The Decentralized Coordination of Intelligence

The Decentralized Coordination of Intelligence

By Alec Rawls © 1999 (2150 words)

I Introduction

Markets are an example of decentralized coordination. Competitive prices convey the value of resources in alternative uses, inducing people to sell what is more valuable to others than to themselves (often their labor) and buy what is more valuable to themselves than to others, until all resources are (with various qualifications) directed to their most valued uses. Not only does decentralized coordination work but, because it allows people to make use of their private information about where value lies and how to pursue it, it works far better than centralized coordination ever could. Still, for all its power, the decentralized coordination accomplished by markets is limited to a particular sphere. It coordinates the choices people make on the basis of their particular information but does not coordinate the information itself -- people's intelligence -- about the worth of different ends and how they can be pursued. This lack can be remedied. Where markets arrive at prices by compiling bid and ask positions through a large scale process of human interaction, it is possible to decentrally coordinate intelligence through large scale data processing, with results that in many ways mirror the efficiency and incentive properties of markets.

It is easy to think of schemes for achieving a centralized coordination of intelligence. Imagine an inter-net utility that people can log onto to rate, along any number of dimensions, any item of information, entertainment or opinion they comes across. If the utility tabulates the central tendencies in different people's judgments (mean, median and other descriptive statistics), that affords some centralized coordination. One can see what other people think is worthwhile and use that as guidance. But the really useful information would be to know, not what others think is valuable, but what oneself would find valuable if one were to judge each item for oneself. This information can actually be gleaned quite easily. Some people have good judgment from my point of view, some have lousy judgment. I just need to separate the wheat from the chaff about who has good judgment from my point of view, then I can use the ratings of those who have good judgment from my point of view to separate the wheat from the chaff about what I would find most worth my time and/or money.

To systematize this intuition, note that as different people rate many items along any number of dimensions it becomes possible to calculate correlations between different people's judgments of worth. (To calculate correlations the different individual raters have to be consistently identified, which means people have to log onto the rating utility to deliver their ratings. Their identities can be either pseudonymous, or revealed, so long as they are kept track of.) The correlations between different people's judgments of value can serve as measures of how each person would rate each other person as a rater of value. These ratings of other people as raters can then be used as weights to construct weighted averages of other people's rankings of particular items. (Negative correlations between raters cause each to weigh the other's ratings negatively. Thus ratings tendered by people who one considers to be lousy raters are not discarded but are still squeezed for their information content.) Generating these weighted sums for each item for a particular person yields a set of personalized estimates of how worthwhile that person would find each item to be. These ratings of different items could then be ordered into a ranking of what the person would most likely find to be worth his time and or money. It would be possible to form both an overall ranking and rankings within sub classifications (such as movies, or articles on a particular subject), and these personalized rankings could be calculated for all participants.

When the rating utility is refined in this way the result is a decentralized coordination of intelligence that mirrors the decentralized coordination accomplished by markets. Each is able to make his own best use of the resources that others make available, using other peoples ratings to project his own judgment. The efficiencies to be gained are similar to when an economy moves from barter to money. Instead of each person sharing intelligence with only a few others, suddenly each person will be able to make use of everyone else's intelligence, using other people as her eyes and ears to extend a projection of her own personal judgment, allowing everyone in effect to look forward with hindsight.

A fully descriptive name for this utility would be a "coordinating rating engine" (to distinguish it from a rudimentary rating engine that just calculates central tendencies). For convenience it could be called by the generic "rating engine" term, or it could be called a "coordinating engine," to emphasize the decentralized coordination of intelligence that it accomplishes. With modern computing power the calculations involved are entirely practical. As the number of users gets large the number of correlation coefficients that need to be calculated gets very large, but they can be processed separately, meaning they can be processed in parallel, meaning size is no object. With the correlation coefficients in hand, calculating the weighted sums is relatively trivial. They might be calculated on demand, or on a schedule of patterned use (e.g. what is worth reading in the paper today) with comprehensive updates done on background. With some programming and an innocuous amount of computing power (initially) the rating engine could be made available on the inter-net in short order.

But any implementation had better be prepared to gear up quickly, because the incentive structure for each person to participate is very robust, mimicking again the qualities of a market economy. The accuracy of the rating engine's predictions about what a person will find most worthwhile are limited only by the accuracy and amount of his own ratings the person has entered into the utility (given a mass of other people's rating to pattern off of and project onto). The better the information the rating engine has about a person's judgment the better it can project that judgment. Thus everyone has a strong incentive to accurately record for the utility their judgments of many things, creating the mass of ratings the utility needs to project each person's judgment.

