Is Romantic Desire Predictable? Machine Learning Applied to Initial Attraction
ncbi.nlm.nih.gov
ncbi.nlm.nih.gov
EDIT: "past lives" refers to the past of each of our lives, not reincarnation. :-)
Example: When cutting a scene with a lot of people eyes, you can really push the limits because of the high likelyhood that the audience will have a certain amount of focus on some eyes.
I tried to major in film (it was a program halfway between film studies and technical work), and remember a big technical thing being not producing an "axis break" (I won't know the terms in english) by changing angles too suddenly or even changing lenses too abruptly from wide to long. Unless you were trying to break the rules on purpose. The big exception was "plane/counter-plane" shot sequences where we alternated between seeing two characters who were talking face-to-face directly. This was 15 years ago and not in a country with an actual film industry.
Then the reason why this worked was said to be insert long French theoretical evocation you identified with the characters. The lacanian gaze, that kind of thing.
So in light of what you just said, the whole "axis break" makes perfect sense: film sequences are trying to show images I might have plausibly apprehended in real life via eye movements. Because I'm always seeing it from someone's point of view. The lacanian gaze.
I think what could really help computer vision is not as much emulating human eye tracking as it is understanding images as short durations rather than instantaneous shots. The point is that during the eye's saccade reality is always changing.
Think of how close-distance flirting works: you can't use your stereo vision to focus on both her eyes at once -- you have to alternate ever so slightly, occasionally meet her eyes focusing on yours and then lose them. All of this happens automatically -- our visual perception of the situation is intrinsically dynamic.
Of course, here we're getting close to Henri Bergson's ideas about time and duration and Hubert Dreyfus' reading of Heidegger as a theoretician of situated existence. This is the same Hubert Dreyfus who told Simon, Minsky, etc. in the 50s they didn't have a chess computer yet and has recently passed away as a skeptic of automated cars.
Did you notice self-driving cars are slowly disappearing as a near-term possibility? We still haven't figured it out -- we're trying to teach computers with frozen cases (MNIST, Imagenet, etc.) when the real problem is training them on actual situations.
They don't need to emulate humans at all. They just need to exist in the same precise sense we exist. Computing power will keep increasing and I assume so will "deep learning" optimizers, but we need to figure out how to give computers a Dasein.
They are training AI agents on simulators, which are dynamic datasets as opposed to ImageNet which is a "frozen" dataset).
> we need to figure out how to give computers a Dasein.
Google and the whole AI community have been hard on the Reinforcement Learning bangdown. RL is based on situating an agent in the environment (real world or sim) where it learns to interact and achieve goals. That's situated AI.
My intuition is that simulators are key to the next AI phase - where a neural net would be like a scientist creating new hypothesis and experiments, and the simulator would be like a lab, testing out those ideas (simulation + MC search + neural nets)
I hope so. Reinforcement learning and apprenticeship learning (inverse RL where the machine watches a human dealing with the problem) are probably the ground zero before which everything will look like toys and Kaggle problems.
Eye movements while taking to someone, posture during conversation, and voice characteristics could all be measured.
The point is feeding the network slightly different images as they're captured by a small field of vision. The real perceived object is not static.
You're going to need a temporal, recurrent type of model. Also: a bigger boat.
For example on two different occasions when I was hanging with two different friends of mine we looked at Tinder profiles of women on my phone and both of them had almost the exact same reaction when I started swiping; "wow, you and me have really different taste in women!"
In both cases I was swiping on the women that I found to look the most attractive. Previously I might have thought that everyone would agree that these women were attractive.
Absolutely agree. I'm just saying that even if they were I think it'd be hard to measure attractiveness objectively because different people have different likes and dislikes.
However you probably are right that there are features that are considered attractive by almost everybody and that we just overlook those.
The things people look for in a partner to go home with for the night are different from the things people find "just generally attractive" (oh he's hot) which are different from the things people find good for long-term dating. I mean, obviously, but my point is the layer of complexity doesn't just start at "different things that cause attraction."
I wonder if the results would be different if the participants got to know each other better. In 4 minutes you're not going to know your partner very well, and any judgment would probably be based entirely on physical attraction.
On the other hand, there's lots of evidence suggesting that the things they were measuring are important for things like long-term relationship survival, etc. There's also studies showing that these "thin slices", while accurate, are not as accurate as what you'd get from intimate long-term knowledge of someone.
This isn't anything astonishing, but I have difficulty interpreting this as much more than "things like personality and attitudes are not very predictive of your superficial attractions to potential partners." I don't mean "superficial" dismissively, I just mean it in the sense of "not based on very much information."
Another way of looking at it is: how could you predict attraction in that time frame based on longer-term, deeper-seated attributes, when that kind of timeframe doesn't allow for assessment of those sorts of things? It's one prediction, but I doubt if you stepped back and came at it without knowledge of this study, that you'd make that prediction. I certainly wouldn't (I'm trying to imagine me predicting my relationships from the first 4 minutes--sure, I'd be able to guess pretty well who I definitely wasn't attracted to, and maybe who I was definitely was attracted to, but beyond that? pbfbfbtttt...)
