[ This text was started in 1991 and last updated in January of 1993, before the Web and my experience with software agents. I also have more recent writings - see my work on Collaborative Filtering, Hypereconomy, and the new Newsfilter project ]

Message Rating System Proposal

SUMMARY:

This proposal describes the features of a message rating system that may be used as a smooth extension of practically any existing information storage / transmission system, to provide users with increased control over what they read and who reads their messages and give them an opportunity to benefit from the quality of their postings as well as incentives to improve other users' life. The system is also expected to offer many other benefits, some of which will be described later in this text.

The GOAL of this text is to share the main ideas of the message rating system with people who can hopefully help shape it into a practically implementable system specification and take part in developing it. The text below does not present such a spec, but just describes various ideas on the subject that I consider interesting and appropriate for an informal introductory text.

Please see my essay page for other related suggestions.

Idea in brief

Users rate the quality of the messages they receive. These ratings are reported to the system that collects profiles on the posters' and readers' reputations and interests in various topics. This information is used to optimize the traffic for all users according to their personal interests. The quantity,quality and timeliness of ratings, as well as the quality of messages improve the users' selection of incoming messages, thus making sure that diligent response is in the best interest of every user. I believe that such a system can be used as an extension of many existing Internet services, including e-mail, news, gopher, WWW, etc., as well as areas where the 'underlying interest' remains off-line, such as book and movie ratings.

Glossary

system: - a system of message storage and traffic employing ratings and agents in the mechanisms described below.
message: - any piece of information reachable through the system.
rating: - a [numeric] value assigned to a message. One message can have many different ratings assigned by different users. Also, the agent gives the user a rating for every unread post. [Maybe, different words should be used to distinguish these two kinds of ratings]
user: - any entity that generates messages and/or ratings in the system.
reputation = capital =credits: - the system currency. Belongs to users. May be used by users to pay and receive payments for messages and ratings generated by others and themselves.
agent: - user's personal part of the system. Configured by the user to control message and capital traffic to and from the user.
system manager / owner: - a person or group that runs the system, manages hardware and software, sets the rules, resolves disputes, etc.

How it is supposed to work

Writing messages

When you write a message, you may pay any number of credits for it, thus indicating how much importance you assign to it. You may not want to pay for it if you don't have enough reputation/capital or think this message is not that important, or just don't feel like it, but I may instruct my agent to discard all messages rated low by their own authors. Also if you have a huge reputation on the system, you may hope that all smart people instructed their agents to read all your stuff regardless of the number of credits you put on it. You may also set a minimal price you want to get from each reader for this message.

Reading and rating messages

When you read a message, you may rate it by paying an arbitrary number of credits for it. These credits will go to the author of the message and to those users whose ratings were most helpful in selecting it. Again,you don't have to rate anything (rate it with 0), but in this case the system will not learn your interests,and you will not be able to use the benefits of the rating-based selection, neither you will be paid for your ratings.

Also, a message may have a minimal amount of credits its author wants you to pay for reading it. The payment should be made automatically by your agent before you receive the message.You may instruct your agents to send you only free messages (except from certain people, topics, etc.) If you think that the required payments unfairly restrict your access to some messages, consider the rights of their authors not to share their thoughts with those who [consistently] rate them as worthless or are not knowledgeable enough in the topic to have deserved any reputation.

You don't have a right to read my message if I don't want you to do it! Though the authors may choose to [instruct their agents to] share their messages for free with all or some other users - e.g., new to the system.

Getting messages rated

The messages get rated for you by your agent (this rating is different from the rating you assign to a message after reading it). An agent is basically a custom - configured program that takes into account the repu- tation of the author of the message, other users' ratings and a number of other things. When using other users' ratings, an agent takes into account correlation of their ratings with yours on a given group of messages. A group of messages may consist of all messages specified by their topic, author, title, keywords or whatever else you may instruct your agent to keep track of. The suggested default would be a topic (e.g. as defined by prefix on extropian list).