This also gives the rating engine a built in flame-resistance: anyone who tries to use the engine to oversell what he has a vested interest in will see his rating as a rater fall and his judgments discounted accordingly. Not only will the attempt to achieve exaggerated influence backfire as people start discounting his opinion, but the rating engine will become a less accurate predictor of how he would honestly rate things. A person might achieve exaggerated influence in one instance but at the cost of diminished influence elsewhere and diminished usefulness of the rating engine as a guide for where his time can best be spent, according to his own judgment. Even this scope for manipulation might be attacked by different modes in which the rating engine could operate. For instance, it might be used to rate, and therefore locate, self-serving misinformation.

In general (as with an economy) the value one gets out of the rating engine is a function of the contribution one puts in, inducing wide, high quality, participation. Once it gets started it will clump together like a planet, just as decentralized economic coordination does. Nothing is more valuable than accurate information about where, in one's own judgment, one's attention would best be directed. The "coordinating rating engine" is the mechanism through which the information revolution's flood of information will be directed.

Section II below develops a calculational scheme for the rating engine. I am holding it back temporarily for possible business reasons.

Section III turns to the complex relationship between the decentralized coordination of intelligence, liberty and monopoly. The rating engine turns out to be a most peculiar animal: a monopoly smashing natural monopoly. The value of the coordinating engine to any one person is a function of how many other people are using it, hence the most efficient market structure for the coordinating engine industry is one huge rating engine used by everyone. Yet the decentralized coordination of intelligence will eliminate the information bottlenecks where gatekeepers currently exercise monopoly power. Empowering individual judgment empowers individual liberty, allowing our economic system of liberty (capitalism) to easily forge new markets, exposing markets currently controlled by gatekeepers to open competition. The gatekeeper role played by present day editors will be the first to fall, with many other dominoes set to fall in turn. Anti-trust can attack monopoly power only in rare large scale cases and with the crude instrument of regulation. By eliminating the information asymmetries that monopoly power often depends on, the rating engine has the potential to expose many small bastions of monopoly power to competition.

Section IV considers one most critical application of the rating engine: the coordination of sense and reason. If users spend some of their energy focusing on this rating dimension it should be possible to systematically separate sense from non-sense. Disinformation and demagoguery will have a half-life of a week instead of years. Acting as each other's eyes and ears, sharing intelligence with those whose judgment one can trust, people will be able to download accurate information about what they would agree with or want to vote for if they were to study any particular matter. Pushing the implications, not only would people be able to call up accurate guidance about who they would want to vote for if they were to study the candidates, but since they will know what they would want to vote for on any issue, representatives become superfluous. We will be able to trade in our representative democracy, with its inherent corruption, for the real thing: a fully competent direct democracy, wielding all the powers of Congress. Again, I hold some elements back for possible business reasons.

Section V looks at how the decentralized coordination of intelligence, by providing a systematic method for collecting and disseminating judgments about merit and worth, can serve as an engine of justice, directing credit (and remuneration) where credit is due.

II The calculational model

Witheld temporarily for possible business reasons.

III Liberty and Monopoly

Consider the historical model of editorship. So long as the only avenue to publication was the printing press it was necessary to have a few people to decide for the rest what items would be made widely available through publication. The drawback of this system is that the opinions of most of those who read a given piece come too late to enter the publication decision. Editorial judgment is by necessity delegated to a few people who must pre-select what the others will have available to read. Editorship in this case forms a bottleneck where leverage and monopoly and can operate. This paradigm extends far beyond the realm of publication. In general, present institutions rely on editors, credentialism and political power to determine which views and expressions will be promoted as most worth attention. All such institutions only make use of a few people's privileged evaluations, creating monopoly power and letting most evaluative judgments go untapped.

Electronic communications fundamentally alters the possibilities. No longer is it necessary to have an editor pre-select what will be made available. So long as something is posted on a web-site it is available to everyone at all times. Virtually everything, certainly any text, can be universally published for close to zero cost. There is no need to decide what will be published. Everything can be. The question of editorship is thus reduced to the problem of passing on judgements about what is worth attention. Traditional editors could still be used for this. A person or editorial board whose judgement some readers want to listen to could still make recommendations about what to pay attention to. But there is no reason only to listen to make use of a few people's editorial judgement. On the one hand, many people will have editorial judgements to pass on, as we with the forwarding e-mail messages that people think are worthwhile and other word of mouth. Such mechanisms decentralize editorial intelligence without coordinating it. Or we systematically separate the wheat from the chaff by using the rating engine to decentrally coordinate intelligence. Everyone's judgment can be collected and passed on to everyone, everywhere, automatically.

