As expected, following the recent leak, Valve has officially introduced a system for determining the usefulness of user reviews on the Steam digital distribution service.
Previously, reviews were sorted by the number of helpful notes from other players, but, as Valve notes, many began to publish reviews with ASCII graphics, jokes, memes and other content that was not useful for potential buyers.
Valve emphasized that it sees nothing wrong with uninformative reviews (and therefore will not delete them), because such content often amuses existing players, but it does not always help new consumers.
In this regard, Valve is introducing a new system for determining the usefulness of reviews when they are displayed on a product page: now, when sorting reviews in the store, those that best help players make a purchasing decision are shown first.
The system is enabled by default (can be disabled in the review settings of each game) and does not affect the rating of product reviews – it simply changes the order in which they are displayed on the store page.
Valve uses several methods to categorize reviews, including user complaints, evaluation of reviews by Steam’s moderation team, and a number of machine learning algorithms (to scale human decisions).
Valve notes that the categorization is not yet complete – the system is currently undergoing open testing. To date, Steam users have published more than 140 million reviews.