An Algorithm Can Find You a Stable Match. Chemistry Is Another Matter.

Key takeaways

• A stable matching leaves no two people who both prefer each other to their assigned partners.
• In the basic Gale–Shapley model, making proposals gives one side an advantage.
• An algorithm’s guarantees depend on preferences that may poorly reflect real attraction.
• A public matchmaking service should assess mutual interest and participants’ ability to decline, alongside match counts.

A government can arrange the introduction. It cannot certify the chemistry. For an initiative such as Singapore GovTech’s FirstDate, algorithmic matchmaking raises a deceptively simple question: what would a “stable” match actually guarantee?

Stable is a mathematical condition

The classic problem behind the Gale–Shapley algorithm is called the stable marriage problem. Despite the name, it concerns pairing members of two groups according to their preferences.

The basic model makes tidy assumptions. Both groups have the same number of people. Everyone ranks every member of the other group, with no ties.

Suppose Alex and Jordan have each been matched with someone else. Alex prefers Jordan to their assigned partner. Jordan feels the same way about Alex.

They have a mutual reason to abandon their assigned matches and choose each other. Mathematicians call them a blocking pair. A matching is stable when no such pair exists.

That definition contains no promise that everyone gets their first choice. Someone can end up with a partner far down their list while the overall matching remains stable.

Nor does stability tell us whether a date will be enjoyable, whether attraction will develop, or whether a relationship will last. The word is doing considerably less work than a dating-service customer might expect.

The side that proposes gets an advantage

The algorithm works through a sequence of proposals.

Members of one group propose to their first choices. Each recipient temporarily keeps their favorite proposal and rejects the others. Rejected participants then approach the next person on their list whom they have not already asked.

A recipient can replace a tentative match if a better proposal arrives. The process continues until no further proposals are needed. Under the basic assumptions, the result is stable.

But the procedure has a consequential property: it favors the proposing side.

For the same set of preference lists, there may be several possible stable matchings. Gale–Shapley gives each proposer their highest-ranked partner available in any of those stable matchings.

That qualification matters. It does not guarantee their favorite person overall. It guarantees their best outcome within the set of stable outcomes.

Choosing which group proposes is therefore a substantive design decision. Two services could display the same cheerful “Match found” message while applying rules that produce different results for participants.

The interface does not reveal who got the mathematical advantage.

Your profile preferences may not survive dinner

In the mathematical model, the preference lists are ready before the algorithm starts. In dating, producing a meaningful list may be the harder problem.

Age, hobbies, and location can help someone decide whom to meet. They cannot fully capture whom that person will enjoy talking to.

Someone unremarkable on a profile might become the most interesting person in the room. A seemingly perfect candidate might make an hour feel unusually long.

The algorithm can work with the preferences it receives. It cannot guarantee attraction those preferences fail to capture.

If a service infers preferences from questionnaires or behavior, the quality of that inference matters alongside the matching procedure. A mathematically correct result can still disappoint when the input misrepresents what someone wants.

Preferences also change. A matching that is stable under today’s rankings may be unstable under next month’s.

There is no relationship warranty hidden in the phrase “stable marriage.”

Count the second dates, too

A government matchmaking service needs a broader definition of success than assigning everyone a partner.

Match counts tell us how many introductions the system produced. They do not tell us whether those introductions were welcome or worthwhile.

A more revealing question is whether both people wanted to meet again. One enthusiastic participant and one reluctant participant is a different outcome from mutual interest. Participants should also be able to decline an unwanted introduction easily.

The rules deserve explanation, too. What preferences does the service use? Which side proposes? Those choices help participants understand their results. Government involvement alone does not establish the quality of any individual match.

An initiative such as FirstDate can be valuable simply by creating opportunities to meet. That could mean introducing someone who fits a participant’s stated preferences—or someone they might otherwise have overlooked.

Stable matching solves a specific mathematical problem. A worthwhile dating service still has to show that its introductions work for the people receiving them. The algorithm can arrange the first meeting; the reason to have a second has to emerge across the table.