Consider the effect on the magazine and popular book businesses. The rating engine will rate all items, be they published on paper under the gate-keeper model or published electronically by independents. Readers will read whatever the rating engine predicts they will find most worth reading, no matter the source. Simple web pages will be able to charge by the page and out-compete national magazines amongst those who are predicted to rate their material higher. The gate-keepers will have no leverage. Their pockets of monopoly power will be eliminated. Contrast this with the situation now in book selling. People who enjoy novels face a huge problem of sorting the wheat from the chaff. They love one book out of fifty and it is not worth it to them to wade through the forty nine looking for the one so they find ways to turn the odds in their favor. They may rely on best seller lists, or the editor of a series. Most commonly they return to authors whose previous books they have found worthwhile. Thus until new writers have an initial success they are in a catch 22. No matter how good the book, the gate-keepers know that most readers will not read an unknown writer so they rarely give new authors a chance, even if they think people would value a new book extremely were they to read it. In our supposed age of plenty, even the very best writers are prone to starve. The rating engine will change this. Electronic publishing is close to costless and intelligence about what is worth reading will automatically be disseminated to everyone. Those books will go to paper that the rating engine predicts will be valued the most by the most people. Merit will be rewarded. Credit (and remuneration) will go where credit is due. We will become prosperous in art as well as science. Not a bad little externality.

The effect of the rating engine here is to greatly facilitate the creation of new markets. Instead of having to forge a market for his wares through word of mouth, or by getting help from a gate-keeper, an author gets automatic universal dissemination of all feedback. This ease of market creation can be magnified even further by people sometimes using the rating engine as a guide to editorial activity. One could ask the rating engine to locate, not items with the maximum expected value to oneself, but items where the one's judgment would have the highest expected editorial productivity, or value to others. (Specifics witheld for possible business reasons.) In this role, the rating engine could serve as an engine for facilitating the initial discovery and propagation of merit. Imagine a world where every important contribution is husbanded, no matter how lone the source. The decentralized coordination of intelligence can enable it.

The fundamental character of the rating engine design is that it does not try to centralize judgment but tailors ratings to what each individual would find most worthwhile. Options for tailoring the rating engine to fit the user's intentions are a continuation of this theme. One might, for instance, just look to see what a particular person whose opinion one trusts especially thinks on some subject. Weighting ratings by the raters rating as a rater is one available resource, but people can also use their more particular information however they want. There would also be multiple ways to employ the rating engine in an editorial mode. (Witheld for now for possible business reasons.)

Other possible modes of operation arise in the academic world. Consider the case of physics where the gate-keeper model has already been superseded. Working papers are posted on web-sites and the community is small enough that word gets around about which papers contain important new results or promising directions. The paper journals have become simply journals of record, publishing the papers that physicists have taken up and followed. The mere fact of electronic publishing, even absent any systematic mechanism for making use of feedback, has proved sufficient to dominate the gate-keeper model and blow it out of the water. Mated to a powerful rating engine that people can use to extend the reach of their own judgment to every corner of the inter-net at once, the gate-keeper model will be superseded far and wide.

This is crucial. The reason the gate-keeper model fell in physics is not just because it is slow, but because it is inherently corrupt (whether or not it is less corrupt than un-refereed models). There is the unavoidable problem of industry builders who get ensconced in editorial roles at academic journals and then direct those journals to build on their own kind of work. Such people are not necessarily bad. They think that their own predilections do point to the most value, and their disinterest in other directions may make perfect sense for them. No one can or should do everything. The problem is that the system is inherently biased. Everyone's judgment should register (and be given the weight that others have learned to give it) not just the judgment of a few. We can do better now. We don't need to employ the monopolistic gate-keeper model and that means monopoly will be superseded, because monopoly is inefficient. Head to head with competition it will get crushed every time.

Where it is desirable to retain peer review, this can be enabled by building a peer review function into the rating engine, allowing users to look at how some credentialed class rates views on a subject, or those members of a credentialed class who report they have examined a study carefully and assert expertise. A user could either look at the raw ratings (mean and distribution) from the specified sub-group, or he could weight those ratings by each rater's rating as a rater, to see what those academics whose judgment he agrees with think. Any academic who oversold her qualifications for rating a particular item, or the thoughtfulness of her rating of an item, would tend to have her rating as a rater fall. As with the rating engine in general, the peer review function has built in flame resistance.

