Paul Resnick

GroupLens: an open architecture for collaborative filtering of netnews

Collaborative filters help people make choices based on the opinions of other people. GroupLens is a system for collaborative filtering of netnews, to help people find articles they will like in the huge stream of available articles. News reader clients display predicted scores and make it easy for users to rate articles after they read them. Rating servers, called Better Bit Bureaus, gather and disseminate the ratings. The rating servers predict scores based on the heuristic that people who agreed in the past will probably agree again. Users can protect their privacy by entering ratings under pseudonyms, without reducing the effectiveness of the score prediction. The entire architecture is open: alternative software for news clients and Better Bit Bureaus can be developed independently and can interoperate with the components we have developed.

Recommender systems

Recommender systems assist and augment this natural social process. In a typical recommender system people provide recommendations as inputs, which the system then aggregates and directs to appropriate recipients. In some cases the primary transformation is in the aggregation; in others the system’s value lies in its ability to make good matches between the recommenders and those seeking recommendations. The developers of the first recommender system, Tapestry [1], coined the phrase “collaborative filtering” and several others have adopted it. We prefer the more general term “recommender system” for two reasons. First, recommenders may not explictly collaborate with recipients, who may be unknown to each other. Second, recommendations may suggest particularly interesting items, in addition to indicating those that should be filtered out. This special section includes descriptions of five recommender systems. A sixth article analyzes incentives for provision of recommendations. Figure 1 places the systems in a technical design space defined by five dimensions. First, the contents of an evaluation can be anything from a single bit (recommended or not) to unstructured textual annotations. Second, recommendations may be entered explicitly, but several systems gather implicit evaluations: GroupLens monitors users’ reading times; PHOAKS mines Usenet articles for mentions of URLs; and Siteseer mines personal bookmark lists. Third, recommendations may be anonymous, tagged with the source’s identity, or tagged with a pseudonym. The fourth dimension, and one of the richest areas for exploration, is how to aggregate evaluations. GroupLens, PHOAKS, and Siteseer employ variants on weighted voting. Fab takes that one step further to combine evaluations with content analysis. ReferralWeb combines suggested links between people to form longer referral chains. Finally, the (perhaps aggregated) evaluations may be used in several ways: negative recommendations may be filtered out, the items may be sorted according to numeric evaluations, or evaluations may accompany items in a display. Figures 2 and 3 identify dimensions of the domain space: The kinds of items being recommended and the people among whom evaluations are shared. Consider, first, the domain of items. The sheer volume is an important variable: Detailed textual reviews of restaurants or movies may be practical, but applying the same approach to thousands of daily Netnews messages would not. Ephemeral media such as netnews (most news servers throw away articles after one or two weeks) place a premium on gathering and distributing evaluations quickly, while evaluations for 19th century books can be gathered at a more leisurely pace. The last dimension describes the cost structure of choices people make about the items. Is it very costly to miss IT IS OFTEN NECESSARY TO MAKE CHOICES WITHOUT SUFFICIENT personal experience of the alternatives. In everyday life, we rely on

Reputation systems

Systems T he Internet offers vast new opportunities to interact with total strangers. These interactions can be fun, informative, even profitable. But they also involve risk. Is the advice of a self-proclaimed expert at expertcentral.com reliable? Will an unknown dotcom site or eBay seller ship items promptly with appropriate packaging? Will the product be the same one described online? Prior to the Internet, such questions were answered, in part, through personal and corporate reputations. Vendors provided references, Better Business Bureaus tallied complaints, and past personal experience and person-to-person gossip told you on whom you could rely and on whom you could not. Participants’ standing in their communities, including their roles in church and civic organizations, served as a valuable hostage. Internet services operate on a vastly larger scale

Trust among strangers in internet transactions: Empirical analysis of eBay' s reputation system

One of the earliest and best known Internet reputation systems is run by eBay, which gathers comments from buyers and sellers about each other after each transaction. Examination of a large data set from 1999 reveals several interesting features. First, despite incentives to free ride, feedback was provided more than half the time. Second, well beyond reasonable expectation, it was almost always positive. Third, reputation profiles were predictive of future performance, though eBays net feedback statistic is far from the best predictor available. Fourth, there was a high correlation between buyer and seller feedback, suggesting that the players reciprocate and retaliate.

