In 2018, psychologists from Yale, Harvard, Essex and Cornell ran an experiment that should have ended overthinking forever. They put strangers in a room, made them talk, split them up, and asked each person two questions. How much did you like them? And how much do you think they liked you? The answers did not match, and they did not mismatch randomly. Across every version of the study, people rated themselves as less liked than they actually were. Strangers in labs. Adults at personal development workshops. College roommates, who kept underestimating how much the other person liked them for an entire academic year.
Sit with that last one for a second. People lived together for months, shared a bathroom, saw each other at their worst, and still walked around convinced the other person was lukewarm on them. Now apply that to the person you talked to for forty minutes at a party and decided afterward you came on too strong. Your read on how that went is not data. It is a bias, and it has a published name: the liking gap.
You are not bad at reading people. You are bad at reading people's reactions to YOU. Those are two different skills, and the second one is broken in a known, predictable, measurable direction: downward.
Why your brain lowballs you
Boothby's team traced the gap to where your attention goes mid-conversation. The other person is experiencing your jokes, your questions, your warmth. You are experiencing your own internal commentary: the thing you phrased weird, the pause that felt eternal, the story you immediately regretted. Afterward, they remember the vibe. You remember the blooper reel. Then you both go home and you alone conclude it went badly. And here is the cruelest detail in the whole paper: the shyer the person, the wider their gap. The people who most need to know they are liked are the ones whose estimate is most wrong.
Before you decide this only applies in person: researchers extended the liking gap to online conversations in 2024 and found the same thing. People like you more than you think in the DMs too. The bias travels.
The other half: your shot is invisible
Now for the truly unfair part. While you are underestimating how much they like you, you are simultaneously overestimating how obvious your interest is. Psychologist Jacquie Vorauer named this signal amplification bias: people who fear rejection believe their overtures broadcast far more interest than anyone can actually detect. You think you have been flirting for three weeks. From the outside you have been, at best, friendly. Vorauer's paper calls these invisible overtures, which might be the most accurate term in all of psychology. Shots you technically took that never registered on anyone's radar.
Put the two biases together and you get modern dating's silent collision: two people who like each other, each convinced the other is not interested, each certain their own interest is embarrassingly obvious. Nobody moves. Everybody goes home and spirals.
The recalibration protocol
You cannot logic a bias out of your head. What you can do is build correction into your decisions, the way pilots trust instruments over their inner ear. Five instruments:
Whatever you scored the interaction on the way home, your negativity bias already deducted points that were never lost. If your gut says it went 5 out of 10, the research says the real number sat closer to 7. Act on the corrected score, not the felt one.
Write down the three moments you are cringing about. Now answer honestly: did THEY visibly react to any of them, or only you? In the liking gap studies, partners rarely even noticed the moments people ruminated over. No external reaction means the moment happened in your head only.
Signal amplification bias means your hints arrive at roughly a third of the volume you sent them at. Stop stacking hints. Send one unambiguous signal: a direct compliment, a concrete ask with a day attached, a stated reason you want to see them. Clarity is the only thing that reliably survives the perception filter.
If it went well, say so within two days, with a specific callback to something they said. Speed reads as real interest. You are not chasing. You are compensating for a documented transmission failure.
When it feels mutual but frozen, remember the Hinge numbers. Both sides wanting more depth while both sides wait is the default state of modern dating. The person who breaks the deadlock is not the desperate one. They are just the only one who read the literature.
TEN HINTS VS ONE SIGNAL
The correction is not that everyone secretly loves you. Some people are genuinely not interested, and one clear signal will surface that fast, which is also a win. The correction is that your default read sits below reality, so most of the borderline cases you wrote off were never actually borderline. Rejection exists. You have just been pricing it at triple its market rate.
Here is the way your PhD friend would run the math. The downside of a clear signal that misses: one awkward week, and the studies on rejection show you overestimate both how bad it feels and how long it lasts. The downside of an invisible overture that was never received: the entire relationship that would have happened. You are not protecting yourself by hinting. You are just losing slowly, in private.
“People systematically underestimated how much their conversation partners liked them and enjoyed their company.
— Boothby, Cooney, Sandstrom and Clark, Psychological Science (2018)
Still not sure if the interest is real or you are reading vibes that are not there? Upload the chat. Delulu Check scores initiative, reciprocity and warmth from the actual messages, so you can stop guessing what they feel and read what they do.
Run the Interest Check
LIKED THIS?
Get the next one in your inbox
Stop overthinking. Get real answers about your relationship.
TRY DELULU CHECK