When peer review is processed through the rating engine it is a simple matter to compare the judgments of a particular peer group with the judgment of other peer groups or with the judgment of all people. Such comparisons have important service to perform in academia. Academic fields have always shown a certain tendency to become self-referential, descending into esoterica that has no relevance to the world outside. When combined with the powerfully illiberal left leaning tendencies common in some parts of academia, departmental isolation can lead to infestations of willful error. It is a staple of the left for instance to ignore the possibility of honest scholarship and assert that all argument is agenda driven, then use that slander to justify the left's own practice of making whatever arguments can be made for preconceived conclusions instead of following reason and evidence. Literature departments across the country are full of literature Ph.D.'s claiming they are doing philosophy when they extrapolate residual uncertainties of language and communication into claims that all meaning is radically uncertain (in an attempt to escape the requirements of making sense). Race and sex activists direct their efforts, not towards understanding, but towards intentional misunderstanding, finding ways to see themselves as victims in order to claim redress. Wherever this ilk achieves power it does its best to metastasize, hiring and publishing only its own.

Enter the rating engine. By looking at peer review across disciplines, anyone will be able to see how philosophers in philosophy departments grade "philosophers" in literature departments on the epistemological foundations of Marxist literary criticism: F minus. No longer will those who reject honest reason be able to hide behind arcane language, departmental power or ad hominem attacks on critics as racist or sexist. The ideological armies of the left will be exposed as political operatives for whom truth, sense and reason are obstacles to be overcome. The rating engine will identify the worst offenders by name, identify which departments in which schools suffer a majority infestation, giving universities the wherewithal to return to their mission of honest scholarship.

One may think I am being too hard on the left. We can let the rating engine sort that out. The point is that by making the dissemination and coordination of judgment nearly costless the rating engine will pull the rug out from under those bastions of monopoly power that take advantage of asymmetric information. Claiming integrity will no longer be enough. All who do not actually have integrity will be exposed. Similarly with competence. This will put great pressure on the tenure system itself, which constitutes economic collusion by the already employed and is the central locus of monopoly power in academia. At present tenure decisions are totally one sided: what did this person do that is so bad that tenure should be revoked? A rating engine would put the other side of the equation at everyone's fingertips: how much better are the other people we could hire? In physics in the 1980's an entire generation of young physicists was put to the blade to preserve of a cohort of dead wood that had lost touch with the frontiers of the discipline 20 years earlier. With that disparity on constant display there is hope that tenure can one day be extirpated. Academia loves its socialism but, like all socialist ideals, the dream of lifetime security for a privileged few fails to account the costs.

Smashing monopoly is a secondary effect of the decentralized coordination of intelligence. Its primary effect, both inside and outside of academia, will be the guidance it offers about what is worth paying attention to. The more a person puts into the rating engine the more she gets out. An academician's career consists of making contributions to learning and judging the contributions of others. No other group is positioned to reap so much assistance from the ability to systematically act as each other's eyes and ears. Most importantly, the rating engine will connect up academia with the rest of society. So much of the important progress these days is being made by people in business and private life. On the other hand, what understanding does repose in academia (I think of my own field of economics) rarely makes its way into public discussion or policy. In so many areas academics write for each other, no matter how irrelevant they have become. The rating engine will tie everything to a single metric of worth (as judged by the user). It is analogous to prices in an economy. By allowing all inputs and outputs to be measured on a single metric of worth, prices provide a way to measure which productive efforts have the greatest net value. When people start using the rating engine for guidance about what is worth paying attention to, writers and researchers will start wanting to be important not just within some solipsistic circle but to the world, which is the measure by which efforts should be guided.

Moving away from academia, the coordinating engine might be able to eliminate information bottlenecks and their attendant monopoly power at many points within organizations. Imagine a research division of a company using a coordinating engine to let employees be each other's eyes and ears about which research developments are most worth following. Credit and merit would all be out in the open, greatly limiting the scope for anyone to falsely take credit or pass blame. Correlation coefficients would be available for all levels of management to see, allowing higher ups to see how each employee's judgment is esteemed by the rest, instead of having to rely solely on lower level managers for personnel decisions. Just having the mechanism for decentralized coordination of intelligence in place would allow it to be used for passing on all kinds of judgments about merit. Scope for arbitrary power would have to be reduced.

There are limitations on what the rating engine can do. One springs from the problem of sample bias. For the most part sample bias is not a concern for the decentralized coordination of intelligence. Since people are being rated as raters their biases are constantly being accounted for. (Average ratings, by contrast, do transmit sample bias.) There is, however, one application where the system of rating people as raters would require an unbiased sample in order to generate meaningful results. That is in product reliability ratings, where those who had particularly good or bad experiences might be more likely to offer feedback. Since these experiences would presumably be uncorrelated with the person's rating as a rater, this bias would not be offset by accounting the person's rating as a rater. Disclaimers would have to be attached to this use of the rating engine. To supplement at this point, product reviews and unbiased sample results could be referenced where available. (This kind of service is already appearing on-line.) There will be other limitations at the outset as well. What is impressive is how much territory that is clearly amenable to cultivation by the rating engine. Consider:

IV Coordinating sense and reason

Most people will start out using the rating engine primarily to locate the most compatible entertainments. That is great. It is a big market and will jump start the decentralized coordination of intelligence. No longer will people be seen wandering the aisles of video stores, scanning hundreds of titles with no clue which ones they would think are any good. People will simply download a prioritized list of what they can be expected to find most worth their time, filterable for rankings by genre, documentary accuracy, humor, or whatever one is interested in.

