Does dating around make it harder to fall in love? Romantic overfitting explained

I noticed something strange while watching Love on the Spectrum, a show that follows autistic people as they try to date and find relationships.

Dating on the show is not easy.

People misread signals. Some become intensely attached. Some struggle to tell whether another person is actually interested.

But there were also moments that felt almost startlingly simple.

Someone with very little romantic experience would finally meet a person who liked them back.

And sometimes they just fell hard.

There was a kind of gratitude and intensity to it that made me wonder about the opposite problem.

Can too much romantic experience make love harder to reach?

Every relationship creates another comparison point.

Another experience worth recreating.

Another standard the next person may have to meet.

Modern dating adds something even stranger: thousands of people you never dated but can still compare a real partner against.

I call the possible result the BiohackBeast concept of romantic overfitting.

In machine learning, overfitting happens when a model learns its old examples so specifically that it becomes worse at handling something new. Dating may have a similar failure mode.

You get better at knowing everything you might want.

And more likely to expect one real person to combine the best parts of several others.

Inside this article

The quick answer

There are reasons to take this idea seriously.

A 2023 study using U.S. National Survey of Family Growth data found an association between premarital sexual history and divorce. People who had no premarital sexual partners other than their eventual spouse had the lowest divorce risk, while those with more previous partners had higher risk. The relationship remained significant after researchers accounted for several differences in family background, religiosity, personality, and mental health.

That leaves an important question unanswered: is partner history itself changing later relationships, or are people who accumulate more partners different in ways the study could not fully measure?

Other evidence gives us possible mechanisms.

Past experience can shape the standards used to judge relationships. A large meta-analysis found that people were less committed when they perceived better alternatives outside their relationship. And in a 2025 dating-app experiment, people became more selective as they evaluated more profiles.

Even the biology raises questions. In prairie voles, a pair-bonding mammal commonly used to study attachment, previous bonding experience can change how later bonds develop.

We do not yet have the human study that connects all these pieces.

But they point toward an interesting possibility:

What if every relationship changes the standard the next relationship has to beat?

Partner C is not meeting the same person as Partner A

Dating advice usually treats experience like information gathering.

You date someone.

You learn.

You date someone else.

You learn more.

Eventually you “know what you want.”

But the person doing the choosing is changing too.

Self0 + Partner A → Self1

Self1 + Partner B → Self2

Self2 + Partner C → Self3

Partner C is not meeting Self0.

They are meeting someone who now knows what A felt like.

What B felt like.

What they loved.

What they hated.

And what they now hope to find again.

This is path dependence: where you end up depends partly on the path you took to get there.

Your romantic history does not just sit behind you.

It may change what the next person has to compete with.

Past relationships can change the comparison point

Relationship science already has a concept close to this: comparison level.

Your comparison level is the internal standard you use to judge whether a relationship feels good enough. Past relationships and experiences can help shape it.

So one person thinks:

This is amazing.

Another thinks:

I have experienced better.

The partner did not change.

The reference point did.

There is also evidence that preferences and partners can shape each other.

In a prospective study of 763 single people, researchers measured participants' ideal-partner preferences before they entered relationships. Those preferences predicted characteristics of the partners they later chose. But once relationships formed, preferences also shifted toward the actual partner.

A later 13-year study covering 322 relationships found a similar pattern: earlier ideals predicted later partner traits, while people's current ideals became even more similar to their current partner.

So standards are not permanently fixed.

But romantic experience gives the mind more material from which to build them.

That is where romantic overfitting becomes interesting.

Romantic overfitting

Overfitting comes from machine learning.

A computer is supposed to learn the broad pattern behind its examples.

But sometimes it learns those examples too specifically.

It becomes excellent at matching what it has already seen and worse at handling something new.

Dating may have a similar failure mode.

Imagine five relationships.

One person gave you extraordinary sexual chemistry.

Another was unusually affectionate.

Another was highly ambitious.

Another made you laugh constantly.

Another gave you the deepest emotional connection of your life.

The problem is not that these are bad things to want.

The problem is that they came from different people.

And each trait came with tradeoffs.

The ambitious person may have had little free time.

The spontaneous person may have been unreliable.

The calm, stable person may have been less exciting.

The sexually intense relationship may have been chaotic.

Memory can preserve the peaks while forgetting the tradeoffs.

Then Partner F gets compared with:

A's chemistry.

B's affection.

C's ambition.

D's humor.

E's emotional connection.

Partner F is competing with a composite person who never existed.

That is romantic overfitting.

The composite partner can become contradictory

The standard can also become hard to satisfy because desirable traits often pull against each other.

You may want someone:

  • extremely ambitious but always available
  • highly adventurous but completely predictable
  • dominant but never controlling
  • sexually exciting but totally safe
  • independent but constantly attentive

None of those combinations is impossible.

But every added requirement narrows the field.

Experience can teach you what matters.

