I know that this idea is not patentable:
"An app that allows people to find their perfect restaurants to eat in."
It is too general, too abstract:
But what about the implementation to the idea:
"My app will randomly select menu items from various restaurants until the user finds items that they like"
This is a specific implementation of an abstract idea.
So is it patentable?
It is specific enough to not be just an idea but it might not be specific enough to avoid being obvious in light of previous knowledge (prior art). It also needs to be fleshed out in terms of the details of the best way you know to carry it out.
It might be that someone already had the exact invention and either published it or applied for a patent. Then your invention would not be novel. Alternatively, there might be an existing publication/patent/patent application that matches people based on them choosing a favorite flower or fruit. An examiner might argue that since some emoji's represented flowers or fruits your idea is obvious and therefore not patentable.
There have been many patents issued in this area in the past, see this one from eHarmony for example. However, it is getting harder and harder to avoid any computer related claim being labeled as "abstract" and therefore not patentable. An experienced practitioner in that field can be a big help, but no guarantee.
This is a specific implementation of an abstract idea.
So is it patentable?
I'm not a lawyer, but in my opinion, the implementation is not sufficiently specific enough to merit a patent. Even if you get past the issue of novelty and obviousness (which is not a given), you haven't enabled the invention. You haven’t stated how you select the number and type of emoji to present, how the reactions are recorded and most importantly how you measure similarity of reaction. There is a whole field of statistics related to measuring similarity.
Now if you develop a new algorithm to measure the degree of similarity of the different members test results, that algorithm may indeed be patentable. Particularly if you can show it works better than existing algorithms.
