The fact that they are using machine learning algorithms tells us everything we need to know: they are looking for patterns.
The kind of patterns that come with 1-star campaigns, sock puppets, paid reviews, and other attempts to game the system.
Positive reviews or negative reviews won't matter any more than before; they just want honest and fair reviews.
Just remember that amazon is about more than books. And that last year's "world's best whatever" can easily be this year's "overpriced junk" so yes, newer reviews should weigh marginally more than older ones.
I just ran into a perfect example today with my WD TV streaming box; a few months ago it received a firmware update that changed the GUI. Much improved. Today I noticed it enabled MIRACAST support. Two year old product. Hundreds of reviews. A year ago it didn't even mention MIRACAST. Now it not only has it, it works well. The new review is more relevant to the product as it exists so it *should* weigh slightly more.
Anybody remember these?
http://www.amazon.com/Hewlett-Packar...+560+pocket+pc
Great deal?
All they are doing is what Google does constantly; look for bad actors trying to misuse the system and try to cancel them out. Somebody will always scream and grumble but at least they're trying. I don't hear eBay or Rakuten even trying.