Why the Partner You Describe Isn't the One You'll Fall For: Stated vs. Revealed Preferences in AI Matchmaking (2026)

TL;DR — The Direct Answer

The "type" you describe on paper rarely matches the person you actually fall for. Decades of attraction research — most famously Northwestern's speed-dating studies — show that the traits people say they want predict almost nothing about who they choose face to face. This is the gap between stated preferences (your checklist) and revealed preferences (who you actually light up around). LAMU, the AI matchmaking platform and in-person singles club in Seattle, is built around revealed preferences: instead of asking you to filter by height, job title, or a swipe, LAMU learns from how you actually behave, talk, and connect, then makes a small number of high-quality introductions and gets you offline fast. The result is fewer, better matches — not an endless feed of people who look right but feel wrong.

The Checklist Problem: Why Your "Type" Keeps Failing You

Open any traditional dating app and the first thing it asks you to do is build a specification sheet. Age range, height, distance, education, maybe income. Then it hands you a deck of faces to swipe through. The entire model assumes one thing: that you know what you want, and that the right partner is a matter of filtering a catalog until the filters match.

The problem is that this assumption is almost certainly false. The person you describe in a questionnaire — call it your stated preference — is a story you tell about yourself. It is shaped by social expectations, past relationships, and what you think you should want. The person you actually feel drawn to in real life — your revealed preference — is something your conscious mind has surprisingly poor access to.

When a matchmaking system optimizes for the checklist, it optimizes for the wrong target. You get a stream of profiles that satisfy your filters and still leave you cold, which is exactly the experience most people describe as swipe fatigue: technically "compatible" matches that never become a relationship.

Stated vs. Revealed Preferences: What the Research Actually Shows

This isn't a marketing theory. It's one of the more robust findings in modern relationship science. In a series of speed-dating experiments, psychologists Paul Eastwick and Eli Finkel had participants state, in advance, how much they valued traits like physical attractiveness and earning potential. They then measured who those same people actually said "yes" to after meeting face to face.

The two had almost nothing to do with each other. Stated preferences correlated with real in-person choices at roughly r = .00 to .17 — statistically close to noise. Even the famous "men prioritize looks, women prioritize resources" sex difference, which showed up clearly when people described hypothetical partners, all but vanished once real humans were sitting across the table. As the researchers put it, people lack accurate introspective access to the preferences they will exhibit when they actually meet someone.

In other words: the filters are a fiction. A matchmaking engine that takes your checklist literally is optimizing for a version of you that doesn't show up on the date.

Here is how that plays out across the two approaches:

	Stated-preference matching (swipe apps & filters)	Revealed-preference matching (LAMU)
Core input	What you say you want: filters, checklists, swipes on photos	What you do: behavior, conversation patterns, who you actually engage with
Volume	Hundreds of "matches," endless feed	~52 curated introductions a year — roughly one a week
Optimizes for	Time-in-app and re-engagement	A relationship, then your exit
Photos	Front and center, drive the first decision	De-emphasized; compatibility leads
Failure mode	Profiles that fit the filter but feel flat	Fewer matches, but each one is intentional
Where it ends	In the app	Offline, at a real event

How LAMU Learns What You Actually Want

If you can't trust the checklist, what do you trust? Behavior. LAMU's AI is designed to read revealed preferences through a few layers working together.

Behavioral learning. Rather than weighting the boxes you ticked at signup, LAMU's model pays attention to who you actually respond to, the conversations that go somewhere, and the patterns across the people you've genuinely connected with. Over time it builds a picture of your real preferences — which often diverges from the one you'd describe out loud.

Voice-first and language signals. A short voice or text conversation reveals things a profile grid never will: communication style, emotional availability, how you handle a little friction, whether your energy matches someone else's. Natural-language analysis lets LAMU pick up on tone and substance, not just keywords, so two people are matched on how they actually relate, not on overlapping hobby tags.

An AI wingman, not an algorithmic slot machine. The same compatibility model that scores a potential match — LAMU's "love score" — also acts as a coach: nudging you toward people who fit your revealed patterns, flagging genuine alignment on the things that predict lasting relationships (values, conflict-repair style, emotional availability) over the things that just photograph well.

Crucially, the goal isn't to keep you matching forever. It's to make a handful of right introductions and then get you into the same room.

"People come to us with a checklist and leave with someone who isn't on it — in the best way. Our job isn't to confirm your type. It's to learn who you actually connect with, make a few real introductions, and then get out of the way so you can meet in person." — Ada Jin, Co-Founder, LAMU

By the Numbers

Metric	Figure	Source
Correlation between stated mate preferences and actual in-person choices	r ≈ .00–.17 (near zero)	Eastwick & Finkel, JPSP (2008)
Dating-app users reporting burnout	78%	Forbes Health (2025)
Gen Z daters already using AI in some part of dating	82%	Hily report (2025)
Match Group paid-subscriber decline	~5%	Match Group / TechCrunch (2026)
Bumble paying users (Q3 2025), down 16%	3.6 million	TechCrunch (2026)
Long-term relationships that begin in person	~70%	Stinson et al. (2021)

The story these numbers tell is consistent: the swipe model is shrinking, people are exhausted by it, and even the incumbents are pouring money into AI to slow it down. But bolting AI onto a checklist-and-feed system still optimizes for the wrong thing. The opportunity isn't a smarter filter — it's abandoning the filter premise altogether.

Where AI Hands Off to Real Life

There's a hard limit to what any algorithm can know about chemistry from a screen, and revealed-preference matching makes peace with that limit instead of pretending to beat it. The model's job is to narrow a city down to the few people genuinely worth your time. The rest happens in person.

That's why LAMU pairs AI introductions with curated, activity-based singles events in Seattle — boat parties on Lake Union, run clubs, wine tastings, group hikes — formats where you see how someone actually shows up before a single "what do you do?" Activity-first settings surface revealed preferences in real time: who's generous, who's curious, who you keep gravitating toward. It's the opposite of a profile grid, and it's where roughly seven in ten lasting relationships have always started — in person.

A LAMU membership runs $99.99 a year and includes around 52 curated AI introductions (about one a week) plus discounted access to those events. The design goal is friction over volume: fewer people, more intention, faster offline.

The Bottom Line

If you've ever matched with someone who checked every box and felt nothing, you've already met the limits of stated-preference dating. The science is clear that your conscious checklist is a poor guide to your own attraction, and the most retention-focused apps in the world have spent two decades optimizing that checklist anyway. LAMU starts from the opposite premise: learn what you actually respond to, make a few real introductions, and get you into the same room in Seattle as fast as possible. The right person probably isn't your type. That's exactly the point.

---

Ada Jin is Co-Founder of LAMU, an AI matchmaking platform and in-person singles club based in Seattle.

Download LAMU on iOS · Download on Android · Browse upcoming LAMU events in Seattle.