Management of Reputation Using Collaborative Filtering

by Moya K. Mason

With the overwhelming diversity that can be found on the Internet and the sheer number of websites that people have to choose from, something is needed to help them make choices. People often follow the advice of others when it comes to which sites they visit, the movies they watch, and even the products they buy. They decide whether or not to follow this advice by taking a person's reputation into consideration.

Using and building upon early collaboration filtering techniques, reputation management software gathers ratings for people, companies, and information sources. The reputation ratings can then be used to help steer interested parties toward sites and products that may be of interest. These types of software systems help us overcome information overload. Sites such as Slashdot and other online communities offer product review information, use reputation software to filter out noise by not showing comments made by people with lower reputation ratings. In Reputation Systems: Facilitating Trust in Internet Interactions, Paul Resnick and associates suggest that reputation systems require longevity, visible feedback from present and past transactions, and proper attention given to the track records of all parties. They state that reputation systems "help people decide whom to trust, encourage trustworthy behavior, and deter participation by those who are unskilled or dishonest."

When dealing with strangers in online situations, you don't have the advantage of having a prior relationship with them, which would provide insight and clarification to personalities and morals. You really don't have anything to go on when it comes to deciding whether or not to buy something they are selling, especially if any amount of money is at stake. Trust is hard to come by when two strangers are involved in a monetary transaction, especially when there is not a system in place that rewards good behavior and punishes bad. As Resnick points out, "[i]n some sense, a stranger's good name is not at stake. Given these factors, the temptation to "hit and run" outweighs the incentive to cooperate, since the future casts no shadow."

Reputation management software can create a track record for each user that acts as an incentive for them to exhibit good behavior and make them accountable for their actions. This is important because it adds elements of expectation and possible repercussions that can affect future interactions. "Reputation systems seek to restore the shadow of the future to each transaction by creating an expectation that other people will look back upon it," (Resnick) and attempt to create what "Robert Axelrod refers to as the 'shadow of the future'" (Resnick).

Reputation management software can also track the quality of an individual's credibility, the value of specific information resources, and also the product history of a company in the form of ratings. These ratings can then be used to create a web of trust. Chrysanthos Dellarocas, a researcher at the MIT Sloan Institute of Management, says that:

The production of trust is an important requirement for forming and growing open online trading communities. The lack of a common history with potential trading partners as well as the relative ease with which buyers and sellers can change partners from one transaction to the next, gives incentives to both parties to provide inferior service quality or to hold back on their side of the exchange (Dellarocas).

Dellarocas maintains that trust is a very important component in the virtual world of ecommerce and is a requirement when it comes to buying and selling on the Internet, and that "[t]he goal of reputation systems is to encourage trustworthiness in transactions by using past behavior as a publicly available predictor of likely future behavior" (Dellarocas).

Examples of Reputation Managed Sites

The most famous of the auction sites, eBay, uses reputation management software to rate the reputation/trustworthiness of its buyers and sellers. Parties to the transactions are able to give feedback to each other and affect their ratings, based upon such things as the condition of the auction item, how quickly they shipped it out, how well they communicated, and the speed of payment. The system has an important affect on buying and selling amongst strangers because it has a built-in mechanism for the development of trust. Part of the reputation system includes the Feedback Forum, where buyers and sellers can rate each other and comment on the quality of the interaction. Buyers feel safer when they can read the comments posted about a seller and view their track record, while sellers can find out if their buyers have ever reneged on paying in the past, or if they tend to be tardy when it comes to sending payments to sellers. eBay's reputation management system motivates both buyers and sellers to be honest and play by the rules.

Slashdot has two mechanisms in place to manage the thousands of comments that users post each day. Both are based on reputation. The system attempts to judge the information shared and the sharers of the information, in order to gauge the quality and wisdom of both for the benefit of all. First, it gives users the ability to rate each other's posts, called 'Moderation,' in an attempt to reduce the signal-to-noise ratio. The resulting conversation can then be read in a variety of ways, such as by viewing only posts rated above a certain threshold. Posts are scored from -1 to 5, and a reader can set their viewing threshold to any score within that range. If, for example, the threshold is set to 3, the only comments displayed will be those with scores of 3 or above. Setting the threshold to -1 will display all comments. Zero will give almost all comments, while 1 filters out most anonymous comments. Secondly, the system awards karma points to members, based on how their posts are rated. Slashdot permits the posts of those with accumulated karma to be automatically set to a higher rating, as well as potentially giving them the ability to moderate.

Epinions is an independent company that endeavors to offer unbiased advice about products and services. Epinions uses a reputation management software system that helps to inform consumers before they make purchases, giving them access to personalized recommendations, reviews on a large number of products, and price comparisons. Epinions rates the services and products it reviews, and also rates the people who write the reviews. It makes use of ratings, reviews, and user feedback to create reputations and instils a level of trust for both reviewers and the products they rate. In other words, you can have your own Web of Trust once you begin to trust and once you value the product ratings of particular reviewers, based on their track record and your personal preferences. Users can vote on the quality of a review once they have read one. Epinions places the reviews that are rated the highest on top, which helps reviewers build a good reputation and make money. A micropayment is paid out to a reviewer every time someone reads one of his or her reviews.

Google utilizes a reputation management scheme for its search results. The system, called PageRank, is based upon reputation. The value of a site depends upon how well linked it is from other sites, as well as on how high each of those other sites is ranked.

Problems Associated with Reputation Systems

Given that reputation management is still in its infancy, there are many problems that need to be addressed in order to perfect the inner workings of such a system. Dellarocas has developed three possible methods that can be used in combination to make sure that a few users cannot distort the ratings presented by ecommerce sites.

1. Controlled Anonymity: Anonymity between buyers and sellers allows them to give honest assessments by reducing concerns of reprisal.

2. Median Filtering: Rather than using simple average calculations to discover the mean of all ratings, more complex techniques can be used to find the norm of the ratings. For example, dropping exceptional scores provides a more realistic picture.

3. Frequency Filtering: To preserve the integrity of a reputation management system an average rate of submission is determined. Based on that, the scores from the most frequent reviewers should be removed to avoid a skewing of the results.

Resnick and associates have uncovered three related problems when it comes to eliciting feedback in an online reputation system:

Problems When it Comes to Distributing Feedback:

1. Name changes and the use of pseudonyms affect the system. If there is a possibility that a user can return with a new identity, they can basically erase a previous negative reputation and begin again.

2. The inability to transfer reputations or feedback from one system to another. For example, an eBay member of good standing may want to have a review of their ratings made available on Amazon. Currently, there is no universal framework that can be used across systems.

3. Thirdly, there is a problem with the lack of meaningfulness in the feedback. This is due to the various numerical rating systems being used, which make it impossible to express important nuances using such a system.

Related Papers

Short History of Barcodes
Short History of Collaborative Filtering
Notes on RFID Technology, 2001
Freenet File Sharing
Gnutella File Sharing
Early History and Media Coverage of Webraska
Early History of Napster
Bio of Bernardo Huberman
What is Mojo Nation?

Bibliography

Dellarocas, C. Mechanisms for Coping with Unfair Ratings and Discriminatory Behavior in Online Reputation Reporting Systems. Proceedings of the 21st International Conference on Information Systems (ICIS), Brisbane, Australia, December 2000.

Resnick, Paul, Richard Zeckhauser, Eric Friedman, and Ko Kuwabara. 2000. Reputation Systems: Facilitating Trust in Internet Interactions. Communications of the ACM, Vol. 43, No. 12, pp. 45-48.


Copyright © 2017 Moya K. Mason, All Rights Reserved

Back to: Resume and More Papers