It can also create a more complicated person for reality to match.

First love starts with fewer comparisons

There is no strong evidence that first love uses some unique brain circuit that later love cannot access.

But first love starts from a different position.

There are fewer romantic experiences available as reference points.

Instead of entering the relationship with years of previous partners to compare against, an early serious relationship may help create some of those standards in the first place.

Long-term research on ideal-partner preferences supports the broader idea that relationship experiences and partner characteristics can become incorporated into what people later consider ideal.

This raises a possibility that has not been directly tested:

The first person you love may only have to compete with your imagination.

Everyone after them also has to compete with your memory.

Internet dating communities sometimes describe an extreme version of this as becoming “widowed” to an exceptional former partner: someone sets such a powerful reference point that later partners struggle to measure up.

The permanent “alpha widow” version of that claim has not been established scientifically. But the underlying idea is less exotic: previous partners can become reference points, while relationship research shows that ideals can both influence partner choice and shift in response to the partners we actually have.

Dating apps add people you never even dated

Past partners are only half the comparison pool.

Modern dating adds imagined alternatives.

People you have never lived with.

Never fought with.

Never seen tired, sick, stressed, boring, or difficult.

You mostly see the pitch.

Your real partner is a complete human being.

The alternative is often a highlight reel.

A 2025 dating-app experiment tested this by varying how many profiles participants evaluated. Acceptance fell from 24.1% in the low-exposure condition to 17.1% in the high-exposure condition.

In other words, the more people participants evaluated, the less likely they became to accept one.

But having more profiles in front of you is not the only way researchers have tested abundance.

In a separate 2022 experiment, researchers manipulated how many potential partners participants believed were available. When the dating pool seemed unusually large, people reported more difficulty choosing, more fear of ending up single, and temporarily lower self-esteem.

That study tested the feeling that alternatives were abundant. The 2025 experiment tested something narrower: what happens as people actually work through more profiles. There, acceptance rates fell, but self-esteem did not.

A third experiment using a simulated dating app found no clear choice-overload effect at all.

So the evidence does not show that a giant dating pool automatically makes people miserable.

What it does suggest is more specific and more relevant here: under some conditions, abundance changes how people evaluate and reject potential partners.

Alternatives matter even when the relationship is good

A 2019 meta-analysis of the Investment Model combined 202 samples and more than 50,000 people.

People tended to be more committed when they were satisfied with their relationship and had more invested in it.

But commitment fell when the alternatives outside the relationship looked better.

Importantly, alternatives mattered even after accounting for relationship satisfaction.

You can be happy with someone and still become less committed when other options look increasingly attractive.

Now put that psychology inside a phone that can produce new alternatives every few seconds.

The long-term effect has not been directly established.

But the psychological mechanism already exists.

Does more partner history weaken pair bonding?

This is where the evidence becomes provocative but incomplete.

Several human datasets link more premarital partners with worse later relationship outcomes.

In the 2023 U.S. study discussed earlier, people with no premarital sexual partner other than their eventual spouse had the lowest divorce risk. Those with more previous partners had higher risk, and the association survived several statistical controls.

That makes partner history worth investigating.

But observational data cannot tell us exactly what causes the pattern.

People who accumulate more partners may already differ in novelty seeking, impulsivity, attitudes toward commitment, or other characteristics that also affect relationship stability.

The pattern has also varied across historical cohorts, which makes a simple model where each additional partner causes a fixed amount of biological damage difficult to defend.

Higher partner count predicts worse outcomes in several datasets.

Whether repeated partnering itself weakens human pair bonding remains unresolved.

The prairie vole evidence makes the question harder to dismiss

Prairie voles are small rodents that form unusually strong pair bonds, which is why scientists often use them to study attachment.

A 2020 experiment asked a question unusually close to ours: does having a previous bonded partner change what happens with the next one?

Researchers first paired male voles with Female 1 until they formed a clear preference for her.

Then they removed Female 1.

The males spent either 48 hours, two weeks, or four weeks alone before being introduced to Female 2.

The males were still capable of bonding with Female 2.

But bonding with her did not mean Female 1 had stopped to matter.

Researchers first tested whether the male preferred Female 2 over an unfamiliar stranger. Then they ran the more revealing test: Female 1 and Female 2 were placed at opposite ends of the chamber, and the male could choose which partner to spend time with.

Only males that had spent four weeks separated from Female 1 consistently chose Female 2 over Female 1.

After just 48 hours or two weeks, the males could develop a preference for Female 2 over a stranger, but when Female 1 was brought back for a direct comparison, they did not reliably choose Female 2 over her.

In plain English: the new bond could form before the old bond had fully lost its pull.

Previous bonding experience also changed what predicted the strength of the next bond.

With Female 1, how quickly a male began mating did not predict how strongly he would later prefer her.

With Female 2, it did.

