too, would fundamental policy choices.⁸⁸ If two participants have opposite views on the propriety of the government's regulating emissions of any pollutant, it is hard to see how a wiki would mediate their dispute. One will be heard and the other excluded (by an editor accepting one and not the other, or, if there is no editor, by whoever modifies the page last putting in her views); or both will be heard, in which case the resulting communication will be schizophrenic.

This has, indeed, occurred on Wikipedia itself with respect to certain entries that reflected larger political positions. One notorious example occurred in the run-up to the November 2004 presidential election. People started vandalizing the pages of George Bush and John Kerry, and as soon as one person fixed a page it was vandalized again. In light of the problems, an administrator took the [*pg 929] extraordinary (for Wikipedia) step of "freezing" the Bush and Kerry pages, despite the strongly worded Wikipedia norm against such freezing.⁸⁹ But then another administrator, enforcing the norm against freezing, unfroze the pages, only to have them vandalized again. The pages were then frozen and unfrozen, back and forth, several more times.⁹⁰

I have been discussing a model that contains editors. What about collaboration with no editors or controllers? The answer would likely not be satisfactory. There is every reason to expect a continual back-and-forth of deletions and reinsertions between those with competing views -- like the battle over the Bush and Kerry pages, without the periods of freezing. Indeed, it seems reasonable to expect that some partisan (and there need be only one) would program her computer (as for Ebay auctions) to wait until the last moment before the communication was finalized, and then to replace the entire communication with one to her liking in the millisecond before such finalization. And, the inserted material might not even be responsive. That is what happened to the Los Angeles Times website when it experimented with a collaborative editorial: users repeatedly added pornography to the editorial, such that the editors finally decided to end the experiment prematurely.⁹¹

Would it make sense to go further, with a regulation itself (as opposed to comments about a regulation) created by peer production, or a regulation as a wiki? No. The beauty of processes such as open source, peer production, and wikis is that they are ultimately producing a product that individuals can take or leave as they see fit. In contrast to the commercial context of peer production, a rule is not merely one of many products among which users can choose; it is imposed on all regulated parties. In the commercial context, a misbegotten project simply fails to gain adherents and is quickly [*pg 930] forgotten. In the regulatory context, a misbegotten project imposes huge costs, precisely because it actually governs. Second, and relatedly, in the peer production context the goal is "rough consensus" -- substantial agreement among the relevant players, but not universal agreement. And usually no single person or small group of people must be in the majority; if the overwhelming majority agrees on something but there are a few holdouts, the norm is that the agreement prevails, regardless of who the holdouts are. But that system presumes an absence of hierarchy, and in agencies there is a clear hierarchy: outsiders can come to whatever agreement they want, but the ultimate judgment is up to the agency head. The peer production model embodies a structure of rough consensus that presumes a particular arrangement -- in which the agency head would go along with whatever "rough consensus" emerged -- that would represent a significant shift in agency operations. Such a shift is possible, of course, but the argument for it needs to be made. Note that this was one of the objections to negotiated rulemaking; having people sitting around a table negotiating reduced the agency head to just another interested party.⁹²

If Wikipedia is not an attractive model for citizens' communications to agencies, one obvious change would be to keep every person's contribution intact, rather than let someone else edit or delete it. There could still be an element of collaboration if, say, people could rate, and respond to, others' contributions. The model here is Slashdot and its Slash software. As I noted in Part I, anyone can submit a story, but editors filter out those they deem unhelpful. That filtering aspect is similar to the Wikipedia model and has its attendant advantages and disadvantages: it allows for greater focus, but at a cost of deleting some people's views. Putting a small group of editors in charge of filtering everyone's input would put a huge amount of pressure on the political ideology of those editors, with the attendant likelihood that they would filter out messages with which they disagreed. In addition, eliminating some voices is at cross-purposes with the ideal of allowing open communication to agencies. [*pg 931] The stakes, and propriety, of deleting comments are different for a website devoted to providing relevant news than for one devoted to presenting the views of concerned citizens.

But recall that on Slashdot, once stories make it through the opening filter, users submit comments, rate other people's comments, and (when acting as "moderators," or meta-raters) rate other people's ratings. Might such a reputation-based system, without the additional element of exclusion of inputs that editors' found unhelpful, be a good model for communications to agencies? It is possible. The Slashdot model seems more attractive than the wiki model. It would allow for everyone to add input, but each added comment would have a rating that an agency could use (as users of Slashdot regularly do) as a signal of value. So the agency could see, at a glance, which comments were deemed by other users to be of high quality, and which were not. This would avoid the disadvantages of deleting comments while still allowing for quality judgments by other users. If these quality judgments could be trusted, they could help agencies and courts determine which comments contained valuable input.

But could they be trusted? There would still be the danger of quality judgments being centrally based on policy preferences. Asking people to make quality judgments about a news item is quite different from asking them to make such judgments about the desirability of environmental regulation. Indeed, as with Wikipedia, Slashdot may work because the stakes are relatively low: all that is at issue is the rating of an article. When it comes to commenting on important public disputes, users may find that their political preferences trump other considerations.

There may be ways to mitigate the problems created by strongly held political views trumping other aspects of judgment. First, it may be that people recognize strong points and identify them as such even when they disagree. Second, and perhaps more realistically, a large number of negative ratings need not overwhelm the information conveyed by a large number of strongly positive ratings. Among the many points and arguments that might be put forward in favor of a given regulatory position, advocates of that position would have an interest in supporting the ones they deemed strongest. So if, among fifty different communications in favor of a particular policy position, one had a large number of very high ratings and a large number of very low ratings, whereas all the other communications had few high ratings (and whatever number of low ratings), the agency might conclude that the one with the most high ratings was deemed to be [*pg 932] the best point by the advocates of that view. Whether this metric would work would depend on the degree to which opponents of that position would intentionally confuse the system by giving a high rating to an argument that they deemed weak (a risky strategy, given that the agency might treat that high rating as support for the policy position more generally), and the degree to which such bad-faith commenters could be identified (e.g., if the agency could see that they were simultaneously giving high ratings to diametrically opposed policy positions).

For better or worse, such poll skewing -- in which people intentionally support voting positions (usually, candidates) whom they dislike in order to produce a weak opponent -- is frequently alleged (it is of course hard to prove). Indeed, this was a stated fear of political parties that opposed open voting regimes under which those not registered in the party could vote in the party's primaries; they feared that people hostile to the party's agenda would vote for a weak candidate to hurt the party and help their own candidate of another party.⁹³ One problem for those attempting poll skewing in the voting booth is that they fail to use their one vote to support their own candidate. In the virtual context, by contrast, an individual could try to have it both ways, by supporting both the arguments she liked and the ones of her opponents that she deemed weakest. Agencies could try to prevent such behavior by checking a given person's ratings for consistency, but that entails significant costs and can be circumscribed by the person's adopting more than one online identity. Agencies could try to avoid this last problem by allowing each person to have only one online identity, but that entails very great costs. It would be a massive undertaking to ensure that each online identity corresponds (a) to a real person who (b) actually signed up herself -- as opposed to having her relative or friend sign up in her name, knowing that she would never bother to do so.

In short, the familiar debates about the desirability of outside forces affecting agency decisions may have a somewhat different cast in the e-rulemaking context. The difficulty is that one cannot know whether it will in fact do so. More to the point, one cannot determine, a priori, whether the impact will be desirable.

[*pg 933]

C. What Do the Results So Far Reveal?

Commentators are not entirely in the world of speculation -- there are some data on the implementation of e-rulemaking thus far. What do they tell us? Professor Coglianese's contribution to this symposium covers much of what I address in this Section,⁹⁴ so my discussion will be relatively brief.

