It’s no secret romantic relationships are a complex affair.

But a new study is helping people better understand the science of love, utilising machine learning and relationship science to crunch the data and uncover the most important factors in any successful relationship.

The study, conducted by a Western University-led international research team, analysed information from more than 11,000 couples across 43 different datasets around the world.

It set out to identify why some relationships are more successful than others by predicting satisfaction using machine learning algorithms.

And the report’s authors believe the findings can change how we navigate our relationships.

“This collective effort should guide future models of relationships,” says the report.

It’s not you… but it’s not me

The study found that individual characteristics (such as income and mental wellbeing) account for just 21 per cent of variance in relationships.

A partner’s personality, on the other hand, accounts only for around five per cent of the total relationship satisfaction.

When machine learning modelled the data, the individual differences did not regulate or moderate the relationship-specific variables.

"Really, it suggests that the person we choose is not nearly as important as the relationship we build," the study’s first author and director of the Relationships Decision Lab at Western University Samantha Joel told Inverse.

Relationship-specific traits meanwhile account for almost half (45 per cent) of all variance in the quality of a relationship, according to the study.

“The surprising part is that once you have all the relationship-specific data in hand, the individual differences fade into the background,” Joel added.

The traits that count

Although the research showed individual variables were far less important than relationship variables, some still ranked higher than others.

Life satisfaction was the number one individual variable, followed by negative affect, depression or feelings of hopelessness, attachment anxiety and attachment avoidance.

For relationship-based variables, perceived partner commitment was the most important factor, followed by appreciation, sexual satisfaction, perceived partner satisfaction and conflict.

"The dynamic that you build with someone — the shared norms, the in-jokes, the shared experiences — is so much more than the separate individuals who make up that relationship," Joel said.

Can AI really predict love?

And while relationship science – which takes into consideration psychology, sociology, economics, family studies and communication – is an established science, there are questions over whether AI can be used to predict future events.

The Western University machine learning study on relationships even concedes “none of these variables could predict whose relationship quality would increase versus decrease over time”.

Meanwhile, early evidence suggests machine learning models that have been trained on human interactions are performing poorly due to some of the erratic human behaviour to emerge from COVID-19.

“If a machine learning system doesn’t see what it’s expecting to see, then you will have problems,” behavioural analytics company Featurespace founder David Excell told MIT Technology Review.

“The world has changed, and the data has changed.”

However, there is some evidence to suggest AI is capable of successfully predicting relationship outcomes.

A 2017 study by researchers from the University of Southern California trained a machine learning algorithms to learn the relationship between tone of voice and relationships.

The algorithm then analysed video recordings of couple’s therapy sessions and was asked to predict the outcome.

It was correct 79.3 per cent of the time, while human psychological experts only managed to accurately predict the outcome in 75.6 per cent of cases.