Selecting messages

Your typical instruction to your agent can be: "get me all messages you rate over 8.5, all free ones you rate over 6.0, all on topic T, all by user X rated over 0, and if total is fewer than 20 messages, add the best-rated up to 20. Also give me all unrated messages on my professional topic P (I'll have a chance to earn credits rating them, so put them on top please! -all other message sort by their ratings in descended order). And yes, never pay more than 10 credits for a message without my approval." - this shouldn't be difficult to do by providing a configuration info to a standard system agent.

A list of messages that received insufficient rating will be sent to the user periodically; it will contain message #, subject and rating. These messages may then be sent on user's request.

[more on] Paying for ratings

You rate a message by paying for it. The payment goes partly to the author of the message, and partly to the raters who helped you choose it. (The discussion of how to divide the payments between the authors and the raters is beyond the scope of this text). Raters divide their share of credits depending on the correlation of their ratings for this message with yours, and how early they rated it (but not their reputations - fair game!).You may instruct your agent to ignore ratings made by certain users specified by name, minimal reputation or other criteria. Making the payment dependent on who rated the message first makes sure nobody will get a piece of the pie by just repeating somebody else's ratings,and also gives incentives to users to read messages early and keep looking for pearls that may have passed unnoticed.

Your relation with your agent

You have write access to some parts of your agent[s data], e.g. rules for reading other users' messages and distributing yours, read-only access to other parts (e.g. your reputation[s]), and no access to some other parts (e.g., I want you to allow your agent use my ratings, but I don't what you to know what they are. Thus, your agent respects other users' privacy and, to a certain extent, private copyright laws. Also, the agent may keep track of which messages you read and how you rated them, calculates correlation of your ratings with those of other users', etc. You may share all or parts of your agent with other users on any conditions. You may choose to make different parts of your agent, including your name, reputation[s] and rating rules visible to other users on any conditions you may want to impose. You may also have multiple agents.

Multiple reputations

It is logical to keep different reputations for one user on every message topic (or otherwise defined message cluster). Your political opinions might have earned you a great reputation, but it doesn't mean your ideas on nanotechnology have as much value. However, you may be allowed to exchange your reputation credits earned in one area into credits earned by you - or anybody else - in another area, using the current 'market' exchange rate. We could expect that credits earned in related areas will be more or less easily transferable. The system manager may tax such transactions to discourage loss of reputations' relevance (though there are strong reasons not to do that).

More on capital control

The system provides the financial control of its transactions, which may include:

initial reputation 'capital' distribution. A certain sum can be given or lent to every new user.

additional emissions of currency;

control over capital transfer / exchange operations;

fiscal policy: certain operations, like exchanges, may be taxed; also, one may tax users' reputations to make sure they reflect users' current positions. Dormant users' reputations may be taxed at a higher rate. Active posters/raters update their reputations frequently enough to keep them current and may be made tax-exempt. - the taxation issues easily come to mind, but it seems that all their goals may be easier and more efficiently achieved without them.

trading the credits for 'real' money. The money earned can be spent to pay for the system hardware and software development. Also, an exchange may allow good posters / raters to earn their living for the work they do. The negative effects of the rich but incompetent users on the quality of other user's reading will be minimal, since their low-rated messages will be discarded by the smart agents.

Other financial opportunities

Certain standard payments (e.g., for writing a message) may get distri- buted to other users proportionally to their capital. This (though in contradiction with the idea of taxation aimed at preserving reputations' relevance) would give users incentives to work hard at the beginning of the system's existence, to post their best articles or compile FAQs as early as possible, etc., and also will make everybody interested in the expansion of the system. Also, this policy makes financial investments attractive- if you think the list is going to be popular, buy the credits in advance! They will pay for the current system development-and you will return the investment later, even if you don't have any personal interest in the discussion.

Other benefits of the system

(some of these have already been achieved by the DF agent).

Users will have good feedback from the readers;

Nobody will have to be 'officially' excluded from the system for sending stupid, irrelevant or impolite messages. Their messages will just become invisible to everybody who does not like them. Thus we replace authoritarianism and coercion with ANARCHY and SPONTANEOUS ORDER. (still, some cases, like credit fraud, breach of privacy, etc., may require interference of the system manager).