In addition to such mundane uses, the coordinating engine will also have revolutionary uses. People will be able to use it to coordinate all manner of intelligence. A critical dimension for rating items of scientific, political and practical import will be how dedicated the author is to following reason and evidence: how he rates on scales of sense vs. non-sense and truth value. Once those who adhere to honest reason have identified each other through the rating engine they can use each other as their eyes and ears. Disinformation will have a half-life of a week instead of years. Sense will be systematically separated from non-sense for all who can follow enough sense on at least some subjects to tap into this river.

The fact that the engine outputs simple ratings does not in any way limit how deep and meaningful those ratings can be. The sophistication and usefulness its predictions will be limited only by the amount of effort people put into entering their ratings into the system. For the majority of items that are read quickly, or viewed for amusement, a few seconds for a one dimensional rating is all that will be warranted, but when a person takes the time to read something carefully, if he also takes the time to give an accurate judgment of such things as the care and competence with which the author follows reason and evidence, then the rating engine becomes for that person a very sensitive instrument for locating items that manifest care and competence in the following of reason and evidence. (For this calculation the engine would use as its weights the correlations between pairs of people as raters of reason and evidence). As long as other people are putting their intelligence into the system one can access that intelligence by contributing one's own.

Ten years from now you will only need to press a "print" command in order to take into the voting booth a prioritized list of who you would want to vote for if you had thoroughly researched every candidates background and positions. All the rating engine will require is a sufficient sample of your political judgment, rating hundreds of op-ed articles for instance, for how much they get right. On any legislative issue or ballot measure people will be able to identify, thanks to the labor of others, which summary statement on an issue is the one they would agree with if they were to study it. Special interest groups will no longer be able to rely on the ignorance of the larger population of people who each have less at stake on an issue (though collectively they may have more at stake than the special interest group). So long as some people are paying attention the rating engine will assess the issue for each person with a high degree of confidence, making effective decision-making close to costless.

One might think it must be impossible to predict what a person will come to think about an issue were he to study it for himself. In the words of J. S. Mill: "Nobody's synthesis can be more complete than his analysis." But the rating engine is not one person. In projecting someone's judgment it makes use of information that the person does not have. Further, the rating engine neither analyzes nor synthesizes. It does not parse intelligence in any way. It does not form a model of a person's judgment and then project that model. Rather, it simply coordinates different people's judgment: I rate you as a rater. Whatever characteristics of judgment you and I share are always represented in the rating engine's calculations by your actual judgment, properly weighted. Our overlapping characteristics of judgment are not in any way abstracted or specified. We only need to have some overlapping characteristics of judgment (however correlated) for the rating engine to work. Of course it won't predict perfectly. I could make better use of your information if I were to make it mine and actually have your expertise. But that wouldn't be progress. That is where we are now. The rating engine will allow me to make effective use of all the knowledge that I don't have!

A skeptic might also think I'm being rosy about the implications, assuming that decentralized coordination will successfully sort out sense from nonsense. Isn't there also a negative potential to this technology? Couldn't the systematic linking up of like thinkers lead to factionalization, not uniting groups but splintering them apart from each other? Absolutely, but this is perfectly benign. Those who follow reason and evidence will form a great trunk. The splinter will be those who devote their intelligence, not to following reason and evidence, but to making the best sounding case they can for preconceived conclusions, avoiding contravening sense the way a chess player avoids checkmate. This pathology is extraordinarily common. Our public debates are full of demagoguery and manipulation, using disinformation and subtle unreason to sway opinion. But disinformation and unreason cannot survive when people are acting as each other's eyes and ears. Unreason will splinter off, and burn away. Consider some examples.

For the last two decades our politics have been mired in a "political correctness" where anything that can possibly be misunderstood as racist or sexist or mean-spirited or un-caring has been intentionally misunderstood, as a basis for claiming victimization and redress. People have been afraid to challenge it or be painted with the same brush. How far this manipulation can be twisted and still thrive is shown by the success of the current campaign against "environmental racism," a ruse by white radical environmentalists to ban economic development in minority areas.

