There are many behaviours by which one human shows being sexually attracted to another – staring at them, running fingers through one’s hair, standing close, smiling at them, etc. Most of these are ambiguous, meaning they can be explained away by nonsexual reasons. Staring may be due to being lost in thought and looking absently at a single point, which happens to contain a person. Adjusting the hair could happen because the hair feels messy. One could randomly stand close to someone, smile because one is happy for unrelated reasons and so on.
There are obvious benefits of clear messages – no wasted effort chasing someone not interested, no awkward situations, no false accusations that one’s partner was sending signals of interest to someone else. Why has evolution led to messages of attraction that create doubt in the observers?
If someone’s sexual advances are unsuccessful, this is interpreted as a negative signal about the rejected person and lowers their chances in the future. Rejection makes one wonder what the rejecter knew about their admirer that is unattractive. If a person has characteristics that makes others reject them, the offspring of that person are likely to inherit these and also be unsuccessful in mating. Unsuccessful offspring mean the fitness of the rejected person is low, justifying rejecting them. This evolutionary mechanism is called Fisherian sexual selection. Because of it, nobody wants to be seen to be rejected. One way to hide rejections is to hide the wooing and if rejected, pretend to be uninterested anyway (sour grapes).
Someone attempting to cheat on their partner obviously does not want others to see their advances on another person. People gossip, so hidden signals with plausible deniability are useful.
Some people take advantage of those attracted to them (the advantage may differ for men and women), so it is good to send messages of attraction only to those who are attracted in return. Someone who is interested pays more attention to a person, so is more likely to notice ambiguous messages from them. Wishful thinking makes an interested recipient interpret mixed messages favourably. Of course there is a positive probability of a mistake, but the difference between the probability of interested people versus unintended recipients noticing a signal is greater for ambiguous than clear messages. This is like encryption – there is a positive probability of friendlies having lost the encryption key, but the difference between the probability of friendlies versus hostiles understanding the message is greater for encrypted text.
Dating websites have probably figured this out, because they allow private messages. An additional improvement may be self-destructing messages that can only be viewed once. This makes it harder for the recipient of a message to prove someone’s interest to others and thus lower their admirer’s reputation after rejecting them. Randomly generating messages of attraction and sending them to people would give plausible deniability to those who are rejected. The benefit of deniability must be weighed against the loss to the recipients of false signals.