In the pre-e-rulemaking era, pending rulemakings attracted relatively few comments from individuals; the main participants were major stakeholders.⁹⁵ Has the advent of the Web and e-mail commenting brought about changes? Discouragingly, the answer is no. As to quantity, the empirical data suggest that the advent of the Web and e-mail commenting has not produced an increase in the number of comments.⁹⁶ More robust data suggest that the introduction of e-mail comments has not increased comments for the overwhelming majority of rulemakings. That is, if the Internet has had any effect on the quantity of public communications with agencies, that effect is confined to a very small number of unusually salient rulemakings -- ones for which interested groups mount huge public relations efforts (combined with form e-mails for people to send). If there is any increase in comments, in other words, it would appear to be attributable to the ease of sending form comments in response to publicity campaigns.⁹⁷

This is not entirely surprising. In other contexts, reducing costs has not led to an explosion of citizen participation. For instance, barriers to voting have been reduced in the last half century, and yet the percentage of the voting-age population that actually votes has declined during that time. And even experts in the rulemaking [*pg 934] process (for whom the costs of participation are presumably lower) often do not file comments.⁹⁸

If the Internet has had little or no impact on the quantity of public participation, what about the quality of such participation? Again, the results are discouraging. The studies thus far find that the individual comments do not put forward new data or arguments. Individuals have overwhelmingly sent form letters, or form letters with an additional sentence or two that adds no new rationales, data, or arguments that the agency would not have already received.⁹⁹ Are the participants who submit electronic comments engaging in greater deliberative activity? Again, the answer so far is no. An empirical study found that "electronic commenters do not appear to be any more deliberatively engaged than paper commenters."¹⁰⁰ The evidence so far indicates that the additional individual involvement enabled by e-rulemaking has neither produced different regulatory outcomes nor produced the degree of engagement among the citizenry that some have expected.¹⁰¹

[*pg 935]

Indeed, perhaps the most obvious result of the implementation of e-rulemaking is the rise of campaigns using form e-mails aimed at agencies, and the concomitant rise in consulting groups that help to organize those campaigns.¹⁰² Interest groups have done a good job of presenting alarming (often misleading) statistics on their websites as a means of pushing people to submit e-mail comments to agencies.¹⁰³ This has often proved to be an effective organizing tool, but it does not appear to convey useful information to the agency.¹⁰⁴

D. The Case for Modest, Skeptical Experimentation

The analysis could stop here. That is, policymakers could decide that they have enough information to determine that e-rulemaking simply will not have a positive impact that will justify the costs of its implementation (and maybe not have any positive impact at all). That would be a reasonable conclusion, given the disappointing results so [*pg 936] far, but on balance it would be a mistake. Further experimentation with e-rulemaking both might uncover forms of e-rulemaking that are worth the costs and might give policymakers valuable information about the rulemaking process. I believe that neither of these potential benefits on its own would justify further development of e-rulemaking initiatives, but that the combination of the two justifies a modest level of experimentation with new e-rulemaking initiatives.

The possibility that some forms of e-rulemaking will have benefits that outweigh the costs is fairly straightforward in light of the previous discussion. Although the data on e-rulemaking are discouraging, they are also incomplete. Our experience with the current experiments is fairly brief, and broader changes (such as wikis and reputation-based systems) have not been attempted. Insofar as comments from individuals have added little to what agencies already received from existing commenters, it could be that individuals simply will not add much; but it also could be (as some e-rulemaking proponents contend) that merely allowing citizens to e-mail agencies changes fairly little, whereas creating opportunities for meaningful collaboration with or rating by individuals will present points and data that agencies would not otherwise receive. Similarly, the fact that input thus far seems not to have changed agencies' rulemakings could mean that agencies will not be affected by input from individuals; but it could be that they will be affected if more valuable input comes from individuals via collaboration or rating. Put differently, researchers have not run the experiments that would allow them to determine whether the patterns in the offline world (where individual comments have little impact) would apply if the broader e-rulemaking proposals were fully implemented for at least some rulemakings. The existing data are too limited to give much guidance on these issues.

The second potential benefit is less obvious: e-rulemaking initiatives may give policymakers valuable information about the rulemaking process. E-rulemaking is not specifically aimed at increasing transparency and creating more data for researchers. The stated motivation behind e-rulemaking is to confer the benefits discussed in this Article. But the effect of e-rulemaking may be to enhance researchers' and policymakers' understanding of the rulemaking process. Such a greater understanding does not depend on e-rulemaking producing the benefits that its proponents have put forward. It may be that full implementation of e-rulemaking initiatives does not produce more citizen involvement, better input [*pg 937] from citizens, or changes in agency behavior, but that it nonetheless sheds light on how agencies operate by giving researchers additional tools to track how materials are presented to agencies and how those agencies choose to respond (or not to respond) to them. If, for example, e-rulemaking increases the quantity and quality of citizen participation in the commenting process, but these increases have no impact on agencies' behavior, that fact will suggest that agencies' decisions are not affected by those comments and instead are influenced by other inputs. This result would be disappointing to e-rulemaking proponents, but it might be useful for those trying to understand how agencies work -- and in particular the degree to which they are captured by powerful entities. It is not clear that such information would make experimenting with additional e-rulemaking initiatives worth the cost, but adding the value of the information created by the agencies' non-responsiveness makes it a closer question. By contrast, if new e-rulemaking initiatives did not increase the quantity or quality of citizen participation, then the goals of the e-rulemaking proponents would not be advanced and researchers would learn nothing new about agencies. All they would learn is that people believe they have better things to do with their time than comment to agencies, even when such commenting is interactive -- information that would not seem to be worth the cost of creating new e-rulemaking initiatives. The outcome that would most likely produce benefits greater than the costs would arise if it appeared that the additional participation resulting from new e-rulemaking initiatives did have a positive impact on the agency. In those circumstances, e-rulemaking would thus not only change agency behavior for the better but also provide valuable evidence about agency decisionmaking.

Experimentation will not provide definitive answers. Any given agency action will always have its own idiosyncrasies, and the metrics for determining the quality of public participation and the impact of that participation on agencies will always be imperfect. After all, researchers can create clear, objective metrics (e.g., number of comments submitted to an agency by members of the public) that may not have any correlation to an agreed-upon definition of quality. Meanwhile, they can instead use qualitative measurements, but of course those are contestable. They could, for example, look to see whether public participation added valuable ideas, but that would require deciding which ideas really are valuable. But researchers could gain data that, in combination with possibly contestable [*pg 938] judgments and assumptions about causal mechanisms, should provide some useful information. If, for example, wider public participation produced a point or argument that did not appear elsewhere and was not obvious, and that point or argument ended up in the agency's regulation, one might surmise that there was a reasonable chance that the public participation had influenced the agency's decision. Policymakers could then draw conclusions about the value of that influence, depending of course on contestable judgments of value.

There are lots of ways that agencies can tweak their procedures to increase transparency and provide additional data. They do not need to implement e-rulemaking to gain that information. So if the goal is more information about how agencies operate, why not pursue that goal directly and jettison the e-rulemaking proposals (unless they happened to be the best way to produce additional information about the rulemaking process)? The answer, as I suggested above, is that more ambitious e-rulemaking implementation may have the benefits that their proponents have put forward, and failing to determine whether those benefits materialize would be a missed opportunity. By adding a new element of procedures designed to encourage public participation, policymakers can not only gain information about the rulemaking process but also more directly evaluate those new procedures.

I do not believe that this justifies across-the-board adoption of e-rulemaking proposals. The costs of such implementation are too high, and the benefits too uncertain. But I believe it does justify experimentation with such proposals by some agencies, so that agencies can serve as laboratories of democracy. The classic instance of laboratories of democracy is state experimentation,¹⁰⁵ but agency experimentation seems at least as attractive. State experimentation has the advantage of allowing for comparison insofar as states are similarly situated, but it has the disadvantage of potentially significant costs for interstate activities: a single company might have to comply with a dozen different regulatory schemes as its activities cross state lines. Each agency regulates a different sector of the economy, so that comparison among agencies is not always easy (although comparison among states is also not as easy as one might imagine, given the variability among state regimes). But the advantage of [*pg 939] experimentation for agencies is that agencies regulate at the national level, so experimentation does not increase costs for regulated parties.

As I noted in Part I, the federal government has moved away from individual agency experimentation and toward a uniform set of e-rulemaking protocols.¹⁰⁶ Such centralization may have benefits, particularly insofar as it reduces the costs for individuals to participate in agency actions.¹⁰⁷ But this discussion highlights that it has significant costs, because it minimizes the divergence among agencies and thus the possibilities for experimentation. If the choice is between across-the-board implementation of further e-rulemaking initiatives or not, this Article suggests the latter choice. But modest experimentation is the best choice.

CONCLUSION

I have tried in this Article to describe the possible outcomes, and evaluations, of the implementation of e-rulemaking. The key word is "possible." My Article, filled with words like "could" and "might," highlights the uncertainty surrounding the effects of e-rulemaking. This is not an accident. One of my main contentions is that e-rulemaking could play out in a variety of different ways, and for this reason it is difficult to be confident in my judgments.

We know something about the costs: implementation of e-rulemaking proposals will entail nontrivial costs.¹⁰⁸ If e-rulemaking inspires any significant number of citizens to participate, agencies will have to devote considerable resources to considering this new input. And if e-rulemaking does not inspire significant new input, then it will not be terribly costly, but it will also be of little benefit; by hypothesis, whatever effort agencies will expend in creating e-rulemaking procedures will inspire a collective yawn. So the costs of implementation exist, and they should make us wary.

