Singapore's GovTech agency has launched FirstDate, a pilot dating service initially restricted to government workers, that matches users through the Gale-Shapley Stable Marriage Algorithm — the same deferred-acceptance framework used since the 1950s to assign medical residents to hospitals in the United States. Users authenticate via Singpass, complete a 30-plus question questionnaire covering lifestyle, love languages, communication style, and dealbreakers, and receive exactly one match per cycle. Both parties must agree before contact details are shared. The questionnaire itself is a revealing artifact. It mixes close-ended personality sorting ("Team sunrise or team 2am?") with open-ended prompts ("Write a short note for your future match") and preference filters (age range, religion, dietary restrictions, smoking). The system also collects metadata: how many first dates you've been on in the past year, whether you've used dating apps before, and when your last first date was. An earlier GovTech prototype proposed subsidizing first-date meals, which likely explains the cluster of food-related questions. The core problem is mechanical, not cosmetic. Gale-Shapley guarantees stable matches — meaning no two unmatched people would mutually prefer each other over their assigned partners. This is powerful when participants are committed to the pool, as medical residents are committed to the Match. Dating has no such commitment. Users can ghost, prefer someone outside the system, or simply not participate in the next cycle. The mathematical guarantee that underpins the algorithm evaporates when outside options exist. Hinge claims to use the same algorithm and its "most compatible" suggestions are routinely ignored. The questionnaire design compounds the issue. Open-ended text fields and ambiguous lifestyle questions lack objective scoring rubrics. The algorithm requires rank-ordered preferences to function — converting "What's a country that's ruined all other countries for you?" into a compatibility ranking is an unsolved natural-language-processing problem dressed up as matchmaking. The gap between data collection and algorithmic input is where the system's credibility thins. The Aphrodite Project, a university-based matching service using a modified Gale-Shapley approach that delivers one match per year, offers a natural comparison. Even with algorithmic optimization and zero swiping fatigue, ghosting persists. The constraint isn't match quality — it's that human romantic behavior doesn't operate like a two-sided market with binding commitments. Singapore's fertility rate has been below replacement for decades, and government matchmaking has a long pedigree here — the Social Development Network ran for years before shuttering three years ago. FirstDate represents GovTech's attempt to apply algorithmic rigor to a problem that prior bureaucratic approaches couldn't solve. The Singpass identity verification is genuinely useful for reducing catfishing and fraud, a real advantage over commercial apps. But verification solves the trust problem, not the matching problem. The pilot's restriction to government workers is both a pragmatic testing choice and a constraint that makes the algorithm's weaknesses more visible: a small, homogeneous pool with abundant outside options is the worst-case scenario for stable matching. If FirstDate scales nationally, pool size improves but the fundamental mechanism mismatch remains. The algorithm is solving for stability in a domain where nobody is bound to accept the result.