The premise of the "environmental racism" crusade is that sources of pollution tend to be disproportionately located in minority areas, when in fact suppressed EPA studies, along with every scientific study (as opposed to the activists own advocacy studies), indicate that sources of pollution disproportionately occur in majority white areas. The logic of this correlation is obvious. Pollution is a byproduct of productivity and whites, as measured by income, have higher average productivity than "minorities" (which usage only includes those minorities that tend to have low earnings). The environmentalists' contempt for the truth is matched only by their contempt for minority interests. They condemn all industrial development in minority areas on the grounds that the pollution generated would disproportionately affect minorities, and thanks to the Al Gore sponsored decision to find actionable "environmental racism" under the Civil Rights Act whenever pollution has a "disparate impact" on minorities, the environmentalists can even use the EPA as their muscle, condemning disproportionately minority areas to permanent poverty. It is the baldest hypocrisy imaginable: anti-capitalist environmentalists using pretended concern for minorities to sacrifice minorities to their Luddite environmental agenda. Yet the scam is blooming like wildflowers because people are easily cowed by charges of racism when few people have the time to learn the truth.

The decentralized coordination of intelligence will tip the balance the other way. The harm to minorities and the radical anti-capitalism of the "environmental racism" crusade will immediately be exposed to all people everywhere who are not themselves enemies of truth. Attempts to deflect reason with ad hominem charges of racism will also be exposed to everyone for what they are. It is the demagogues who will be humiliated, scorned, expelled and ignored henceforth. People will not even have to learn what "environment racism" is to learn that it is a scam, and they won't even have to learn that it is a scam to get reliable guidance about how they should vote on laws about it. The rating engine will tell them where they should look for understanding if they want it, but if a person just wants a thumbs up or down (on a scale of plus ten to minus ten perhaps) on an existing or proposed "environmental-racism" policy, all he has to do is pay attention to some issues, submitting enough judgment for the rating engine to reliably project off of.

This ability of the rating engine, not just to guide people to what is worth attending to but also to rate positions and proposals, is critical because few people can, never mind do, learn about very many issues. Consider another example. Few people know much about the issues of gun rights vs. gun control and the casual information they do get is filtered through our virulently illiberal press, made up almost entirely of people who believe as an article of faith that they way to reduce crime is to guarantee to criminals that all potential victims will be disarmed. Compounding the difficulty of learning about this issue is the fact that research on guns is characterized by a total split. On one side are economists and criminologists who, following the best methods in their fields, find a large and consistent correlation between legally owned guns and lower crime rates. It is an example of the instrumentality of liberty. When people whose rights are unimpaired by criminal history are allowed to account their private assessment of the dangers around them by arming as they feel is wise, much crime is deterred or interdicted.

On the other side is an advocacy statistics industry, publishing in medical journals and employing a "medical model" that begs the question by likening guns to germs. The medical literature pretends to its audience to be "the" research in the area by the simple device of citing only each other while studiously refusing to acknowledge to the opposite conclusions of properly conducted research. The anti-gun press weighs in by publishing the steady stream of phony factoids produced by the advocacy statistics industry while consistently ignoring the honest research in the field, all to serve the public interest of course, given the premise that guns are bad.

Absent a coordinating engine to sort through the systematic disinformation, uncovering the truth is a daunting job even for someone who is very interested in the issue. Those who don't even want to think about guns are easily seduced by the goal of eliminating guns when the fact is that they can be more secure in their personal preference for eschewing guns if other citizens are allowed to go armed. The criminals don't know who is armed and who isn't and so are deterred from attacking anyone, reducing the likelihood that any particular person will need to defend himself. (Since Texas has started enforcing the right to bear arms its crime rate has fallen so far that the rate of gun ownership has actually started to fall.) (Cite) Even with the disinformation campaign, and with people's susceptibility to demagoguery on this issue, the truth is readily apparent to any honest person who investigates far enough. There simply are not two honest sides to this issue anymore. However few people know it, the facts are in. That makes it a piece of cake for the rating engine to sort out. The advocacy statistics industry and the bias of the press will quickly be exposed as intellectual and moral bankruptcy in the service of totalitarian instincts, hating rather than loving the triumphs of liberty.

There are literally thousands of examples of public discussion divided by corruption and stupidity. It occurs on every legislative issue and in an appalling percentage of intellectual debates. Those whose modus operandi is to serve ulterior motives, or fight for their preconceptions rather than follow reason and evidence, are everywhere. It is a vast pollution that the decentralized coordination of intelligence will easily filter out (but which mere compilation of ratings, without rating raters as raters, cannot). With decentralized coordination, the holders of untenable views will indeed unite, and become progressively more isolated, suffering constant attrition as their more honest constituents are led progressively toward more tenable views. The coordinating engine focuses light like a magnifying glass, igniting nonsense and disinformation like tinder. Ideologies built on distortion and error will not be able to avoid the flame. The more mutually reinforcing the error, the more fuel for final self-immolation.

