I see your point, not so sure about the bell curve here though. There is an assumption that the median average experience is equal to arithmetic average experience, which doesn't seem to be rooted in reality. The quality of service is constantly upgraded, hance prevalance of 5 star reviews, if everything is ok, there is no need to lower the score. There is a gratitude.
Owners and service providers and incentivised to provide nice experience. And on the other side of spectrum there are people.whose expectations were not met. And this is a situation when some people decide to leave a review, and only then. Does it make these groups of people full of shit? Even if they might be the same people sometimes?
The game theory here seems to be much more complex than influence of emotions on the system. Sure, some people are terrible at giving feedback, and sometimes, we all have this bad day, when one small detail can make a difference. It seems to be more than simple algorithm.
Which is exactly why a 5 option rating system isn't even appropriate within the context of the given example. Just another way we can see that the rating systems are inherently flawed to the core. For something like a restaurant review it makes way more sense to just have 3 options, as most reviewers are never going to need more precision than that.
This is also something I have talked about at length, and I even came up with a way to make different rating systems compatible with each other using odd-point rankings.. If there are an odd number of options available every ranking can be reduced down to a percentage that meshes will with the other ranking systems. The most common systems being 3-point, 5-point, and 15-point rankings. Where 3 point is pass/neutral/fail, 5 point is 5-stars or A-B-C-D-F, and 15 point is A-B-C-D-F with a +/- option. The maximum option would be 101 point scale where reviews are graded directly with the percentage itself from 0%-100%