[*pg 940]

The empirical data so far are discouraging, but we do not know enough about the benefits of e-rulemaking to rule out the further implementation of e-rulemaking proposals. Most fundamentally, we do not know whether the input enabled by e-rulemaking will be beneficial, we do not know whether the agency will respond to it, or how courts will respond to the agency's action, and thus we do not know whether, in light of the costs and the benefits, it makes sense to undertake ambitious e-rulemaking initiatives. If I ended the analysis here, I would end with a position of skepticism but uncertainty: my assessment is that e-rulemaking likely is not worth the costs, but that we cannot quantify the possibilities with confidence.

The analysis need not stop there, however. Policymakers can encourage modest experimentation with new e-rulemaking procedures, as that will allow them to evaluate such procedures and gain additional information about the rulemaking process. Our skepticism about e-rulemaking -- informed by the disappointing empirical data so far -- counsels against a massive investment in e-rulemaking initiatives. But our uncertainty over the possible benefits of e-rulemaking, combined with the possibility of new data on rulemaking, counsels in favor additional experimentation.

It bears noting that our uncertainty about the benefits of e-rulemaking stems in significant part from the fact that rulemaking has long been an insiders' game. Public participation has never been terribly easy or encouraged. A member of the public could file a formal comment (which is time-consuming) or send a form letter or e-mail (with little reason to believe that it would make a difference), but she had no easy mechanisms to collaborate with other citizens. Ignorance about the likely level and quality of public participation has thus been in part a function of the relative inaccessibility of the rulemaking process to members of the public. Meanwhile, ignorance about how agencies decide -- and in particular whether they are influenced by communications that are not generated by interest groups -- flows from a lack of good empirical data about the effect that such communications actually have on agencies. Both forms of ignorance could be alleviated by experimentation with e-rulemaking.

Thus the promise of e-rulemaking, in my view, is that it may produce valuable changes in the rulemaking process, and it may give us more information about the rulemaking process. Skeptical experimentation, then, will yield two kinds of data -- more information about the attractiveness of e-rulemaking and more information about how rulemakings take shape -- without imposing [*pg 941] the larger costs that full implementation of e-rulemaking would entail. Even if e-rulemaking initiatives prove to have little impact, they may thus serve the salutary purpose of providing better information upon which to base future initiatives to change the regulatory process.