Stated positively instead of negatively: accurate guidance about what is worth paying attention to will lead judgment itself to proceed apace. To believe that the empowerment of individual judgment will inevitably improve judgment is just to understand the instrumentality of liberty: that the great engine of progress in the discovery and pursuit of value is the freedom of individuals to choose their own direction and make their own best use of the resources that they can command through voluntary agreements with others. Instead of all following one lead, all follow their own judgment, allowing more intelligence to be made use of. More value is discovered, which discoveries are shared, leading to greater attainment of value. Liberty has proved itself the great engine of progress a hundred times over and it will with the decentralized coordination of intelligence also. The fact there will always be numerous disagreements between people who follow sense and reason is the mark of the human condition, the incompleteness and growth of our art and understanding, the vitality of a truly civil society.

Even more positively, my dissertation research (out of which this scheme for the decentralized coordination of intelligence emerges) develops a complete understanding of the basic structure and implications of moral science (click Moral Science at www.rawls.org)The principles of a complete analysis of value, from ends to means, are available and can serve as an engine in themselves for sorting sense from nonsense. Powered by the decentralized coordination of intelligence, this sorting of sense from nonsense will easily reach the natural majority of people whose instinct is place sense and reason ahead of ulterior motive and preconceived conclusions. We need not have the complete basic structure and implications of moral science in hand to trust -- and be certain -- that the empowerment of judgment will power the improvement of judgment, but it does strengthen the argument for certainty.

Once use of the coordinating engine has matured a bit by shaking out the nonsense and misrepresentation so common in our public culture today, to the point where people can be confident of the guidance it offers about how they would want to vote if they were to study any piece of legislation and its subject, I presume we will want to move to direct democracy. Once The People can costlessly vote their own will on each issue, We will no longer tolerate a system of representation that regularly contravenes our will. (Consider the large majority that -- correctly -- reviles carpool lanes, but this non-sense has become pork, hence unstoppable.) The system of representative democracy is inherently corrupt, both in the electoral process and in the power it invests in some to speak for others. It is not a vitiating corruption. Absent a workable system of direct democracy, representative democracy is certainly the next best thing. But we will no longer need it. We won't need legislators to write bills for us. Any of us will be able to do that, alone or in collaboration. The coordinating engine will direct people's attention to what is worth consideration, allowing those who attend to a particular issue to act as the eyes, ears, and brains, for the rest of us. Neither do we need the special wisdom and expertise of elected representatives, when we can instead draw on the much larger wisdom and expertise of all.

V Capturing the Natural Monopoly

Temporarily witheld for possible business reasons.

VI Justice and Desert

The prediction engine systematically promotes merit, as each person sees merit. Thus it can serve as an engine for giving credit, and ultimately remuneration, where these are due. While the prediction engine also has implications for the public sphere, as with the possibilities for direct democracy, the great potential of the rating engine is the hope it offers for advancing justice in the private sphere. A just tax system is limited in how much justice it can accomplish because it is stuck with the incomes generated by the private sphere as its starting point. These may be substantially unjust, with no way to know how they are unjust. But just as the decentralized coordination of economic activity enables competition and brings reward substantially into line with productivity, thereby promoting one element of justice, so too can the decentralized coordination of intelligence enable competition and bring reward more into line with productivity. To explore how far the rating engine will naturally provide this service and how far it might be pressed into this service we must look to how injustice arises.

Under ideal conditions of competition and market completeness and perfect information, a market economy will pays people according to their marginal productivity. But ideal conditions are the exception and most people are subject to arbitrary or monopoly power on various levels. Asymmetric information allows all kinds of scope within organizations for usurping credit and shifting blame. Even at the theoretical level the question of how to allocate credit for joint products can be problematic, as when multiple contributions are critical so that different people's marginal product each amount to total product. Or a person might be a member of a group whose behavioral tendencies he does not share, which fact people who do not know him cannot know, perhaps deterring them from giving him a chance. Intellectual property rights are very imperfect. If an idea is valuable but not patent able, or patent able but not patented, inventors can easily be cut out of their due. The particulars of how people get shortchanged are as numerous as street-corners.

