Collaborative filtering systems can support exploratory searching by providing users with information derived from the experiences of previous users. Collaborative filtering is based on the observation that people are good editors and filters for their friends. Currently, when a user reads an article that he thinks will be of interest to a friend, he copies the article and sends it to the friend. Often the friend will not even know that the information contained in the article is available, and so could not have constructed a filter rule or query to extract the information directly. As suggested by Malone, collaborative filtering takes advantage of the social interactions between people to create a filter.
Another advantage which collaborative filtering systems have over other automatic filtering systems is that we can expect human beings to be better at evaluating documents than a computed function. Automatic filtering systems attempt to find articles of interest to their user, often using some scoring function to evaluate features of the documents and returning the documents with the highest scores. People can effortlessly evaluate features of a document that are important to other people, but would be ``AI-complete'' to detect automatically. Examples of such features are the writing style and ``readability'' of a document, or the clarity and forcefulness of an argument the document contains. Imagine the difficulty an automatic filtering system would have figuring out which of two cake recipes is ``easier to follow.''
A further motivation for collaborative filtering comes from the following observation made by Hill et al.[14] The objects we use in everyday life accumulate wear and tear as a normal part of their use: pages in books become wrinkled, bindings creased, and margins smudged with fingerprints. This wear is usually frowned upon as indicating that the object is wearing out and needs to be replaced. Consider however, how helpful are these markings gathered unobtrusively over time: Reference books will open to the most commonly used pages when dropped on a desk. The well thumbed paperback books in a library are the most commonly read ones. The most frequently used recipes in a cook book will have the most stains. The wear on these objects acts as an index to information they contain.
Now compare the rich environment of real objects to the much poorer one in which computer users operate. When a user reads a computer file he usually has no way of telling whether he is the first person to ever read it, or if he is looking at the most commonly used reference on the system. Collaborative filtering works by associating with computer documents the history of their use. Access to this history provides users with the type of subtle hints that we already take advantage of when making read/don't read decisions in the real world.