FOOTNOTES

†	Professor of Law, Duke University School of Law. B.A., Yale University, 1987; J.D., Yale Law School, 1991. I would like to thank Cary Coglianese, Joan Magat, Arti Rai, Stuart Shulman, and Neil Siegel for helpful comments.
1.	See Steven J. Balla, Between Commenting and Negotiation: The Contours of Public Participation in Agency Rulemaking, 1 I/S 59, 59 (2005) ("Researchers and practitioners are on the cusp of a new era in rulemaking, one in which the continued application of information technology has the potential to transform both the conduct and management of rulemaking . . . ."); Thomas C. Beierle, Digital Deliberation: Engaging the Public Through Online Policy Dialogues, in DEMOCRACY ONLINE: THE PROSPECTS FOR POLITICAL RENEWAL THROUGH THE INTERNET 155, 155 (Peter M. Shane ed., 2004) ("Accompanying the last decade's enthusiasm for the Internet economy were equally extravagant expectations about information technologies' impact on democracy and governance."); Barbara H. Brandon & Robert D. Carlitz, Online Rulemaking and Other Tools for Strengthening Our Civil Infrastructure, 54 ADMIN. L. REV. 1421, 1478 (2002) ("The Internet, properly utilized, can make policymaking more transparent, and enable Americans outside the Beltway to participate more effectively in the activities of the federal government."); Cary Coglianese, E-Rulemaking: Information Technology and the Regulatory Process, 56 ADMIN. L. REV. 353, 355 (2004) [hereinafter Coglianese, E-Rulemaking] ("Electronic rulemaking, or e-rulemaking, offers the potential to overcome some of the informational challenges associated with developing regulations."); Cary Coglianese, The Internet and Citizen Participation in Rulemaking, 1 I/S 33, 34 (2005) [hereinafter Coglianese, Internet and Citizen Participation] ("[M]any lawyers and policymakers look with hope to new information technologies as a way of overcoming rulemaking's democracy deficit."); Stephen M. Johnson, The Internet Changes Everything: Revolutionizing Public Participation and Access to Government Information Through the Internet, 50 ADMIN. L. REV. 277, 279 (1998) ("The Internet and other modern technological innovations can expand public access to agency decisions . . . ."); Orly Lobel, The Renew Deal: The Fall of Regulation and the Rise of Governance in Contemporary Legal Thought, 89 MINN. L. REV 342, 439 (2004) ("Both federal and state agencies are constructing Web sites with rulemaking documents, which allow citizens to submit electronic comments on proposed rules."); Jeffrey S. Lubbers, The Future of Electronic Rulemaking: A Research Agenda, ADMIN. & REG. L. NEWS, Summer 2002, at 6 ("The second purpose of the transformation of rulemaking is a participatory one-making it possible for participants to participate in real time . . . ."); Beth Simone Noveck, The Electronic Revolution in Rulemaking, 53 EMORY L.J. 433, 433 (2004) [hereinafter Noveck, Electronic Revolution] ("Informal rulemaking . . . is about to be transformed by the silent revolution of e-government . . . ."); Beth Simone Noveck, The Future of Citizen Participation in the Electronic State, 1 I/S 1, 3 (2005) [hereinafter Noveck, Citizen Participation] ("Now the crafting of regulations will shift from a paper-based to a fully electronic lawmaking process."); Stuart W. Shulman, The Internet Still Might (But Probably Won't) Change Everything, 1 I/S 111, 111 (2005) ("The evolution of information and communication technologies has the potential to revolutionize citizen participation in the democratic process."); Steven L. Clift, E-Government and Democracy: Representation and Citizen Engagement in the Information Age 2 (Feb. 2004), http://www.publicus.net/articles/cliftegovdemocracy.doc ("Leading governments, with democratic intent, are incorporating information and communication technologies into their e-government activities.").
2.	Probably the most comprehensive lists can be found in Noveck, Citizen Participation, supra note 1, and in Cary Coglianese, The Internet and Public Participation in Rulemaking 7 (Regulatory Pol'y Program, Working Paper RPP-2003-05, 2003), available at http://www.ksg.harvard.edu/cbg/research/rpp/RPP-2003-05.pdf. See also Robert Carlitz & Rosemary Gunn, E-Rulemaking: A New Avenue for Public Engagement, 1 J. PUB. DELIBERATION Article 7, at 5 (2005), available at http://services.bepress.com/cgi/viewcontent.cgi?article=1008&context=jpd; Clift, supra note 1.
3.	For this reason, I do not focus on suggestions that agencies select a small group of citizens and get their feedback in the form of a focus group or virtual jury. See Coglianese, Internet and Citizen Participation, supra note 1, at 12 (noting the possibility of virtual juries and regulatory polling); Noveck, Citizen Participation, supra note 1 (same).
4.	See Brandon & Carlitz, supra note 1, at 1422 ("[T]he federal government should build a transparent online environment that encourages public input."); Coglianese, Internet and Citizen Participation, supra note 1, at 33 (expressing the popular view that the Internet may revolutionize citizen participation in rulemaking); Elena Larsen & Lee Raine, The Rise of the E-Citizen: How People Use Government Agencies' Web Sites (Pew Internet & American Life Project 2002), available at http://www.pewinternet.org/pdfs/PIP_Govt_Website_Rpt.pdf (noting rise in the use of the Internet to send comments on policy choices of public officials).
5.	See, e.g., CORNELIUS M. KERWIN, RULEMAKING: HOW GOVERNMENT AGENCIES WRITE LAW AND MAKE POLICY 113, 189 (3d ed. 2003); Coglianese, Internet and Citizen Participation, supra note 1, at 38-40 (summarizing data on who participates in agency rulemaking process); Marissa Martino Golden, Interest Groups in the Rule-Making Process: Who Participates? Whose Voices Get Heard?, 8 J. PUB. ADMIN. RES. & THEORY 245, 265 (1998) (noting that interest groups and businesses dominate rulemaking submissions).
6.	Furthermore, according to one school of thought, such a transformation of rulemaking would give it much greater democratic legitimacy. Many theorists have emphasized the importance of discussions among citizens in a democracy, suggesting that democratic legitimacy requires meaningful political exchange among citizens. See, e.g., BENJAMIN R. BARBER, STRONG DEMOCRACY: PARTICIPATORY POLITICS FOR A NEW AGE 136 (1984) ("[T]here can be no strong democratic legitimacy without ongoing talk."); EMILE DURKHEIM, PROFESSIONAL ETHICS AND CIVIC MORALS 89 (1957) ("The more that deliberation and reflection and a critical spirit play a considerable part in the course of public affairs, the more democratic the nation."); ROBERT C. POST, CONSTITUTIONAL DOMAINS 184-87 (1995) (discussing and adopting such a theory of democracy). In this spirit, transforming rulemaking into a participatory process would increase its democratic legitimacy.
7.	See ELAINE CIULLA KAMARCK & JOSEPH S. NYE, JR., GOVERNANCE.COM: DEMOCRACY IN THE INFORMATION AGE (2002); Beierle, supra note 1, at 155 (arguing that Internet-based rulemaking participation opens the prospect for vastly increased participation); Brandon & Carlitz, supra note 1, at 1422 (recognizing that "the Internet could fundamentally change how the public participates in federal policymaking," but noting need for further innovations in participative forums); Clift, supra note 1, at 2 (considering potential democracy-enhancing aspects of Internet rulemaking); Daniel C. Esty, Environmental Protection in the Information Age, 79 N.Y.U. L. REV. 115, 170 (2004) ("People with good ideas-even those who never get to Washington or their state capitals-thereby can have a chance to shape policy outcomes."); Johnson, supra note 1, at 279 (noting the potential for the Internet to expand rulemaking participation); Noveck, Citizen Participation, supra note 1, at 1 (noting the potential for enhanced participation but arguing that further innovation in the participative forums is required).
8.	See generally Coglianese, E-Rulemaking, supra note 1 (considering the interaction of administrative agencies and the public in e-rulemaking); Noveck, Citizen Participation, supra note 1 (same); Shulman, supra note 1 (same).
9.	See generally UNITED STATES GOVERNMENT ACCOUNTABILITY OFFICE, ELECTRONIC RULEMAKING: PROGRESS MADE IN DEVELOPING CENTRALIZED E-RULEMAKING SYSTEM (2005), available at http://www.gao.gov/new.items/d05777.pdf (chronicling impact of e-rulemaking efforts).
10.	The Federal Communications Commission, for example, makes all submissions available on its website at http://www.fcc.gov.
11.	Cary Coglianese et al., Unifying Rulemaking Information: Recommendations for the New Federal Docket Management System, 57 ADMIN. L. REV. 621, 628 (2005).
12.	See id. at 628 ("Essentially, the FDMS will make the public records of rulemaking activities, in their entirety, readily available via the Internet to anyone interested in tracking government rulemaking."); see also Cindy Skrzycki, Funds for E-Docket Filed Under 'No,' WASH. POST, Jan. 10, 2006, at D1 (noting delays in the completion of the docket management system).
13.	See generally Coglianese, Internet and Citizen Participation, supra note 1 (arguing that existing efforts to apply information technology to rulemaking are unlikely to vastly increase citizen participation, but acknowledging that newer technologies may facilitate such a result); Noveck, Citizen Participation, supra note 1 (suggesting that changes in the software used for e-rulemaking are necessary to facilitate greater participation); Shulman, supra note 1 (expressing skepticism over the ability of the Internet to change rulemaking significantly, at least in light of current technologies).
14.	The "join" versus "endorse" distinction might matter a great deal to the creator of the original communication if the potential endorser was someone with whom the creator did not wish to associate. If, for example, a scientist writes a comment arguing for a particular position on scientific grounds, she might not be enthused about a self-identified member of the Flat Earth Society "joining" that comment, if joining meant having the Flat Earth member's name listed alongside that of the creator. Mere endorsement, by contrast, presumably would not have the same connotation, because the endorsement would not entail actually joining, but instead would be a separate document of support.
15.	If desired, the raters' raters could be rated, and so on.
16.	Amazon.com, for example, uses such a system.
17.	www.slashdot.com.
18.	Slashdot bills itself as "News for Nerds." Yochai Benkler, Coase's Penguin, or, Linux and the Nature of the Firm, 112 YALE L.J. 369, 393 (2002).
