Statistically it might make sense to segregate and discriminate groups of people. But morally?

It's like how the public wouldn't really accept that if statistics show that all planes are blown up by people from country X, everyone from country X is thoroughly searched and no one else.

Also note that in case of the dating app, if you assume people like X want to date with people like Y and you don't allow them to match anyone else you, you can never validate/improve your assumptions.

Well, depending on how the service marketed itself, I don't think there's anything wrong with doing that. If people want to use an internet dating site to find someone they are happy with, and if they are more likely to be happy with someone similar to them, then you are providing what they asked for. If they find they aren't happy with similar people, I'm sure there are other dating services which deliver that specifically.

It's not our job and it's not right to force our values on others because we think it will make them better people when that's not what they asked for. That said, I think it would be different if a news organization was doing the equivalent, and serving items that confirmed prior ideas. the difference there is that in one case they are asking to be happy, and in the other they are asking to be informed. This does of course assume people actually want to be informed and news isn't just an entertainment medium for the customer as well now...

This was my line of reasoning as well. You can't tailor an entire business around the 1% of cases where a sheep herder in Asia and a multimillionaire from Texas really hit it off.

In this example it sounds like the app was making assumptions and pre-segregating matches based on implied attributes rather than specified preferences.

So even though a user might only say "show me matches between 18-35 years old" the app would apply additional filtering based on the user's supplied attributes in an effort to improve "relevance".

I'm not saying you're wrong, or that this was wrong/right but I did want to clarify what the OP seemed to be saying.

It makes business/economical sense, and some of the algorythm's rules were logical (ie : have people of same country/language meet), but I think the way we discriminated on assumed income and race to be a bit more questionable.

Another thing is that they used to bought ads targeting female users in developing countries to "feed them" to users of richer countries hoping to bait them to pay money (to send digital gifts for example).

The connection criteria should go deeper than the list the OP mentioned. Real connections happen when your principles, beliefs, goals and dreams resonate with another. Your skin colour, location etc. can be completely irrelevant in that equation.

Were the OP working on an algorithm that would go 'next level' and extract those qualities from the base data given, then I am sure that would have been a far more rewarding endeavour than just matching skin colour or income level.

(Note: I am of sub-continental heritage and married to an Australian girl whom I met while going to ballroom dancing lessons).

I'm not saying its not possible, but the list of aggravating factors and even the definition of romance or love would all fall into the equation and quickly turn the service worthless. If a significant user base can find what they need with the minimal and most topical filters it wouldn't make sense to dilute it any further.

Let's say you're an Indian guy living in Midwest America.

The algorithm will put you in the bottom pile, because your race is different than the women in your local area

Or get hounded by insider gold diggers if they find out you're wealthy.