The Value of Reputation on eBay: A Controlled Experiment

We conducted the first randomized controlled field experiment of an Internet reputation mechanism. A high-reputation, established eBay dealer sold matched pairs of lots—batches of vintage postcards—under his regular identity and under new seller identities (also operated by him). As predicted, the established identity fared better. The difference in buyers’ willingness-to-pay was 8.1% of the selling price. A subsidiary experiment followed the same format, but compared sales by relatively new sellers with and without negative feedback. Surprisingly, one or two negative feedbacks for our new sellers did not affect buyers’ willingness-to-pay. Copyright Economic Science Association 2006

Using Social Psychology to Motivate Contributions to Online Communities

Under-contribution is a problem for many online communities. Social psychology theories of social loafing and goal-setting can lead to mid-level design goals to address this problem. We tested design principles derived from these theories in four field experiments involving members of an online movie recommender community. In each of the experiments participated were given different explanations for the value of their contributions. As predicted by theory, individuals contributed when they were reminded of their uniqueness and when they were given specific and challenging goals. However, other predictions were disconfirmed. For example, in one experiment, participants given group goals contributed more than those given individual goals. The article ends with suggestions and challenges for mining design implications from social science theories.

REFEREE: trust management for Web applications

Abstract Digital signatures provide a mechanism for guaranteeing integrity and authenticity of Web content but not more general notions of security or trust. Web-aware applications must permit users to state clearly their own security policies and, of course, must provide the cryptographic tools for manipulating digital signatures. This paper describes the REFEREE trust management system for Web applications; REFEREE provides both a general policy-evaluation mechanism for Web clients and servers and a language for specifying trust policies. REFEREE places all trust decisions under explicit policy control; in the REFEREE model, every action, including evaluation of compliance with policy, happens under the control of some policy. That is, REFEREE is a system for writing policies about policies, as well as policies about cryptographic keys, PICS label bureaus, certification authorities, trust delegation, or anything else. In this paper, we flesh out the need for trust management in Web applications, explain the design philosophy of the REFEREE trust management system, and describe a prototype implementation of REFEREE.

Eliciting Informative Feedback: The Peer-Prediction Method

Many recommendation and decision processes depend on eliciting evaluations of opportunities, products, and vendors. A scoring system is devised that induces honest reporting of feedback. Each rater merely reports a signal, and the system applies proper scoring rules to the implied posterior beliefs about another raters report. Honest reporting proves to be a Nash equilibrium. The scoring schemes can be scaled to induce appropriate effort by raters and can be extended to handle sequential interaction and continuous signals. We also address a number of practical implementation issues that arise in settings such as academic reviewing and online recommender and reputation systems.

Slash(dot) and burn: distributed moderation in a large online conversation space

Can a system of distributed moderation quickly and consistently separate high and low quality comments in an online conversation? Analysis of the site Slashdot.org suggests that the answer is a qualified yes, but that important challenges remain for designers of such systems. Thousands of users act as moderators. Final scores for comments are reasonably dispersed and the community generally agrees that moderations are fair. On the other hand, much of a conversation can pass before the best and worst comments are identified. Of those moderations that were judged unfair, only about half were subsequently counterbalanced by a moderation in the other direction. And comments with low scores, not at top-level, or posted late in a conversation were more likely to be overlooked by moderators.

PICS: Internet access controls without censorship

87 W ith its recent explosive growth, the Internet now faces a problem inherent in all media that serve diverse audiences: Not all materials are appropriate for every audience. Societies have tailored their responses to the characteristics of the various media [1, 3]. In most countries, there are more restrictions on broadcasting than on the distribution of printed materials. Any rules about distribution, however, will be too restrictive from some perspectives yet not restrictive enough from others. We can do better. We can meet diverse needs by controlling reception rather than distribution. In the TV industry, this realization has led to the V-chip, a system for blocking reception based on labels embedded in the broadcast stream. On the Internet, we can do better still, with richer labels that reflect diverse viewpoints and more flexible selection criteria. The Platform for Internet Content Selection (PICS 1) establishes Internet conventions for label formats and distribution methods while dictating neither a labeling vocabulary nor who should pay attention to which labels. It is analogous to specifying where on a package a label should appear and in what font size it should be printed without specifying what it should say. The PICS conventions have caught on quickly. and other software vendors announced PICS-compatible products. AOL, AT&T WorldNet, CompuServe, MSN and Prodigy provide free blocking software that will be PICS-compliant by the end of 1996. RSACi and SafeSurf are offering their particular labeling vocabularies through online servers that produce PICS-formatted labels. A labeling infrastructure for the Internet offers a flexible means of content selection and viewing. Without Censorship 1 PICS is an effort of the WorldWide Web Consortium at MITs Laboratory for Computer Science, drawing on the resources of a broad cross-section of the industry. Project history, a long list of supporting organizations, and details of the specifications may be found at http://w3.org/PICS.