Males that mated sooner with Female 2 later showed a stronger preference for her.

So the second bond differed from the first in two specific ways:

  • The first partner could continue competing with the new partner for weeks.
  • How quickly mating began predicted the strength of the second bond, even though it had not predicted the first.

That is the path-dependent finding.

The male did not lose his ability to bond.

Instead, Partner 2 entered a bonding system that was still carrying Partner 1, and previous experience changed what predicted how strongly the new bond formed.

One important limitation is that this experiment only tested male voles.

Female prairie voles can differ from males in how quickly they form partner preferences and in which hormonal systems contribute most strongly to bonding. Running the same Male 1 → separation → Male 2 design in female voles could produce a different pattern.

That would be especially interesting because female mate choice plays a major role in sexual selection across many species.

Humans are not prairie voles, so this does not prove that previous human relationships alter later attachment in the same way.

But in another pair-bonding mammal, the first bond changed the conditions under which the next bond formed.

Oxytocin probably does not explain this by itself

If repeated partner experience changes later bonding, oxytocin is an obvious molecule to investigate.

It is heavily associated with social attachment and is often described as the “love hormone.”

But newer research shows that pair bonding is much harder to reduce to one chemical.

In a 2023 genetic experiment, prairie voles bred without functioning oxytocin receptors could still form pair bonds.

Another prairie vole study of bonding and separation found changes in oxytocin-producing neurons after pair bonding that moved back toward baseline after enough time apart.

So if repeated partnering changes future bonding, oxytocin alone is probably not the explanation.

Learning, reward, comparison, novelty, and attachment history offer more interesting possibilities.

Sexual novelty adds another layer

Sexual novelty affects reward too.

The Coolidge effect describes renewed sexual interest when a novel mate appears after sexual interest in a familiar mate has declined.

The effect is well established in several animal species, and dopamine appears to play an important role.

Humans also show sexual habituation in laboratory experiments: repeated exposure to the same erotic stimulus can reduce arousal, while novelty can restore some of it.

A review of the Coolidge effect and sexual novelty discusses this reward response across species.

Laboratory sexual novelty is obviously not the same thing as repeatedly changing real partners.

But it establishes an important piece of the puzzle:

Novelty itself can become rewarding.

Repeated partner rotation could therefore create another learning loop:

new person → anticipation → reward → familiarity → new person

This overlaps with the BiohackBeast idea of accelerated intimacy, where partner dance can repeatedly combine intimacy cues with novelty and partner turnover.

In both cases, the interesting variable is not intimacy alone.

It is intimacy combined with novelty and partner rotation.

The telegony rabbit hole

There is one stranger biological analogy.

Telegony is the old idea that a previous male can somehow influence offspring later fathered by another male.

Classical human telegony is not supported.

But a 2014 experiment in Australian flies produced something genuinely strange.

Researchers raised male flies on either a rich or poor diet, creating larger, high-condition males and smaller, low-condition males.

A female first mated with one of those males while her eggs were still immature.

Two weeks later, after those eggs had matured, she mated with a different male and laid them.

The second male fathered about 87% of the offspring.

Yet the first male still predicted their size.

Females that had first mated with a high-condition male produced larger adult offspring, even though the second male fathered most of them.

When females were merely exposed to the first male without mating with him, the effect disappeared.

That pointed researchers toward something transferred during sex. Their leading explanation was that molecules in the first male's seminal fluid affected the female's immature eggs before the second male later fertilized them.

So this was not the first male's DNA somehow sneaking into another male's offspring.

It was a non-genetic carryover from an earlier mating.

No comparable effect has been demonstrated in humans.

But the experiment makes one narrow biological point beautifully: a mating event does not always end biologically when the partner leaves.

For humans, the much stronger evidence for carryover is psychological.

We remember.

We compare.

And the next partner meets a person shaped partly by what came before.

Maybe love requires less optimization than modern dating teaches

This brings me back to Love on the Spectrum.

There is no evidence that autistic people somehow love more purely.

But a 2024 review of autism and online dating found that autistic adults often face greater barriers to meeting partners and tend to have less relationship experience, even though many want romantic relationships.

Maybe what looked so different on screen was something simpler.

Two people were paying attention to the person in front of them.

Not an ex.

Not 500 profiles.

Not an imaginary combination of five previous partners.

Just one person.

Modern dating assumes more experience makes us better at knowing what we want.

Maybe it does.

But experience also gives us more people to compare.

More peaks to remember.

More alternatives to imagine.

And more traits to combine into someone who may not exist.

That is romantic overfitting. Past partners and imagined alternatives can merge into a composite partner built from the best pieces of different people.

The more detailed that composite becomes, the harder it may be for one real person, with their own strengths and tradeoffs, to measure up.

The first person you love may only have to compete with your imagination.

Everyone after them also has to compete with your memory.

Back to blog

Leave a comment

Please note, comments need to be approved before they are published.