I wasn’t thinking of any particular paper there, but if you look at some papers you’ll usually find the following reasons: 1. The app algorithms are usually bl

I wasn’t thinking of any particular paper there, but if you look at some papers you’ll usually find the following reasons: 1. The app algorithms are usually black boxes, so who knows whether they’re really using scientific methods 2. Apps often include questions such as height due to customer demand, which are known to be very bad predictors 3. The interaction between the factors is very significant and complicated/unknown 4. People don’t answer truthfully 5. People don’t know what they want What is known however is that there’s a strong placebo effect initially, i.e. the mere suggestion that somebody is a good match although they aren’t increases the success of first dates, but not in the long term. There’s also research on how these systems can be designed to be addicting, which is surely being done. One of the strongest known predictors instead is “shared emotional experiences”, e.g. a study about making people walk over a wobbly bridge together. Hence why “running groups” and so on are popular nowadays among people disillusioned by algorithms