Monday, February 16, 2009

Meta-information Markets

It is a general characteristic of information that it is often impossible to know, or even to have a good estimate of, how useful a piece of information will be until one has it.

People do buy books, of course, but it is quite common to feel afterward that the purchase of a particular book was not really worth it, and conversely a great many worthwhile books go unread by people who have no way of knowing in advance how much they would appreciate those books. Similarly with other forms of information.

In this particular post I will discuss what might be called meta-information, by which I mean information which leads to other information which the reader will (hopefully) find useful or otherwise interesting. The information pointed to may be further meta-information, but somewhere at the end of the chain there must be something will value for its own sake.

The problem in its full generality is too complicated for me to discuss here. I will assume that there is some entity which I will call the "publisher" who will be able to obtain some payment from the reader, direct or indirect. I will assume for now that there are no coercive entities that need concern us; the publisher is operating openly and "legally". However, unscrupulous publishers potentially may employ "shills" to promote their goods, and their competitors or others who dislike their ideas for whatever reason may say untrue negative things about their works. My goal is to briefly sketch how some sort of useful "reviewer" system might work.

In my proposed system, the reviewer describes the work using with some sort of standardized system which allows the reader to quickly find works of interest to him. Semantic web type stuff. Either the reviewer is a relatively large organization, or many mutually independent reviewers have agreed to use the same system. The reviewer's primary function is to accurately describe the work and only secondarily to give a subjective assessment of its quality. The publisher pays the reviewer in hopes of receiving a wider readership. The reader has access to reviews for free.

The reason the publisher pays rather than the reader is that the reader is highly uncertain as to the value to himself of the service of the reviewer, whereas the publisher has the relatively simple task of assessing whether the additonal sales due to the review justify the reviewr's fee. Even though the reviewer is paid by the publisher, the reader can have more confidence in the review than he could in advertisements because the reviewer is not merely acting on behalf of the publisher but is using some sort of objective criteria, and the reviewer's value is solely based on his reputation for honesty and accuracy. Yes, reviewers would review each other.

Whatver the details of the reviewing system, provided that all reviews are specific and digitally signed, it seems to me that the system as a whole should be verifiably almost completely "honest". If Alice reviews Bob and says something "I agreed to read six books reviewed by Bob at random, here are the six books and his reviews, I found them completely accurate", this is not sufficient for a reader who trusts Alice to trust Bob 100%, but it's pretty good evidence that he's generally reliable. Provided that it can be kept unambiguous whether or not a review is "correct", one false review could be enough to destroy a reviewer's reputation. In order to establish themselves, reviewers might have to at first review a few books for free and pay to have themselves reviewed by established reviewers.

No comments: