Star ratings have been widely proven to be a total failure. Most people treat them as binary: either something is nice, and gets 5, or is crap, and gets 1. The neutral score for things that are OK tends to be 5, sometimes 4.
This is particularly pronounced whenever there are any consequences attached to the ratings. If you give anything but 5 to your Uber driver, you're risking them losing their job. If you give anything but 5 on an e-commerce platform, the seller may lose a lot of money. 5 being the "neutral" score became normalized.
That reminds me about how in Germany a reference for a job must be positive. So it ends up with a spectrum of "he/she worked here" vs "he/she was REALLY good" (in German).
I've come to the (sad) conclusion that review systems are inaccurate, even when well intended. There was likely a moment during their existence where they were accurate. I would guesstimate this moment is generally before it gained momentum. E.g. early days of IMDB, early days of Amazon, and early days of tbat German law mentioned earlier (I don't know the name).
Since Metacritic is an aggregation of paid and amateur reviewers alike, it might very well be more accurate.
> Most people treat them as binary: either something is nice, and gets 5, or is crap, and gets 1
I really think GoodReads is the exception to this though. When I look through other people’s ratings I frequently see them using the full spectrum.
Furthermore, it’s very common for written reviews to say that a book really deserves some fractional number of stars, but Goodreads forced them to round.
While I agree with you, it is worth commenting on the idea that a 5 out of 5 Uber ride isn’t quite the same thing as a masterpiece of cinema which someone gives 5 stars to on a platform like Mubi. I always interpreted a perfect Uber ride as clean, safe, no BS. A four star Uber does mean that something was a bit off about the experience whereas a four star movie was still wonderful but not quite earth shattering like A 5 star stone cold classic such as L’Aventura or Battle Royale.
This is particularly pronounced whenever there are any consequences attached to the ratings. If you give anything but 5 to your Uber driver, you're risking them losing their job. If you give anything but 5 on an e-commerce platform, the seller may lose a lot of money. 5 being the "neutral" score became normalized.