Software boffins at the University of Illinois at Chicago are working on ways to detect fake Amazon reviews. And they’re developing an algorithm for a software bot, sponsored by a Google Faculty Research Award. They try to detect groups of people working in cahoots, by looking for obvious things such as:
* a sudden rash of glowing five-star reviews for a brand new ebook, suggesting the author has rounded up his family and friends to boost the book;
* a sudden clustering of good reviews (all reviews appearing in a four-day window, although I guess this could legitimately be a ‘cascade’ effect arising from a genuine good review);
* and they’re using stylometrics to detect similarities in language across different reviews, which might suggest they were all written by the same person under different names.