A few thoughts:
1. Five star rating systems by themselves contain such little information that they are basically pointless. If there's a critical mass of ratings, then they can be somewhat useful. Yelp is a 5 star system, but without the ability to skim a handful of reviews written by other users, it probably wouldn't be nearly as useful.
2. The system presented in this post might work ok for one's personal ecosystem, but as soon as I strayed into a new city where my contacts were 0 to a few, the utility of this service would rapidly decline.
3. For all their pitfalls, Yelp and Google reviews do a decent enough job at enabling discovery, especially when users take into account the amount of reviews for an establishment, the specific area they're searching in, and the specific search term they've entered. Sure, there are fake reviews, astroturfers, and gaming going on, but it's not too hard for a trained eye to filter these things out.
But a five star rating system that aggregates the opinions of my friends and colleagues (i.e. people whose tastes tend to be positively correlated with my own) is much more useful to me than a system that aggregates ratings from some general, random or self-selecting population.
If I travel to a city where I don't know anyone, it's true that I initially won't have any ratings (from this particular service), but as I prefer not to dine alone, it's only a matter of time before I make some friends or meet some people and I have ratings again.
My experience of Yelp ratings and reviews is similar to kansface's experience - little correlation with my own tastes except at the far low end. I also don't know many friends or colleagues who rate or review places on Yelp, but I think many of us would start rating places if we knew it would be helpful to our friends, immediately accessible to them, and trivial to do.
But a five star rating system that aggregates the opinions of my friends and colleagues (i.e. people whose tastes tend to be positively correlated with my own)
You don't befriend people who have different tastes or opinions than you? Most of my friends don't even like the same food as each other. Some use mac, some use Windows, some use Linux. How sure are you that those ratings would be more useful than those of the general population?
Also, didn't facebook try this out at one point? Facebook would seem to me to be a perfect platform for this type of service.
The main reason I have (at least some) confidence in my friends' ratings being more useful than those of the general population is that the system approximates what my friends and I currently tend to do, which is go to restaurants recommended to us by our friends.
We've got tastes and opinions in spades, but when it comes to picking a restaurant, we're really not that different - our different opinions are discussed over a dining experience we all want to try and/or enjoy.
If we didn't have a tendency to like the same dining experiences, we probably wouldn't be eating together in the first place. The 'smart/group average' example also seems to be a good way of avoiding or minimising that kind of clash if you don't already know the person.
Most of my friends don't even like the same food as each other.
I'm intrigued by this. I can't think of any particular foods that my friends would say they don't like or don't eat - except perhaps obviously unhealthy options with no redeeming features (eg most fast food). I suppose some people might have a very specific pet hate for something, but usually that results in that person not ordering that thing rather than the whole group going elsewhere.
Australia has quite a multicultural society with a rich variety of cuisines, so that may be part of the explanation. Upon reflection, I'm not confident that our approach would always be accepted by friends not from here.
Maybe this is all just an example of the hyperlocal being more useful to an individual than the general. If Facebook implemented it well, and didn't co-opt each of my ratings into some kind of advertised endorsement, perhaps it could work there too.
This system could always fall back to the more general populace review if there isn't enough data from your friends/acquaintances.
I agree a 5-star rating system tells you little. But it's at least some signal and combine that with also displaying how many people have contributed ratings (which would be trivial in these mockup examples) and you now have a pretty strong signal as to the quality of the place without ever touching the content of the reviews. This is a sample of one of course, but of the probably hundreds of restaurants I've been to with 4-star Yelp ratings that have over 20 reviews, I don't think I've been to even one where I wouldn't agree that yes - this deserves at least a 4-star rating.
I don't think Yelp reviews have any correlation with my own (except perhaps at the far low end). I'll go even further and say I enjoy new restaurants far less when I read reviews first. Everyone has a different deal breaker- healthiness, tastiness, wait time, location, noise, price, quantity, service, smoking, kids, cleanliness, etc. Projecting all of those attributes into a single axis is simplistic- too much is lost. The reviews are invariably a litany of every single time the restaurant fucked up with sprinkles of "pretty ok". I would prefer a better system (if not a different company).
