Supporting "Word-of-Mouth" Social Networks through Collaborative Information Filtering

Article excerpt

Altered Vista is an instructional system that supports a form of "contextual" collaborative learning. Its design incorporates an information filtering technique, called collaborative information filtering, which through computational and statistical means, leverages the work of individuals to benefit a group of users. Altered Vista is designed to provide, upon request, personalized recommendations of web sites. It can also provide recommendations of like-minded people, thus setting the stage for future collaboration and communication. An empirical study involving inservice and preservice teachers was conducted using Altered Vista and presents results from an empirical study. The study examined the feasibility and utility of automating the well-known social feature of propagating word-of-mouth opinions within educational settings. It also examined the impact of Altered Vista's ability to recommend a social network of potentially unknown people.

**********

Much has been written about information technology support for intentional, extended, and intensive collaborative learning (e.g., Dillenbourg, 1999; Roschelle & Teasley, 1995). This form of collaboration falls into the category that Bruce (2001) has dubbed conceptual collaboration. Bruce (2001) called another form of collaboration contextual, wherein individual participation is occasional and less intensive. In this case, the pursuit of personal goals by individuals creates incidental by-products, which also contribute to the common good.

This article describes a system, called Altered Vista, which was designed to support such contextual collaboration. Specifically, its design incorporates a recent information filtering technique, called collaborative information filtering which captures an individual's preferences to benefit a group of users (Resnick & Varian, 1997). In its individual, intentional form, Altered Vista solicits ratings and opinions from users about the design, usefulness, and quality of web sites on particular topics. These data become a repository of community knowledge.

In its contextual form, the system can mine the data using collaborative filtering techniques to provide personalized recommendations of web sites to an individual user. Because of this capability, Altered Vista is an example of what is called a "recommender system" (Resnick & Varian, 1997). In addition, because of the underlying collaborative filtering algorithm, Altered Vista can also provide recommendations of like-minded people. Thus, Altered Vista sets the stage for future collaboration and communication.

The next sections describe collaborative information filtering, and discuss its implementation within the Altered Vista system. Results from an empirical study involving mostly inservice and preservice teachers enrolled in classes at two U.S. universities are then reported. In particular, through analyses of user surveys, user comments in an online bulletin board, and system usage, the article reports the feasibility and utility of providing personalized recommendations. Specifically, the ability of the system to support and automate the well-known social feature of propagating word-of-mouth opinions from trusted people is examined.

Second, the article examines the feasibility and utility of Altered Vista's ability to recommend like-minded users. In particular, users' reported interest in people recommendation, related privacy issues, and the broader question of the role of a computer system in suggesting social networks where none previously existed are discussed.

COLLABORATIVE INFORMATION FILTERING SYSTEMS

Within the human-computer interaction (HCI) literature, an approach to categorizing, collecting, and filtering information has emerged, called collaborative information filtering. It is based on propagating word-of-mouth opinions and recommendations from trusted sources about the qualities of particular items (Malone, Grant, Turbak, Brobst, & Cohen, 1987; Maltz & Ehrlich, 1995; Shardanand & Maes, 1995). …