That's the question raised by a new paper reporting that positive online ratings can be strongly influenced by favorable ratings that have come before. The result is a sizable "herding" effect, underscoring the well-known positive bias present in many online ratings sites. But negative comments didn't show the same effect. (click below to read more)
Researchers at the business schools of Hebrew University, the Massachusetts Institute of Technology and New York University, working with an unnamed news-sharing site, set out to "quantify the effects of social influence on users' ratings."
So over the course of five months, they randomly assigned a plus-1 rating, a minus-1 rating or no rating at all to 101,281 comments. The ratings, visible to the site's users, were assigned in proportion to the real-world ratio of such ratings, meaning positive heavily outnumbered negative, but most went into the unrated control group.
Comments starting out with an up-vote were 32% more likely to get another up-vote from their first viewer compared with comments in the control group. This boost mostly endured; five months later, comments that launched at plus-1 had an average rating 25% higher than those of the control group.
Such comments were also more likely to rack up especially high overall scores, suggesting a snowball effect.
Yet the same syndrome didn't occur with comments that started out rated minus-1. On the contrary, an initial negative rating generated a "correction effect" leading to more up votes. The result, after five months, was roughly the same outcome for comments starting at minus-1 as for the control group.
"Our findings," the researchers write, "suggest that social influence substantially biases rating dynamics in systems designed to harness collective intelligence."
No comments:
Post a Comment