Easier Bootstrapping of Your Own Views
A reader who prefers to remain anonymous writes:
It occurs to me as I read your column that my own collection of bookmarks is my vote for most useful and highest quality sites. I'd take issue with the "neutrality" of not voting for or against a site but I respect your point about bandwidth.
I would suspect some visitors would tend to pad the voting for their "favourites" and trash the sites they think are not worthy or respect. Another more subtle problem is in the changing nature of the estimation. Some days I get cranky and nothing I see impressed or helped me. Other days a simple valentine page with a clever implementation of FLASH animation might make me ecstatic. The sober second thought which comes from a second or third visit to a site is where the real "vote" should appear. Perhaps an automation system which tracks repeat visitors would be an alternative to the vote button.
Pattie Maes from the MIT Media Lab sent a similar comment a few days later regarding the use of bookmarks as an easy way to bootstrap a reputation manager's understanding of what sites you like. She also mentioned that my.yahoo.com has been using this exact technique for more than a year: you tell Yahoo! your bookmarks and it directs you to other sites you may like.
Jakob's comments: I completely agree that bookmarks are a great source of data about what sites a user likes. In general, it is a good design principle for Internet services to find ways to reuse existing information and avoid the need for users to engage in extended dialogues and data entry. Web users are extremely impatient and do not like to spend any more time than absolutely necessary.. With respect to "padding" of votes, it would be a simple matter to place a ceiling on the amount of influence or number of votes that are recorded from any individual user. Punishing a site you don't like is actually OK since that will prevent other people with the same taste as yourself from wasting time on that site.. Users who tend to have different preferences than you will be very weakly correlated with your votes and will therefore not have their ratings impacted by your "punishment" vote. In other words, these other users may still get positive recommendations for the site in question (if other people who share taste have liked the site).
Survival of the Biggest?
Clay Niemann writes:
First let me say I enjoy your columns. We tried to keep them in mind when it was time to redesign our company website ( www.tmf.org ).
The point of this message is about the hypothetical Reputation Manager in the Feb. 8 column. Wouldn't it only work for sites that already have traffic going to them? There could be gold mines of valuable content, but if traffic doesn't start somewhere, no one will ever see them. Meanwhile a few sites get some traffic, the higher the traffic the higher the rating, and again the higher the traffic. We would get the same effect present in popular radio: The same 10 songs everyone is sick of played over and over again for weeks at a time. All because the numbers say they have the highest rating.
Also, there is the possibility of a near oligopoly of sites that would judge content, and would naturally have the highest reputation rating (they have the most detailed stats and an intimate knowledge of the rating system--this would give them an obvious advantage). The effects of this can already been seen by the high priced real estate on major search engines and other major entry points. Those entry points are valuable and will only become more so in the future.
Jakob's comments: A site's rating in the reputation manager would not be a simple sum of the number of users voting for the site. The rating would be the proportion of voting users who liked a site, with each user's vote weighed by the extent to which that user has the same taste as you. As an example, consider two sites, A and B:
- Site A has received "thumbs up" votes from 9,000 users and 1,000 "thumbs down" votes
- Site B has received 900 "thumbs up" and 100 "thumbs down" votes
In this simplified example, sites A and B both have a 90% approval rating and would receive the same reputation rating. It might be reasonable to allow the number of users to have some impact on the reputation rating, so A could end up with a somewhat higher score than B, but it certainly wouldn't have ten times as high a rating even though it has ten times as high traffic.
A further refinement is that the votes are weighed. Thus, if Site B tends to get visited by people who have similar tastes as you, then the reputation manager would send you a higher rating for Site B than for Site A (assuming that Site A is visited by average users most of whom are not particularly close to your individual preferences).
It is true that there has to be some way for new sites to get discovered in the first place, so that they can get their initial votes entered into the reputation manager. Most likely, this discovery of new sites will proceed in traditional ways: recommendation links from established sites, users telling each other about the new site by email, and advertising and other promotional campaigns.
Finally, the possible oligopoly of reputation managers is an argument for these services to be independent companies that are funded through micropayments from the users. It is true that it would be hard to trust recommendations from a service that was run by a company that also owned substantial content sites, though it should be possible to guarantee the integrity of the reputation manager through a "church-state" firewall similar to that maintained between editorial and advertising in good newspapers.