The phrase “EQ assessment” may suggest a general emotional-intelligence test: Can you read emotions? Are you empathetic? In matchmaking, however, the questions can look much more operational.
A typical item may contrast two undesirable patterns. One person struggles to read the room, misses changes in a partner’s feelings, and tends to behave in irritating ways. Another makes little effort to improve relationships with family or friends and prefers being alone. The respondent is then asked which pattern would be harder to tolerate in dating or marriage.
That is not merely “Are you sociable?” It is closer to asking: Which failure mode would damage shared life more for you?
It feels like a romantic version of FMEA—Failure Mode and Effects Analysis. Before asking which movies both people love, the system is already asking which recurring household bug would be unacceptable.
1. Is there a hidden ranking? Part of that is public, part is unknown
The terms of Aichi’s public marriage-support system describe the EQ assessment as a “values diagnostic test.” They state that the AI introduces partners based on the EQ assessment result plus desired partner conditions, and that people whose desired conditions mutually match are introduced with priority.[1]
So the claim that the assessment affects matching is not speculation.
What is not public is the internal weighting. We cannot verify that a particular answer contributes a specific number of points to a “boundary” score, or that it moves a certain profile down three positions.
A careful summary is therefore:
- EQ results are used in matching: confirmed
- mutual preference conditions affect priority: confirmed
- exact question weights and latent scores: not public
- the questions appear to map tolerance for shared-life friction: reasonable inference
The entrance to the algorithm is visible. The scoring sheet is not.
2. “What I cannot tolerate” may matter more than “what I like”
Dating profiles naturally emphasize shared likes: food, travel, films, hobbies, and weekends.
Long-term relationships have another problem. A couple may share ten interests, but one repeatedly triggered dealbreaker can dominate everyday life.
Across six studies with more than 6,500 participants, Jonason and colleagues found that relationship dealbreakers mattered strongly, especially in long-term contexts. Negative dealbreaker information was weighted more heavily than positive dealmaker information.[2] A 2022 follow-up study similarly found that learning dealbreaker information reduced willingness to go on another date, with stronger effects for long-term than short-term relationship contexts.[3]
In some situations, “we both cannot stand shouting” may matter more than “we both love cats.”
The front end of matchmaking may look like a points game. The back end of married life is often dominated by recurring penalties.
3. The strongest model is not “same dislikes,” but “my intolerance × your behavior”
Two people disliking the same thing does not automatically create compatibility.
Both may say they hate possessiveness, while one still demands constant location updates. Both may say they need alone time, yet one becomes anxious whenever the other goes out independently.
A more useful matching logic is:
A. What I find hard to tolerate
×
B. How likely you are to do that
This includes boundaries and coordination:
- how much alone time is treated as personal territory
- how much one partner intervenes in the other’s plans
- whether “consultation” means information sharing or permission seeking
- how major joint-impact decisions about money, housing, children, family, or work are made
- whether conflict leads to repair attempts, avoidance, or withdrawal
These variables are less glamorous than shared hobbies, but they are called repeatedly in daily life.
If a system separately measures behavior tendencies and unacceptable partner behavior, then cross-matches them, the design would make practical sense. Public documentation does not disclose whether the Aichi system does exactly that.
4. “More similar” does not automatically mean “happier”
Research also warns against a simple “find your clone” rule.
One study of 248 married couples found that overall profile similarity and some forms of value similarity were associated with higher marital satisfaction.[4] Yet a 2023 study reported that Big Five personality similarity played a negligible role in explaining relationship and life satisfaction.[5]
So similarity itself is not a universal cure.
Some dimensions benefit from alignment; other differences are manageable. The important task is not to find an identical person but to avoid mismatches that are both frequent and high-impact.
Different movie tastes are easy to route around. A person who needs substantial solitude paired with someone who interprets solitude as rejection may hit the same conflict every week.
5. This “EQ assessment” looks closer to a values-and-friction map than a pure emotional-intelligence test
The service itself calls the instrument a values diagnostic test.[1]
That matters because an academic measure of emotional intelligence and a matchmaking values assessment are not identical things. Here, the practical function appears closer to mapping where two people might repeatedly clash in shared life than to compressing emotional skill into one global number.
A question contrasting “poorly reads social cues” with “does not work to improve relationships” illustrates this well.
There is no moral answer. Choosing A does not make someone better than choosing B.
The useful interpretation is not virtue ranking, but which failure mode you are least able to live with.
6. Then comes the final boss: pool size
Even an excellent ranking algorithm cannot create candidates that do not exist.
If 1,000 people remain after basic screening, ranking the best-fitting 20 can matter a great deal.
If only a handful remain, AI is mostly sorting a short list, not summoning new humans.
The Aichi terms explicitly warn that, depending on preference settings, a user may not receive a match every month.[1]
A simplified model of matching quality is:
candidate pool × hard filters × compatibility ranking
If the first factor becomes tiny, making the last factor extremely sophisticated has diminishing returns.
AI is a sorting function, not a magic circle.
7. Conclusion: the useful goal may be fewer surprise landmines, not a perfect soulmate prediction
Treating an EQ assessment as an oracle that discovers one’s “true personality” would be excessive.
Its more practical value may be to:
- identify what someone genuinely cannot tolerate
- compare that line with a partner’s behavior tendencies
- surface boundary and joint-decision friction earlier
- reduce recurring high-impact mismatches rather than merely maximizing shared likes
Public information confirms that EQ results and desired conditions are used as matching inputs.
The internal weights remain undisclosed, and a small candidate pool limits the power of any ranking system.
So the conclusion is less mystical and more useful:
AI does not invent the perfect partner. But if it reduces a few “Wait, THAT was the landmine?” surprises, it may already be doing valuable work.