As discussed in the previous section, the rating engine will strip away certain kinds of monopoly power, particularly in the realms of editorship and any gate-keeping role in the world of ideas where it will systematically promote merit, broaching markets and moving us near the perfect market ideal. In this realm it automatically goes a long way towards directing credit and remuneration where they are due. The other speculations of the previous section -- the fall of tenure (the locus of monopoly power in academia) and the weakening of credentialism more generally -- will also be great victories for justice if and when they occur. For the most part though, the specific uses which the decentralized coordination of intelligence will be put to use cannot be predicted. It is general purpose technology that individuals and organizations will experiment with. Rating dimensions themselves will be subject to innovation, targeting what are discovered to be key avenues for coordination. I cannot offer any systematic insight into how far the decentralized coordination of intelligence will be able to root out monopoly except to say that rooting out monopoly is its nature. Wherever the coordinating engine shines its light, monopoly power will be undermined. But if I cannot be systematic, I can still point out one very interesting place where the rating engine might be pressed into service to directly serve credit and due: in the realm of intellectual property rights.

The ideal patent rights system would provide people with incentives to declare their advances, then it would allow everyone to make use of those advances. Our current system falls short on both fronts. Inventors have one year, after an idea is made public, in which to file a patent on an invention incorporating the idea. This creates a powerful disincentive to make incomplete inventions public, and many inventions are very difficult to complete, even if the central idea is sound. But until an inventor has patent rights, going to others for help in completing an invention is very risky. Those who don't have the resources to carry an invention through to completion themselves can be stuck, and often end up keeping their ideas secret in hopes that they will someday be able to prosecute them. Instead of providing incentive to make ideas public, our current system often does the opposite. (The European system, where any publicity before patent is filed vitiates patent ability, is even worse in this respect.) At the same time, the system of granting twenty years of monopoly power to successful patent claims rewards invention but at a high cost. Others cannot freely use the idea and anyone who wants to seek further patent able advances in the same vein knows that they will have to arrive at a negotiated agreement with the existing patent-holder in order to make any money off of what they find, increasing the riskiness of research in that direction.

Consider an alternative. People make their advances known through electronic publishing and register their declarations electronically with the patent office, which keeps a record of who declared what when. Ideas put forward can be either partial or complete. Other inventors use the rating engine, perhaps along with search engines, to seek out good ideas that are relevant to their own efforts, or new ideas that they might be able to take a little further. As ideas reach fruition, the various contributors can collectively seek a patent. If the invention meets the requirements for patent ability (usefulness, non-obviousness, etcetera) the contributors are issued rights to a percentage, say 15%, of the gross sales of any product that incorporates the invention. Anyone will be free to make use of the invention under a standard open licensing agreement where they have to pay this percentage of gross to the inventors. If further patent able innovations are made along the same lines, and the new innovations are patented and incorporated into a product, the manufacturer pays the same percentage of gross to the inventors, who now include the inventors of the later innovation. Such patent rights need not expire quickly, as patent rights do now, since they do not impede general use of the patented idea, though they should expire at some point, to keep the system from getting too complicated.

All it would take to make a system like this work is a process for allocating credit to the various claimants. At present the patent courts have the very difficult job of trying to separate possibly overlapping patent rights on the basis of priority. Dividing credit might in many cases be easier. Further, the patent system might be able to use the rating engine itself to coordinate judgments about how credit should properly be divided. There could even be a tax on the open licensing fee to be used to pay people for the time they spend assessing the record of claimed contributions to an idea and judging how credit can best be divided. Some such automatic process would have to be set up because, as innovations pile on top of innovations, credit would continually have to be reallocated. I am aware that I am being very vague about how the rating engine could function in this capacity, but it does have some advantageous qualities for this use: it is flameproof, it is broad based, with strong incentives to participate, and it is sensitive to important new information, with editorial modes that can draw attention to conflicts. There ought to be some way to make it work. The key would be using every means to mark to market as many judgments of value as possible.

Going off on a bit of a tangent, LINUX is a very interesting example of decentralization. It is a free version of the UNIX computer operating system, developed through an open ended collaboration among independent programmers called open source programming, all of whom can make use of the existing code to develop further contributions on condition that others can make use of their contributions on the same terms. It is a wonderful example of people's willingness and desire to coordinate in a decentralized fashion, but from the point of view of justice the outcome is terrible. None of the programmers stand to receive any financial reward for their contribution, and the resulting free software stands to keep other operating system developers from reaping their due. Remember, the anti-trust charge against Microsoft regarding Netscape is that it attacked Netscape with the ultimate predatory price for its own inter-net browser: zero. The intention behind Linux is the opposite: to be generous and self-less, but the effect is identical: the ultimate predatory price, undermining justice in the industry. I am no expert on open source lisencing agreements but it seems to me that open source as a method for coordinating contributors does not require that the source code be open to users as well. It does not have to be free. Rather, the coordinators of the project could set up a system for allocating credit among contributors while in the market for operating systems acting as a profit maximizing competitor. The programmers would then get paid their proper due, given a reasonable system for allocating credit. I nominate the rating engine.

Top of Page