How similar you actually are does not predict whether you will like each other
What predicts attraction is not how much two people actually have in common, but how much they feel they do.
Andrew Huberman · Science of Attraction, Compatibility & Romance | Dr. Paul EastwickYou can measure everything about two strangers, score them as an 83 percent match, and still "be no better than a coin flip at" predicting whether they will click.
That is what happens when researchers try to predict attraction from actual similarity, said Paul Eastwick, who studies attraction. When studies pair people on traits measured before they meet, without letting them spin the results, the predictions come out at chance.
What does predict liking is perceived similarity, a loose feeling of shared values and easy connection. It is untethered to any particular trait, so people get wide latitude for motivated reasoning.
When they like someone, they find the similarities. If both love Japanese cinema or share the same politics, that becomes what really matters. If they disagree about politics, they decide that issue just does not matter much to them.
That is also why matching on stated preferences can make dating apps feel useless. Even for demographics and socioeconomic status, where similarity effects look most promising, the improvement is tiny, lifting the odds only to about 53 to 47.
Questions people ask
What exactly is perceived similarity?
Perceived similarity is a general, free-floating sense that you have a lot in common and see the world the same way. It is not tied to any one attitude or preference, which is why people can attach it to whatever fits the relationship.
Does sharing the same politics matter for couples?
Yes, but only slightly. People with very different political views are much less likely to pair up in the first place. Among couples that do form despite a mismatch, the difference predicts satisfaction much less than people expect, because partners tend to downplay it.
Why do dating apps feel so hit or miss?
They often match people on stated preferences and measurable traits, which is actual similarity. Since that predicts initial liking at about chance levels, two high-match profiles may feel nothing in person, while two low-match profiles may click.