19.	"Regardless of [Slashdot's] scalability, they do not structure decisions, just a series of conversations." A. Michael Froomkin, Habermas@Discourse.Net: Toward a Critical Theory of Cyberspace, 116 HARV. L. REV. 749, 871 (2003).
20.	Beth Noveck discusses some of these at length. See Noveck, Electronic Revolution, supra note 1, at 495-510 (2004) (discussing models for online citizen policy juries).
21.	WikiWikiWeb, http://c2.com/cgi/wiki?WikiWikiWeb, is credited as the original Wiki. Froomkin, supra note 19, at 861. WikiWikiWeb allows anyone to add or change content. In 2005, it instituted a delay, posting the following notice: Note to all wiki spammers: As of 1-2-2005 no changes to this wiki, either by editing or adding new pages, will be picked up by search engines until 10 hours have passed. All spam on this site is usually deleted in minutes, an hour at the most, so it is now pointless to try to add spam of any type to this wiki. Wiki Wiki Web Faq, http://www.c2.com/cgi/wiki?WikiWikiWebFaq (last visited July 10, 2006).
22.	ERIC S. RAYMOND, THE CATHEDRAL AND THE BAZAAR: MUSINGS ON LINUX AND OPEN SOURCE BY AN ACCIDENTAL REVOLUTIONARY (2002).
23.	Wikipedia Editing Policy, http://en.wikipedia.org/wiki/Editing_policy (last visited July 10, 2006).
24.	By "ultimate action," I mean to include not merely the rule the agency proffers, but also subsequent judicial review (if it occurs), and even congressional changes to the rule (which is unlikely but of course possible). I discuss Congress and the courts in the next section.
25.	The virtue, here, is in the eye of the beholder-particularly based on one's views of the desirability of individuals influencing agency behavior. See infra Part III.C.
26.	Fred Emery & Andrew Emery, A Modest Proposal: Improve E-Rulemaking by Improving Comments, 30 ADMIN. & REG. L. NEWS 8, 8 (Fall 2005), available at http://www.abanet.org/adminlaw/news/adlaw_fall2005.pdf.
27.	I suggest a few hundred thousand different comments on the theory that there would be some duplication that a sophisticated computer program would recognize. If each member of my hypothetical one percent used his or her own words, or if five percent of the adult populace commented instead of that hypothetical one percent, then humans would have to read well over a million comments.
28.	See generally Hui Yang & Jamie Callan, Near-Duplicate Detection for eRulemaking (Carnegie Mellon Univ. Language Tech. Inst., 2005), available at http://erulemaking.ucsur.pitt.edu/doc/papers/dgo05-huiyang.pdf (discussing computer programs that should be able to detect form letters with very slight differences from the model form letter).
29.	See Thomas C. Beierle, Discussing the Rules: Electronic Rulemaking and Democratic Deliberation 11 (Resources for the Future, Discussion Paper No. 03-22, 2003) (suggesting that e-rulemaking may produce "a cacophony of unreflective comments [that] tempts rule writers to lapse into preference aggregation, counting up support and disagreement in an inappropriate application of a voting model"); Coglianese, Internet and Citizen Participation, supra note 1, at 55-56 ("[S]ystematic and substantial increases in citizen comments would shift regulators' attention away from selecting the policy option that best fulfills their statutory mandate or the public interest, and instead lead regulators to strive to satisfy the views expressed by those who file comments. This problem would move the process closer to making rules by plebiscite, which will risk giving undue prominence to expedient or even erroneous considerations.").
30.	Notably, in response to the FDA's proposed regulation of cigarettes and smokeless tobacco there was "a massive tobacco company-orchestrated campaign [that] generated some 300,000 pieces of mail - nearly half of all of the mail received by the agency on this topic." Regulations Restricting the Sale and Distribution of Cigarettes and Smokeless Tobacco to Protect Children and Adolescents, 61 Fed. Reg. 44,395, 44,418 n.34 (Aug. 28, 1996) (to be codified at 21 C.F.R. pt. 801). Such a concerted effort may reveal more about the organizing ability and wealth of those on one side of a regulation than it does about citizens' strength of preference.
31.	In this regard it is unsurprising that there was more organized opposition to the Communications Decency Act of 1996 among Web users than there was organized support for it.
32.	See Peter Strauss, The ABA Ad Law Section's E Rulemaking Survey, 29 ADMIN. & REG. L. NEWS 8, 10 (Spring 2004), available at http://www.abanet.org/adminlaw/news/adlaw_spring2004.pdf (stating that "it is not hard to imagine manipulative campaigns exploiting tools of spam to proliferate comments dramatically"); Emery & Emery, supra note 26, at 8 (noting that such a "rulemaking arms race" is already occurring).
33.	See infra notes 102-104 and accompanying text.
34.	See, e.g., 15 U.S.C. § 2643(d)(7) (2000) (requiring the collection and dissemination of information about asbestos "based on the best available scientific evidence," which "shall be revised, republished, and redistributed as appropriate, to reflect new scientific findings"); 16 U.S.C.A. § 669c(d)(1)(D) (2000) (requiring the development of a "wildlife conservation strategy based upon the best available and appropriate scientific information and data").
35.	See, e.g., 47 U.S.C. § 761(a)(1) (2000) (providing that the FCC "may not issue a license or construction permit to any separated entity, or renew or permit the assignment or use of any such license or permit, or authorize the use by any entity subject to United States jurisdiction of any space segment owned, leased, or operated by any separated entity, unless the Commission determines that such issuance, renewal, assignment, or use will not harm competition in the telecommunications market of the United States").
36.	See, e.g., 47 U.S.C. § 251(d)(2) (2000) ("In determining what network elements should be made available . . . , the [FCC] shall consider, at a minimum, whether (A) access to such network elements as are proprietary in nature is necessary; and (B) the failure to provide access to such network elements would impair the ability of the telecommunications carrier seeking access to provide the services that it seeks to offer.").
37.	The irrelevance of public opinion to agencies' legislative mandate does not, of course, mean that there is no role for comments from the public. The sections of the Administrative Procedure Act governing informal rulemaking put public comments at the center of the process. See 5 U.S.C. § 553 (2000) (stating that agencies "shall give interested persons an opportunity to participate in the rulemaking through submission of written data, views, or arguments"). Congress's directions to agencies as to the basis upon which they should act determines which arguments contained in those comments are relevant to the agency's decision and which are not. Agencies will accept virtually any comments, in other words, but only some put forward arguments that are within the agency's purview.
38.	See, e.g., 47 U.S.C. § 309(a) (2000) (authorizing the FCC to regulate licenses consistent with "the public interest, convenience and necessity").
39.	See JAMES M. LANDIS, THE ADMINISTRATIVE PROCESS 28 (1938) (noting that the administrative process "promotes expertness . . . [which] makes for much more effective public responsibility"). On the centrality of Landis' vision, and role, in the creation of New Deal agencies authorized to act in the public interest, see generally DONALD M. RITCHIE, JAMES M. LANDIS: DEAN OF THE REGULATORS (1980).
40.	In re 2002 Biennial Reg. Review--Review of the Comm'n's Broad. Ownership Rules & Other Rules Adopted Pursuant to Sec. 202 of the Telecomm. Act of 1996, 18 F.C.C.R. 13,620 (2003) [hereinafter 2003 FCC Ownership Order].
41.	See Prometheus Radio Project v. FCC, 373 F.3d 372, 386 (3d Cir. 2004) ("[N]early two million people weighed in by letters, postcards, e-mails, and petitions to oppose further relaxation of the [media ownership] rules."); 2003 FCC Ownership Order, supra note 40, at 13,957 (Copps, Comm'r, dissenting). As Commissioner Copps stated in his dissent to the Commission's order (when the comments had not yet topped the one-million mark): This proceeding has generated three-quarters of a million comments now-more than any other proceeding that I am aware of in the history of the FCC. Of those comments, all but a few hundred are from individual citizens. And of those, nearly every one opposes increased media consolidation-over 99.9 percent! Id.; see also Jennifer 8. Lee, Deregulating the Media: Comments Showed Solid Opposition, N.Y. TIMES, June 3, 2003, at C8. ("More than 520,000 comments on the proceeding were sent in by citizens . . . . [and] nearly all of them were against relaxing media ownership rules.").
42.	This does not mean, however, that there was no point in so many people filing those comments. The need to wade through so many comments occupied a fair amount of time and thus may have slowed down the FCC. More importantly, the strategy of the leaders of the groups that pushed for the comments was to set up the FCC for reversal in the courts, not to change the FCC's own decision. See infra note 63 and accompanying text.
43.	Indeed, the FCC's characterization of these comments so indicates. In its media ownership order (when it had received hundreds of thousands of comments, but not yet one million), the FCC stated that We received more than 500,000 brief comments and form letters from individual citizens. These individual commenters expressed general concerns about the potential consequences of media consolidation, including concerns that such consolidation would result in a significant loss of viewpoint diversity and affect competition. We share the concerns of these commenters that our ownership rules protect our critical diversity and competition goals, as they are designed to do, and we believe that the rules adopted herein serve our public interest goals, take account of and protect the vibrant media marketplace, and comply with our statutory responsibilities and limits. 2003 FCC Ownership Order, supra note 40, at 13,624. The reference to "brief comments and form letters" and to the "general concerns" they contained, combined with the FCC's response in the next sentence at a similarly high level of generality, do not inspire the sense that the FCC found these comments terribly helpful or influential.
44.	An example of the latter would entail individuals finding a comment with which they agreed and making a similar argument more effectively.
