> Most books are rated between 3 and 4. Its really hard to tell if that rating is accurate

Sometimes I agree with you (because it's annoying to me too), but other times I feel like that's accurate. The difference in quality between a book that's in the 75th percentile quality-wise among the books I've read and a book that's in the 25th percentile is not very large! I'd say, by and large, most books that get published are pretty good. Few of them are completely flawless or life-changing, and that's okay. 3-4 seems about right for 50% of the books I read.

If you combine that basic fact with a range of people with differing tastes, you get even more reversion to the mean, so just about every book has a 3-4 rating.

The same goes with beer. If you check the major beer rating sites, you'll see most beers end up with a 3-4 rating. Personally I think movies have a much wider range in quality, but you still see this effect somewhat with IMDb: a huge proportion of movies is in the 6-8 range.

You might think that the ratings would be more useful if what we got was a percentile rather than an absolute rating, and that might be right... or it might disguise the fact that I really would get close to the same amount of enjoyment out of a 3.5 as a 3.9, even though they're separated by 30 percentiles or whatever.

It's strange to me that this is the pattern for things like books, movies, or beer but with Uber drivers the pattern seems to be "Give 5 stars unless something was wrong."

Before I learned this Uber etiquette I would rate drivers the same way. "Well, he got me from A to B without issue, but was there anything that set this ride apart and elevated it?"

That is due to Uber corporate misunderstanding ratings and penalizing drivers who get anything less than a perfect rating.

One aspect of this problem might be that there's no incentive to have a wider dynamic range in your ratings. A naive recommendation system which just suggests highly rated books similar in some metric to ones I rated highly could be improved enormously if it took into account books I didn't like. In return for accurate ratings, I would get good recommendations.

Yeah, most review sites have the 5 or nothing problem. I look at the distribution and then find some medium to low starred reviews to see why I might not like a book that otherwise seems a good match for me.

In aggregate the star rating is pretty good for literature and non-fiction. For mass fiction it's fairly useless.

But that's a problem with people doing the ratings, not Goodreads. I see this problem with IMDB, trakt, myanimelist, anilist, etc... Because for many people, it's either 1, 4 or 5.

I think that's honestly how most people score. If you read the entire book and it was okay you give it a 3 if it was great a 4 best book ever a 5. If you hated the book you may not finish or review.

The opposite is uber where rating affects future service and puts a lot of power over others in your hand. Not giving 5 is socially unacceptable like not tipping vs tipping less and complaining.

It's all about framing though isn't it? If Uber were '1 (miserably below expectations) .. 3 (met expectations) .. 5 (above and beyond)', you wouldn't see that.

It doesn't really matter as long as everyone's on the same page I suppose. Sure, you clip stuff off the end.. but doesn't everyone just want a taxi that gets them A-B without problem? Who's actively seeking 'above and beyond'? Maybe 'would've scored >5 if I had them' is not a problem, because nobody cares?

(I should maybe say I've never actually used Uber!)