Why we know our own criteria badly
The preferences people state before an encounter predict very poorly the people they then choose. The result is solid, and it unsettles the identikit portraits.
An unusual set-up
From the 2000s onwards, Paul Eastwick and Eli Finkel made use of a precious experimental situation: speed dating sessions. Participants first state what they are looking for, then meet a series of people for a few minutes each, then designate those they would like to see again.
For the same person, and within the same day, one thus has both their stated criteria and their actual choices. Few protocols allow this direct comparison, and that is what gives the result its weight.
The result
Stated criteria predict actual choices very poorly. Someone who declared they attach great importance to ambition does not retain any more of the partners judged ambitious by the other participants. The correspondence between the portrait described and the person chosen is weak.
The authors also note a classic difference that does not survive the move to action: stated preferences differ markedly between men and women, actual choices much less so.
What this does not mean
The reasonable interpretation is not that people lie, nor that their criteria would be unimportant over time. It is more limited: knowledge of one's own preferences is imperfect, and an abstract criterion does not work the same way as a person who is present.
There is a difference between knowing what one is looking for and recognising it when it stands in front of you. Describing an ideal mobilises categories, choosing mobilises an overall impression, and the two operations do not summon the same information.
Lists
This result throws light on the fate of lists of criteria, readily drawn up after a disappointment. They are useful for ruling out what one no longer wants, and far less so for designating what one does want.
They also suffer from a construction flaw: they are drawn up from the most recent experience, and therefore calibrated against a particular case rather than on a regularity. That is how one requirement gets replaced by its mirror image, without the result being any better.
The effect on algorithms
Services promising matching based on stated criteria rest precisely on what this result weakens. A review published in 2012 by Finkel and colleagues concludes that nothing establishes the superiority of these algorithms over simpler matching.
What these services really change is access: they considerably widen the number of people one can meet. That is a substantial effect, but it has to do with putting people in contact, not with the quality of the selection.
Key points
Criteria stated before an encounter predict actual choices badly, including among those who state them with confidence.
Knowing what one is looking for and recognising it in someone's presence are two distinct operations.
Lists of criteria serve better to rule out than to designate, and they are often calibrated against the most recent disappointment.
This article describes documented mechanisms and disagreements. It recommends no position, no party and no vote, and commits to no reading other than that of the work it cites.
And you, where do you stand?
The Civix test situates your convictions on six dimensions and matches you with one of the twelve profiles, without ever telling you what to think.
Start my test