45.	This does not necessarily mean that costs per regulation would increase. It is possible that an increase in costs at one stage (here, the commenting process) will produce a greater reduction in costs at another stage (by reducing the time it takes to issue the rule, or the likelihood of litigation after the regulation is promulgated). In such a situation, the total costs per regulation would not increase. The problem is that there is neither evidence nor much reason to support this proposition, and some evidence that cuts the other way. As for the reasoning, I am talking here about an agency devoting more resources to determining the content of messages. There is no reason to expect that spending more time reading messages will systematically shorten, or otherwise make more efficient, the regulatory process. Nor will the fact of reading additional input from individuals tend to reduce litigation against the agency, unless (a) the agency changes its regulation in accordance with the wishes of that additional input (an assumption that I discuss next), (b) the people placated by that change would have challenged the regulation, and would have been the only ones to do so, and © the change does not aggrieve a new set of potential litigants who are at least as likely to bring litigation. As for the evidence, as I discuss shortly, negotiated rulemaking was designed to let stakeholders reach a consensus about regulations before the notice and comment process began, in the expectation that this would shorten the notice and comment process and reduce the likelihood of litigation. The actual experience with negotiated rulemaking, however, has been that it does neither. See infra notes 67-70 and accompanying text.
46.	And, of course, even if Congress does increase agencies' funding to cover the costs, that money has to come from somewhere-either from cuts in other programs or from additional borrowing or taxation.
47.	If, in response to a regulation mandating that an agency reduce the emissions of a minor pollutant "to the extent feasible," Person A sends an irrelevant message (e.g., "Please do not regulate this pollutant, because I like inhaling it" or "Please reduce emissions to X level, because otherwise I'll die in the next few days"), and Person B then makes a slightly different argument (e.g., "Many people like to inhale this pollutant, so please keep the level reasonably high" or "Many people fear this pollutant, so please ban it to allay their concerns"), these messages are still irrelevant to the statutory mandate. The additional argument does not give the agency useful guidance.
48.	See William Funk, When Smoke Gets in Your Eyes: Regulatory Negotiation and the Public Interest-EPA's Woodstove Standards, 18 ENVTL. L. 55, 95 (1987) ("The [Consumer Federation of America] may have represented the interests associated with the mentality of a Consumers Reports reader, but it did not appear to lobby on behalf of poor, rural folk for whom the rule will provide little benefit and perhaps significant burden."); Mark Tushnet, Foreword: The New Constitutional Order and the Chastening of Constitutional Aspiration, 113 HARV. L. REV. 29, 79 n.240 (1999) ("[S]o-called public interest groups . . . . are staffed and funded by the relatively well-to-do, and their programs are determined by what the professional elites perceive to be the interests of the persons on whose behalf they are acting."); C. Frederick Beckner, III, Note, The FDA's War on Drugs, 82 GEO. L.J. 529, 548 (1993) ("[Many] self-styled public interest groups do not represent the interests of the public as a whole, but instead exploit consumer fear in order to aggrandize the political power necessary for their existence.").
49.	For those readers who think the previous sentence is fanciful or hopelessly naive, because agencies are captured, then the additional messages enabled by e-rulemaking may make new points, but they are likely to be disregarded by the captured agency. See infra note 50 and accompanying text.
50.	E. Donald Elliott, Comment, Re-Inventing Rulemaking, 41 DUKE L.J. 1490, 1492 (1992).
51.	As I discuss in Part III, the evidence so far suggests that e-rulemaking initiatives have not produced an increase in individual involvement in the rulemaking process. If an increase in citizen participation does not, in fact, materialize, that presents a challenge not merely to a suggestion of congressional involvement but more fundamentally to the desirability of e-rulemaking initiatives in the first place. See infra Part III.
52.	Before INS v. Chadha, 462 U.S. 919 (1983), was decided, Congress routinely included and acted upon legislative vetoes in legislation, which allowed one or both houses of Congress to reverse agency actions. Id. at 968 (White, J., dissenting) ("[O]ver the past five decades, the legislative veto has been placed in nearly 200 statutes."). After Chadha, the difficulty of passing congressional legislation rejecting or modifying agency actions led Congress to pass legislation creating an expedited process for Congress to respond to agency decisions. See Contract with America Advancement Act of 1996, Pub. L. No. 104-121, §§ 251-253, 110 Stat. 847, 868-74 (1996) (codified at 5 U.S.C. §§ 801-808 (2000)). That legislation raised the prospect of Congress's countermanding many agency actions, but in fact it has been successfully used only once since its enactment (to overturn ergonomic rules imposed on workplaces). See GARY LAWSON, FEDERAL ADMINISTRATIVE LAW 108 (3d ed. 2004) (discussing the legislation and noting the rarity of its use); Morton Rosenberg, Whatever Happened to Congressional Review of Agency Rulemaking?: A Brief Overview, Assessment, and Proposal for Reform, 51 ADMIN L. REV. 1051, 1052 (1999) (same); Julie A. Parks, Lessons in Politics: Initial Use of the Congressional Review Act, 55 ADMIN. L. REV. 187, 188-89 (2003) (discussing the one successful utilization of the congressional review process).
53.	The exact percentage of agency rules that are challenged in court is not clear. The most fulsome debate has revolved around EPA rulemakings. Many commentators have suggested that 80 percent to 85 percent of EPA rulemakings are subject to challenge, but the better estimates appear to be that the number is lower-closer to a quarter of all EPA rules and a third of major EPA rules. See Cary Coglianese, Assessing Consensus: The Promise and Performance of Negotiated Rulemaking, 46 DUKE L.J. 1255, 1298-1301 (1997) (finding that 26 percent of all EPA rules issued from 1987 to 1991 and 35 percent of major EPA rules issued from 1980 to 1991 were challenged in court); Christopher H. Schroeder & Robert L. Glicksman, Chevron, State Farm, and EPA in the Courts of Appeals During the 1990s, [2001] 31 Envtl. L. Rep. (Envtl. Law Inst.) 10,371, 10,375 (Apr. 2001) (reporting that 33 percent of EPA major rules issued during the 1990s were challenged in court). But see Richard J. Lazarus, The Tragedy of Distrust in the Implementation of Federal Environmental Law, 54 LAW & CONTEMP. PROBS. 311, 334 (Autumn 1991) ("Both environmental organizations and industry took advantage of the increased judicial access and together challenged between 80 and 85 percent of EPA's major decisions."); Coglianese, supra, at app. D (collecting assertions that 80 percent of EPA rulemakings were subject to challenge). Significantly, those who are in the agencies apparently believe the higher numbers. See Seidenfeld, Cognitive Loafing, Social Conformity, and Judicial Review of Agency Rulemaking, 87 CORNELL L. REV. 486, 514 (2002) ("[T]here is a general perception, shared by those who study judicial review of administrative rules and those in the agencies, that most rules are subject to challenge."). More importantly, agency members do not know, in advance, which rules will be challenged. So, once a significant percentage of rules are challenged, agency members know that they need to write the rule to be prepared for future judicial review. As Professor Mark Seidenfeld noted, "there is good reason to believe that rulemaking staff at many agencies do worry from the outset about pleasing a court should the rule be challenged, and therefore do not commit to an outcome before taking such review into account." Id.
54.	See 5 U.S.C. § 706(2)(A),(C) (2000).
55.	See NLRB v. Universal Camera Corp., 340 U.S. 474, 477-91 (1951) (delineating the "substantial evidence" standard of review for findings of fact in formal agency proceedings).
56.	See Motor Vehicle Mfrs. Ass'n v. State Farm Mut. Auto. Ins. Co., 463 U.S. 29, 43 (1983) (finding that the "arbitrary [or] capricious" standard of § 706(2)(A) applies to agency policy determinations and adopting hard look review); El Conejo Americano Inc. v. Dep't of Transp., 278 F.3d 17, 19-20 (D.C. Cir. 2002) (applying the "hard look" standard).
57.	As the Supreme Court elaborated in State Farm, The scope of review under the "arbitrary and capricious" standard is narrow and a court is not to substitute its judgment for that of the agency. Nevertheless, the agency must examine the relevant data and articulate a satisfactory explanation for its action including a "rational connection between the facts found and the choice made." . . . [A]n agency rule would be arbitrary and capricious if the agency has relied on factors which Congress has not intended it to consider, entirely failed to consider an important aspect of the problem, offered an explanation for its decision that runs counter to the evidence before the agency, or is so implausible that it could not be ascribed to a difference in view or the product of agency expertise. State Farm, 463 U.S. at 43 (quoting Burlington Truck Lines v. United States, 371 U.S. 156, 168 (1962)). Thus, an agency's failure to respond to alternatives to its regulation or arguments that undercut its action is "arbitrary [or] capricious" within the meaning of 5 U.S.C. § 706(2)(A).
58.	Sierra Club v. EPA, 356 F.3d 296 (D.C. Cir. 2004), provides a useful instantiation (and discussion) of the differences between Chevron deference and hard look review, id. at 306-07.
59.	This is not necessarily a bad thing, of course; I discuss its desirability in the next section. See infra Part III.
60.	If there are no additional participants, then e-rulemaking will have produced little benefit and nontrivial costs. See supra Part II. If there are additional participants but they do not contribute any new ideas, the agency will learn only that a bunch of people "voted" (via an e-mailed message) a particular way. I discuss the limited value of such a blunt signal in the previous section. See supra notes 29-43 and accompanying text.