The rating system will drastically reduce message traffic. In the current systems,after you choose the presentation of information (e.g. plain texts, or color GIFs), and start spending all the time you have for reading your mail, all further increase in traffic consists of the messages YOU DO NOT WANT TO READ, and the signal/noise ratio tends to 0 together with your satisfaction. The rating system would allow you to keep this ratio very high - and improving.
You don't have to read and rate hundreds of messages to make your agent barely usable. Just get (buy/rent/borrow) a few agent parts from people whose interests and opinions you have reasons to respect, put them together - and start having fun! Later, you will get to know the system better,and learn how to earn more credits and cheaply get messages that would bring you maximal satisfaction for the time and effort you are ready to put into it.
There is no limit to the number of users or postings in the system. System growth may only translate into better selection of everyone's reading, not more noise.
The system may help people with similar interests find each other; the users then can query the publicly accessible parts of others' profiles and attempt to contact each other in order to form a new discussion group, local baseball team, business or family.

Who will benefit most from such a system?

First, who won't:

Established scientific and other professional disciplines are already sufficiently supported by established media; business is interested more in keeping the information to itself rather than spreading it; government's reports will never compete with each other on the market, casual chats or alt.sex.lets.do.it.now ads hardly require a sophisticated system for rating the conceptual depth of the postings...

Now, we are left with non-formalized, volatile communities that would like to increase the quality of their information exchanges and are ready to share their opinions with anybody who may be interested in them, and for whom the relevance and quality of the information and expertise of the authors are highly important.

Looks like most, if not all, people and groups I like fall into this category.

Who can put such a system together?

People who have:

interest in using such a system (see above);
current access to the Internet;
experience in writing software;
respect to free market communities;
some idea of digital cash;
understanding of copyright and privacy issues;
sufficient vision in information technology development;
a good share of adventurism.

If you are aware of any group of such people, please let me know! ;-)

Some general thoughts on messaging systems

Currently available messaging systems, consisting of segmented flow of messages and a non-rating reader, have a natural limit (lets consider only discussion streams, as opposed to, say, .forsale kinds) in segment size, after which the segments become unbearably cumbersome and have to either kick off and/or suppress their own participants or break up to survive. This process has already destroyed many virtual communities. The solution, in my opinion, lies in both better categorizing of information (this has a great variety of well developed methods) and assessing quality of individual items (the task, of which AI will not be capable for at least another [human] generation).

Another bad consequence of this situation is that the topic cyberspace becomes increasingly fragmented which puts smaller and smaller meme populations into reproductive isolation and inhibits cross-fertilization of ideas.

Thus, while the Net does a great job in carrying information over geographic boundaries, it falls far behind any old-fashioned off-line conversation in its ability to cross borders in concept space.

So, the rating system will not only drastically increase the quality and diversity of your reading, protect your interests, reward your efforts and save system bandwidth.

It is a revolutionary engine that may transform the current cyberspace feudalism into a market economy and bring the fracturing global body of knowledge into shape.

I could hardly think of any project of greater importance.

Implementation plan: modest -> daring -> SciFi

Stage I ~1993)

Start with one mailing list. We may base an incoming message's rating on the reputation of its author on this topic. Individual message ratings would require delaying of messages, or a request from each user on logon to send the ratings for the posts s/he has received; then a modified reader can sort the messages according to their ratings. Also, ratings update can be sent periodically to all participants (or appended to every message they receive). Even using an unmodified reader, and receiving all the messages (say, 500), the users will get the list of those few that are best for their interests, read it, and take actions manually.

Such a variant may be practically implemented with relatively little work, and may bring limited, but immediate benefits to most of the list members.

Stage II ~1997)

Global information service with more or less fully implemented encryption and archiving features, elements of hypertext and digital 'reputation' cash. Smart people get rich, curious people learn. Fools stay off-line.

Stage III ~2029)

Brain-resident terabit modem with VR interface suggests you the best of global multisensia according to your current interests. Which means the ultimate freedom of choice. Plus global integration of consciousness. Both of which help people prepare for the Singularity which immediately follows, unnoticeable for non-participants.

January 23, 1993.