Dating Algorithms: Continuing the Conversation...

OPW - Jan 23 - At iDate we had a stellar session on Dating Algorithms. Dr Steve Carter from eHarmony, Dr Rense Lange from Integrated Knowledge Systems who worked on the TRUE.com dating algorithms, and Michael McNicholas, CEO of PerfectMatch, took questions. Lets continue the conversation in the comments.

Comments

[George Blair-West]

I’m an Australian psychiatrist and relationship therapist, writing a book on how to find a partner that brings out the best in you.

This question is primarily for Steve Carter – you mentioned that eHarmony has been collecting longitudinal data on couples who subsequently married - which is necessary data to answer this question.

By way of context - There are obvious complexities with matchmaking with perhaps the greatest hurdle being the problem with asking people to describe an ideal match when the research shows they’re not good at predicting what they actually want. Studies of speed dating, for example, have shown that the potential partners that daters reported feeling attracted to, did not actually fit the descriptions they had given beforehand of their ideal match.

And then Matching and Assortative Mating theories presume similarities are central to attraction. But, in one study the degree of similarity in political orientation, religiosity, and life values was not associated with levels of marital satisfaction. And then a study of newlyweds found that enjoying the same leisure activities caused husbands to feel closer to their wives, but didn’t matter at all to the wives.

Then the research into genetic difference triggered by Claud Wedekind’s seminal study into scent and an individual's major histocompatibility complex has shown that women are attracted on the basis of genetic complementarity i.e. not too similar so as to avoid in-breeding (risk of genetic defects) and not to different so as to avoid out-breeding (which can dilute robust genes).

My question is then, without releasing details of your algorithms, has eHarmony identified differences e.g. in personality traits, rather than similarities, that correlate with marital satisfaction? If so, can you give us some examples of such differences?

[Steve Carter, Ph.D., VP of Matching, eHarmony]

Hi George,

If you look at marital quality from the perspective of an individual, you can find personality factors where “more different” = “better marriage” if. For example, Neuroticism has a strongly negative univariate relation to marital quality. People who score high on Neuroticism are annoying as hell. So, if you measure satisfaction among married individuals, and rank order it on the absolute difference between them and their spouse on a measure of Neuroticism, you will find that highl Neuroticism people are more likely to be happy in their marriage if they are married to a Low Neuroticism spouse. Statistically, you’ve now demonstrated complementarity!

However, when you look at a dyadic measure of marital quality (i.e., a summed or averaged measure of marital quality from BOTH members of the marriage) a different picture will emerge. True, High Neuroticism individuals are most likely to be happy when paired with Low Neuroticism individuals, but those Low Neuroticism individuals are NOT being optimized. Indeed, the best marriages are between two Low Neuroticism individuals, and the worst are between two High Neuroticism individuals. So, is this relation between Neuroticism and Marriage Quality still complimentary?

The fact is, when you are looking to prove a theory of similarity or a theory of complementarity, you can create data that will support your POV. However, if you want to engineer a solution for recommending partners, you will want to optimize for the happiness of both members of the dyad, and you will find that “similarities” seem to outweigh “differences” as predictors of success the vast majority of the time.

For further discussion on this, I strongly suggest that you correspond with our Dr. Gonzaga who has been leading our compatibility research efforts since 2007.

[Nick Tsinonis]

1. To the psychologists: For long-term compatibility, how much weight would you put against: interests vs socio-economic background vs personality vs physical attraction? What other major factors are there to consider?

2. To the data scientists: Have you had any surprising discoveries about dating or compatibility from data collected and analysed where the question was not pre-determined?

[Datinginsider]

This fascinates me to no end. Why do the matching attributes seem so generic and simple? Seems like it's more of a weighting issue and every dating site is different in terms of emphasizing certain attributes as more heavily weighted. I've always found energetic compatibility useful from a layman's perspective, but nobody seems to take that into consideration. Same could be said for astrology, but I digress.

I would love to see a detailed differentiation between top site algorithms and verified results (yeah I know...). Dating sites should respect people enough to share more details about how they match. Slippery slope perhaps, but nobody has even tried to be open about their methodology, it all gets mired down in academic-speak, so to say. As far as I'm concerned, Match, EH, Chemistry and POF are all pretty much exactly the same when it comes to returning relevant matches, it's more about the types of people on each site (ducks). It's just marketing-spend battles which is so sad. You guys are the unsung heros of the industry yet you are forced to hide behind ads featuring 1/2 naked people, a damn shame if you ask me.