61.	The finding of unlawfulness will frequently entail the court invalidating the regulation, but in some cases the court will find the regulation unlawful but nonetheless remand the matter without vacating the regulation. This practice is controversial, both as to its permissibility under 5 U.S.C. § 706(2)(A) (which provides that a reviewing court "shall . . . hold unlawful and set aside agency action, findings, and conclusions found to be . . . arbitrary, capricious") and as to its desirability as a policy matter. Compare Checkosky v. SEC, 23 F.3d 452, 462-66 (D.C. Cir. 1994) (Silberman, J., separate opinion) (supporting remand without vacation under certain circumstances), with id. at 490-93 (Randolph, J., separate opinion) (rejecting the legality of remand without vacation); compare Milk Train, Inc. v. Veneman, 310 F.3d 747, 755-56 (D.C. Cir. 2002) (ordering remand without vacation), with id. at 757-58 (Sentelle, J., dissenting) (rejecting the legality of remand without vacation); compare Ronald M. Levin, "Vacation" at Sea: Judicial Remedies and Equitable Discretion in Administrative Law, 53 DUKE L.J. 291, 360-63 (2003) (defending the attractiveness of remand without vacation), with Daniel B. Rodriguez, Of Gift Horses and Great Expectations: Remands Without Vacatur in Administrative Law, 36 ARIZ. ST. L.J. 599, 601 (2004) (contending that remand with vacation is unattractive).
62.	See supra note 41 and accompanying text.
63.	Interview with Mark Cooper, Director of Research, Consumer Fed'n of America, in Durham, N.C. (Sept. 24, 2005); Interview with Michael Weisman, Co-Director, Reclaim the Media, in Durham, N.C. (Sept. 24, 2005).
64.	It is possible that courts (or Congress) will react to agencies' difficulties in responding to all the new arguments in comments by relaxing the requirements of hard look review. That is, judges or legislators might find that a massive increase in the number of arguments to which an agency must respond imposes significantly higher costs without corresponding benefits, and thus relax the requirement that agencies in fact take a hard look at all significant arguments against their regulation. If so, then e-rulemaking will have had a profound effect on the review of agency rulemakings. The desirability of such a change (which is, of course, purely speculative) is beyond the scope of this paper.
65.	Professor Coglianese is an exception, but even he discusses only briefly the question of the desirability of increasing public participation. Coglianese, Internet and Citizen Participation, supra note 1, at 55-57.
66.	See id. at 39-40 (noting that "public participation can be viewed as intrinsically valuable for citizens themselves, for such participation fosters important personal virtues"); Stuart W. Shulman et al., Electronic Rulemaking: A Public Participation Research Agenda for the Social Sciences, 21 SOC. SCI. COMP. REV. 162, 167-73 (2003) (proposing that electronic rulemaking may foster a more discursive democracy through greater public participation in the process).
67.	See Steven J. Balla & John R. Wright, Consensual Rule Making and the Time It Takes to Develop Rules, in POLITICS, POLICY, AND ORGANIZATIONS: FRONTIERS IN THE SCIENTIFIC STUDY OF BUREAUCRACY 187, 204 (George A. Krause & Kenneth J. Meier eds., 2003) ("Our research demonstrates . . . that rules to which regulatory negotiation was applied took longer to issue than those developed through conventional proceedings, despite the fact that agencies were more likely to conduct regulatory negotiations in situations that were amenable to relatively rapid resolution. In general, we find no evidence that consensual rule making reduces the time it takes to develop rules."); Coglianese, supra note 53, at 1335 ("Negotiated rulemaking does not appear any more capable of limiting regulatory time or avoiding litigation than do the rulemaking procedures ordinarily used by agencies. The agency that has used negotiated rulemaking the most, the EPA, has not seen its negotiated rules emerge in final form any sooner than rules not subject to formal negotiation. Once promulgated, negotiated rules still find themselves subject to legal challenge. The litigation rate for negotiated rules issued by the EPA has actually been higher than that for other significant EPA rules.").
68.	Laura I. Langbein & Cornelius M. Kerwin, Regulatory Negotiation versus Conventional Rule Making: Claims, Counterclaims, and Empirical Evidence, 10 J. PUB. ADMIN. RES. & THEORY 599, 625 (2000).
69.	See Cary Coglianese, Assessing the Advocacy of Negotiated Rulemaking: A Response to Philip Harter, 9 N.Y.U. ENVTL. L.J. 386, 431 (2001): Langbein and Kerwin report that of all the types of participants in negotiated rulemakings, the representatives from EPA and state government gave negotiated rulemaking the highest overall ratings. This is important to recognize because approximately eleven percent of the negotiated rulemaking participants they interviewed were EPA officials and approximately twenty-five percent were representatives from state and local government. In contrast, the sample of individuals who filed comments in conventional rulemakings obviously included no one from EPA and included only three representatives from state and local government. Thus, approximately thirty-six percent of the respondents from negotiated rulemakings were individuals who might be considered 'enthusiasts,' given their higher overall ratings, while only approximately six percent of the comparison group were. Id.
70.	See Coglianese, supra note 53, at 1261 ("Despite all the postulations about how negotiated rulemaking will save time and eliminate litigation, the procedure so far has not proven itself superior to the informal rulemaking that agencies ordinarily use."); Funk, supra note 48, at 96-97 (arguing that negotiated rulemaking undermines the principles and values of administrative rulemaking by changing an agency's objective from serving the public interest to seeking consensus among private parties); Mark Seidenfeld, Empowering Stakeholders: Limits on Collaboration as the Basis for Flexible Regulation, 41 WM. & MARY L. REV. 411, 513 (2000) (arguing that collaborative regulation cannot help to solve existing regulatory shortcomings because "they can succeed in overcoming the adversarial propensities of at least some stakeholders only within narrow regulatory environments").
71.	It also bears noting that with e-rulemaking the new participants do not meet face to face with each other. The point of e-rulemaking is to allow people to communicate through the Internet, rather than face to face. Although this may have many advantages, it obviously does not include the experience of physically meeting together as a community and reaching conclusions around a table.
72.	See Jon Elster, The Market and the Forum: Three Varieties of Political Theory, in FOUNDATIONS OF SOCIAL CHOICE THEORY 103, 121 (Jon Elster & Aanund Hylland eds., 1986) (describing the benefits of participation as the mere by-product, not the main goal, of political activity).
73.	Perhaps the most famous exponent of this view was James Landis, who was the architect of the new administrative agencies created under Franklin D. Roosevelt. See RITCHIE, supra note 39, at 1-5 (introducing Landis's philosophies); see also Gail L. Achterman & Sally K. Fairfax, Public Participation Requirements of the Federal Land Policy Management Act, 21 ARIZ. L. REV. 501, 508 (1979) ("Public involvement programs . . . may easily mobilize dissent and heighten polarization, public frustration, and dissatisfaction."); Jim Rossi, Participation Run Amok: The Costs of Mass Participation for Deliberative Agency Decisionmaking, 92 NW. U. L. REV. 173, 178 (1997) ("A threshold amount of participation is necessary to deliberative decisions, but at some point participation creates significant institutional costs for deliberative administrative process.").
74.	See, e.g., Ernest Gellhorn, Public Participation in Administrative Proceedings, 81 YALE L.J. 359, 380-81, 403 (1972) (arguing for the virtues of public participation in agency decisionmaking); Earl Bonfield, Public Participation in Federal Rulemaking Relating to Public Property, Loans, Grants, Benefits, or Contracts, 118 U. PA. L. REV. 540, 540 (1970) (contending that the interested public should participate in the rulemaking process).
75.	For a good summary, see generally Nancy Roberts, Public Deliberation in an Age of Direct Citizen Participation, 34 AM. REV. PUB. ADMIN. 315, 315 (2004) (chronicling the history of and debate over direct citizen participation).
76.	What actually happened in the case of the Alexander Hamilton entry is that, after the McHenry article appeared, a Wiki community member posted that entry on a webpage devoted to entries that needed to be improved/updated, http://en.wikipedia.org/wiki/Wikipedia:Peer_review, and soon most of the errors and clumsiness that McHenry identified were removed.
77.	Robert McHenry, The Faith-Based Encyclopedia, TCSDAILY, Nov. 15, 2004, http://www.techcentralstation.com/111504A.html; see also Can you trust Wikipedia?, THE GUARDIAN (U.K.), Oct. 24, 2005, available at http://www.guardian.co.uk/g2/story/0,,1599116,00.html (compiling critiques of various Wikipedia entries).
78.	Jim Giles, Internet Encyclopaedias Go Head to Head, NATURE (Dec. 14, 2005), available at http://www.nature.com/news/2005/051212/full/438900a.html (reporting on the Nature investigation that compares Wikipedia and Britannica).
79.	Id.
80.	Indeed, one of Wikipedia's founders left the project and identified "the dominance of difficult people, trolls, and their enablers" as a major problem for Wikipedia. Larry Sanger, Why Wikipedia Must Jettison Its Anti-Elitism, Dec. 31, 2004, available at http://www.kuro5hin.org/story/2004/12/30/142458/25; see also Stacy Schiff, Know it All: Can Wikipedia Conquer Expertise?, NEW YORKER, July 31, 2006, at 36 ("Even Eric Raymond, the open-source pioneer whose work inspired [Wikipedia head] Wales, argues that '"disaster" is not too strong a word' for Wikipedia. In his view, the site is 'infested with moonbats.'").
81.	Wikipedia: Patent Nonsense, http://en.wikipedia.org/wiki/Wikipedia:Patent_nonsense (last visited May 4, 2006); see also Wikipedia: Replies to Common Objections, http://en.wikipedia.org/wiki/Wikipedia:Replies_to_common_objections (last visited May 4, 2006) (noting ways to combat unhelpful contributions).
82.	Wikipedia: Articles for Deletion, http://en.wikipedia.org/wiki/Wikipedia:Votes_for_deletion (last visited May 4, 2006).
83.	Wikipedia: Criteria for Speedy Deletion, http://en.wikipedia.org/wiki/Wikipedia:Candidates_for_speedy_deletion (last visited May 4, 2006).
