Each week we¹ll be looking at a new piece of research that¹s relevant to or affects the reputation space. Today we start with a report first covered on MIT’s blog from Vassilis Kostakos, of Carnegie Mellon University and the University of Madeira.

Increasingly, people are paying more attention to review and comparison websites for an accurate portryal of how well a product or service performs, rather than relying on editorial coverage or advertising.

Sites like eBay – or even the Apple App Store – actively promote this as a key strength of their network (with Amazon even bringing in their top sellers to talk to media), but Kostakos’ research suggests that a small group of highly active users do most of the talking.

Kostakos, presenting his findings at the International Conference on Social Computing last month, stated that ratings systems can paint a distorted picture of a product if only a handful of people do most of the voting (much like the ‘power users’ on Digg).

He went on to add that though ratings systems may tap into the ‘wisdom of the crowd’, by looking at the patterns of users on Amazon, IMDb (the Internet Movie Database) and BookCrossings ( a book review site based in the US), it seems that very few people actually contribute on a regular basis. On Amazon for instance, only 5 percent of active Amazon users cast votes on more than 10 products.

This in turn means that scoring a ‘top 10’ ranking on such a site doesn’t mean that those purchasing or using a product necessarily feel positively about that item, as the voting doesn’t represent the general opinion of the community, just a small snippet.

Most consumers either have a very strong feeling towards something (love or hate), or just don’t care. This means that ratings tend to be either very low or very high, producing an extreme result. Even then, people inherently want to be nice, and so most results are positive.

For many, this comes as no surprize, comment-based systems can be far more useful for findings advocates and for building up a database of customer recommendations. However, sites that integrate a variety of review features (‘thumbs up’, numercial, ‘was this review helpful to you’ etc) are often most reflective of mass public opinion.