From Academia: Personality Traits Don’t Predict Compatibility

2017 research published in the journal Psychological Science has investigated whether answers to a personality test can help to predict romantic attraction.

The study is entitled: ‘Is Romantic Desire Predictable? Machine Learning Applied to Initial Romantic Attraction’.

It used machine learning to look at the tendency of singles to romantically desire other people (‘actor variance’), to be desired by other people (‘partner variance’) as well as desire for specific partners (‘relationship variance’).

The authors note that many online matchmaking and dating sites claim to be able to match people on their potential for compatibility, but few of these claims have been scientifically vetted.

The study reads: “For romantic matching algorithms to be effective at all, one or more of the following three assumptions must be met: It must be possible to predict the emergence of romantic interest in the form of (a) who desires others on average (i.e., actor variance in desire), (b) who is desirable on average (partner variance), and (c) who uniquely desires whom (i.e., relationship variance (…)).

“If the first or second assumptions were true, an algorithm could help people form relationships by excluding exceptionally misanthropic (i.e., low actor effect) and/or undesirable (i.e., low partner effect) people from the group of eligible daters. But it is the third of these components—unique desire—that is the raison d’être behind commercial approaches to matching.”

Relationship variance, therefore, is the outcome which tests claims of ‘tailored’ matches and ‘perfect’ matches suited to each user.

Most studies in this area have found that attraction is difficult to predict from information collected before two people have met. Attraction tends to build over time, the authors note.

The models developed in the study were able to predict “4% to 18% of actor variance and 7% to 27% of partner variance; crucially, however, they were unable to predict relationship variance using any combination of traits and preferences reported before the dates.”

This suggests that apps may be able to filter people on how desirable they are and how fussy they are based on questionnaires, but that they may struggle to predict compatibility in any 1-to-1 sense.

Find the full study here.

Scott Harvey

Scott is the Editor of Global Dating Insights. Raised in Dorset, he holds a BA from The University of Nottingham and an MSc from Lund University School of Economics and Management. Previously he has written about politics, economics and technology for various online publications.

Global Dating Insights is part of the Industry Insights Group. Registered in the UK. Company No: 14395769