84.	Wikipedia: List of Administrators, http://en.wikipedia.org/wiki/Wikipedia:List_of_administrators (last visited May 4, 2006).
85.	Wikipedia: Resolving Disputes, http://en.wikipedia.org/wiki/Wikipedia:Dispute_resolution (last visited May 4, 2006). For a fuller discussion of the Wikipedia dispute resolution process, see Joseph M. Reagle, Jr., A Case of Mutual Aid: Wikipedia, Politeness, and Perspective Taking, http://reagle.org/joseph/2004/agree/wikip-agree.html.
86.	See Jimmy Wales, Experiment on new pages [E-mail to Wiki English listserv], available at http://mail.wikipedia.org/pipermail/wikien-l/2005-December/033880.html ("Today, as an experiment, we will be turning off new pages creation for anonymous users in the English Wikipedia."); Article Creation Restricted to Logged-In Editors, http://en.wikipedia.org/wiki/Wikipedia:Wikipedia_Signpost/2005-12-05/Page_creation_restrictions (discussing this policy); see also Katharine Q. Seelye, Rewriting History; Snared in the Web of a Wikipedia Liar, N.Y. TIMES, Dec. 4, 2005, § 4, at 1 ("Mr. Seigenthaler recently read about himself on Wikipedia and was shocked to learn that he 'was thought to have been directly involved in the Kennedy assassinations of both John and his brother Bobby.'").
87.	The political costs associated with a government agency publicly identifying some submissions as worthless are so great, and the benefit to the agency of the agency being in charge (as opposed to a private entity) sufficiently small, that the chances of a government agency taking on this responsibility seem slim.
88.	See, e.g., Cary Coglianese & Gary E. Marchant, Shifting Sands: The Limits of Science in Setting Risk Standards, 152 U. PA. L. REV. 1255, 1282 (2004) ("In setting [air quality standards], or any other regulatory standard, EPA officials need to draw upon the available scientific evidence on the health effects of different pollutants, but ultimately they must make a decision based on factors other than just the science. Standing alone, scientific data on ozone and particulate matter do not, and cannot, provide a principled justification for the level at which the respective air quality standards are set.").
89.	See Wikipedia: Replies to Common Objections, http://en.wikipedia.org/wiki/Wikipedia:Replies_to_common_objections (last visited May 4, 2006) ("As a community, almost all of us are opposed to what has been called the policy of completely 'freezing' particular pages . . . .").
90.	See Sarah Boxer, Mudslinging Weasels into Online History, N.Y. TIMES, Nov. 10, 2004, at E1 (describing the Bush-Kerry incidents on Wikipedia, and noting that, at the times when the Bush and Kerry pages were frozen, "Senator Kerry and President Bush took their places next to the other untouchables in the Wikipedia: Ariel Sharon, Osama bin Laden, Rush Limbaugh and Salvador Allende").
91.	See Los Angeles Times Suspends 'Wikitorials': Users Flood Paper's Web Site with Foul Language, Pornography, MSNBC, June 21, 2005, available at http://msnbc.msn.com/id/8300420/ (reporting the failure and cessation of the paper's experiment with allowing readers to re-write the editorial).
92.	See, e.g., Funk, supra note 48, at 92 ("The concept of regulatory negotiation . . . [reduces] the agency to the level of a mere participant in the formulation of the rule, and . . . essentially [denies] that the agency has any responsibility beyond giving effect to the consensus achieved by the group."); Patricia Wald, Negotiation of Environmental Disputes: A New Role for the Courts?, 10 COLUM. J. ENVTL. L. 1, 22 (1985) (warning that negotiated rulemakings could produce agreements that amount to "pure political logrolling" among the negotiating parties).
93.	See Cal. Democratic Party v. Jones, 530 U.S. 567, 577-82 (2000) (holding unconstitutional a statute restricting political parties' protected freedom to select nominees).
94.	See generally Cary Coglianese, Citizen Participation in Rulemaking: Past, Present, and Future, 55 DUKE L.J. 943 (2006).
95.	See CORNELIUS M. KERWIN, RULEMAKING: HOW GOVERNMENT AGENCIES WRITE LAW AND MAKE POLICY 191-210 (1994); Coglianese, supra note 94, at 949-52.
96.	See Steven J. Balla & Benjamin Daniels, Information Technology and Public Commenting on Agency Regulations (2005)(Midwest Political Science Association); see also Coglianese, supra note 94, at 954-55.
97.	See John M. de Figueiredo, E-Rulemaking: Bringing Data to Theory at the Federal Communications Commission, 55 DUKE L.J. 969, 988 (2006) (noting that a "spike in comments should not necessarily be interpreted as demonstrating heightened interest from individuals"). But see Coglianese, Internet and Citizen Participation, supra note 1, at 52 (discussing a relative dearth of comments submitted through Regulations.gov in spite of its publicity).
98.	Professor Peter Strauss surveyed the members of the American Bar Association's Section on Administrative Law and Regulatory Practice in 2003. Strauss, supra note 32, at 8, available at http://www.ksg.harvard.edu/cbg/rpp/erulemaking/papers_reports/Strauss_Erulemaking_Survey.pdf. He discovered that many of the 320 respondents said that they looked at websites for research purposes (often agency websites), but that only 45 percent had filed comments in any rulemaking proceedings in the previous three years. Id. As Professor Coglianese noted, "[i]f the majority of the most relevant legal specialists do not file comments in rulemakings, we probably should not expect to see a large proportion of ordinary citizens filing comments, even with a more digitized and accessible rulemaking process." Coglianese, Internet and Citizen Participation, supra note 1, at 54.
99.	See de Figueiredo, supra note 97, at 990 (deflating the relative significance of volume spikes in public e-comments between 1998 and 2004); Stuart W. Shulman, Whither Deliberation? Mass e-Mail Campaigns and U.S. Regulatory Rulemaking 5 (Jan. 2, 2006), available at http://erulemaking.ucsur.pitt.edu/doc/papers/Whither Deliberation.pdf ("To date, there is very little evidence that the actual public comment text submitted meets any of our deliberative thresholds.").
100.	David Schlosberg et al., 'To Submit a Form or Not to Submit a Form, That Is the (Real) Question': Deliberation and Mass Participation in U.S. Regulatory Rulemaking 15 (May 5, 2005) (unpublished paper, on file with the Duke Law Journal).
101.	Shulman, supra note 99, at 14-15. Shulman examined a sample of e-mails sent to agencies as part of a campaign by environmental groups to get citizens involved in the rulemaking process via the Internet. He concluded that: At this early stage in the research (and in the epoch of mass e-mail campaigns) there are few indications that online deliberation is enhanced within the current eGovernment configuration in the United States. The mass e-mail campaign in particular appears to be an odd and possibly counter-productive tribute to twentieth century notions of one-directional, non-deliberative, un-reflexive nose counting. Although the medium could be used to promote better dialogue, debate and deliberation, and public understanding, it falls short of the loftier ideals held out by hopeful political theorists. Id. at 14-15; see also Dietram A. Scheufele & Matthew C. Nisbet, Being a Citizen Online: New Opportunities and Dead Ends, PRESS/POLITICS, July 1, 2002, at 55, 69 (presenting empirical findings and concluding that "the role of the Internet in promoting active and informed citizenship is minimal"); Darrell M. West, E-Government and the Transformation of Service Delivery and Citizen Attitudes, 64 PUB. ADMIN. REV. 15, 15-16 (2004) (finding that e-government has failed to transform service delivery or public trust in government); Sylvia Tesh, The Internet and the Grass Roots, 15 ORG. & ENV'T 336, 336 (2002) (arguing that the Internet is merely another tool in the narrow range of citizen options in policy-making participation, and that it moreover robs grassroots movements of the motivational power of physical public hearings).
102.	E.g., Capitol Advantage Home Page, http://www.capitoladvantage.com (last visited July 10, 2006); GetActive Home Page, http://www.getactive.com (last visited July 10, 2006); Kintera Advocacy Product Page, http://www.ctsg.com (last visited July 10, 2006). According to two consultants to agencies, this is part of a larger problem: The Internet as a means for expanded public participation in rulemaking has inspired a sort of rulemaking arms race. Some commenter organizations are investing excessive time and money in technology that will enable them and their members to produce large numbers of comments as quickly as possible in response to any rulemaking. Some commenter organizations are convinced that their position is strengthened by taking ten salient points and masquerading them as thousands of unique thoughts from thousands of thoughtful taxpayers. Under the current e-rulemaking plan, interest groups spend money on the latest software to generate thousands of e-comments, and agencies are forced to invest in sophisticated software that will enable them to mine the thousands of comments to identify the ten salient points. This is a silly, wasteful, and circular game the rulemaking world has engaged in. Emery & Emery, supra note 26, at 8.
103.	E.g., http://www.cleartheair.org/mercury (presenting concerns about mercury and then providing a link along with the statements "Take Action by telling the EPA to strengthen the mercury rule. Click to send a comment to the EPA docket."); see supra note 30.
104.	For a discussion of the insignificance of e-mail quantity as an indicator, see supra notes 29-43 and accompanying text.
105.	See New State Ice Co. v. Liebmann, 285 U.S. 262, 311 (1932) (Brandeis, J., dissenting) ("It is one of the happy incidents of the federal system that a single courageous State may . . . serve as a laboratory . . . .").
106.	See supra notes 11-12 and accompanying text.
107.	It is not clear that centralization will reduce costs for users. For instance, it may be that most of those who would want to send communications to the Environmental Protection Agency or the Federal Communications Commission would find it more convenient to go directly to the agency's website and/or would find a commenting system tailored to that agency more valuable or easy to use. The empirical data on this question are too sparse thus far for researchers to be able to draw conclusions on this question. The point in text, of course, is that there are other reasons to be wary of uniform implementation of e-rulemaking initiatives.
108.	Or the e-rulemaking proposals will entail small costs but have trivial benefits. See supra Part II.