I wasn't able to predict my most meaningful relationships from a month or so of being around them, sometimes on daily bases. Looking back at them, I might have had some kind of subconscious reaction I wasn't able to process, but I wasn't aware of what it meant at the time, which renders it useless.
However, do dating services ever do consistently better than chance (whatever chance means in such situations)? Plenty of Fish IIRC used to publish lots of stats.
Of course, a first date is also a pretty weird setting. The advantage is, it lasts long enough that there's a chance the "shy and embarrassed" can wear off, and the more normal person can come through.
That might be more attractive to other shy people, I'd guess.
Do other shy people look for a partner who is shy?
> These results suggest that compatibility elements of human mating are challenging to predict before two people meet.
Arranged marriages certainly seem to permit a capacity for arbitrary pairings, but when you consider the hereditary aspects behind an arranged marriage, where parents are choosing for children, the reality is that a parent's approximate attraction to someone they select for their children, even by proxy through that someone's parents, will probably land closer to what their children would intuitively approach on their own, than random chance.
A marriage is less likely to fail when it causes enough upsides to be worth the downsides.
People don't like the concept of trophy wives, gold diggers and strategic marriages because the pros and cons each side brings to the table are incredibly dissimilar in nature and the perceived value of them is highly subjective so to most people they look like a bad deal for at least one party. On the other hand only two people have to think it's a good deal for it to work out.
Which is not to say that leaving it up to the kids themselves is always successful...
Cut the "matching", just use first impressions as the primary filter.
Though, this raises all sorts of new problems in the romantic ecosystem.
If anyone has a subscription to YouTube Red I would recommend watching this episode from many years ago:
EDIT: Oh, Voyager. That's also on Netflix, and probably other places. S4E22, "Unforgettable".
FWIW, I have YouTube Red, and this video says I need to buy a $150 season to watch it.
If anyone has hulu, here's a link
Star Trek: Voyager S4 E22 Unforgettable https://www.hulu.com/watch/441748#i0,p20,s4,d0
We don't have to think of it philosophically.
If researchers can train some ML algorithm to predict it, then it is predictable, end of story.
If not, it's open (it could still be predictable and they couldn't come with the right algorithm, or it could just be too fuzzy, or even impossible to measure).
That said, if we drop it from "love" to simple "desire", then most of it should be fairly predictable and measurable with very small margin of error.
Humans are not that much of unique snowflakes as they say. In a given culture, the majority of the population will converge a lot in what they find attractive and what they don't -- this has been studied before.
Besides attraction, other things might come into play in predicting relation happiness - such as dialogue style, extraversion/introversion, specific needs of one in relation to the strengths of the other and specific situations in life the couple has to deal with. Some of these things might emerge from a psychological interview, others happen dynamically and are impossible to predict.
And most people are very time sensitive - the same person might be available now that was not a week ago. The availability for dating of some people measures in just months between long term relationships, so they are like 99% of the time unavailable even if they are compatible. Searching for a mate is hard because most compatible people are presently attached and you got to get them at the right time. That might be mitigated by going on a dating site, but unfortunately dating sites are full of people who are there for other reasons than serious relationships.
> Matchmaking companies and theoretical perspectives on close relationships suggest that initial attraction is, to some extent, a product of two people's self-reported traits and preferences.
Right from the first sentence, I can tell the results are going to be awful.
Here's the truth: appearance is significantly more important than anything else. I mean the face, the headshot from the chest and up, the body. But mostly the face. This is for both male and female attractiveness. The rest of one's "profile", their traits and preferences, just make for conversation topics unless there's something so polar opposite that it becomes a deal-breaker. I.e. in my research these are "threshold factors" not continuous variables for correlation.
> Each participant met each opposite-sex participant attending a speed-dating event for a 4-min speed date. Random forests models predicted 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.
And there are the poor results, as I expected.
We are shallow. Or so it seems. But we are actually incredibly complex processing machines. I can tell your levels of strength, aggression, trustworthiness, and intelligence fairly decently from looking at your face. I can also tell what your diet is like, how healthy you are.
Maybe not explicitly, but subconsciously. And women are amazing at this (we're better at looking at breasts and hip-to-waist ratios). Here's an interesting video getting into some of this: https://www.youtube.com/watch?v=rVE6kZW88lc
So how do I literally 10x a man's results in terms of his OkCupid profile? In terms of measured message responses received, and dates achieved?
I get them in shape. I fix their diet. I cut them to <12% bodyfat while adding pounds of lean muscle. I increase their testosterone levels as much as naturally possible over a 90 day period to start (that's my basic service). And then I take proper photos -- men, don't ever use a smartphone for a headshot. Smartphones have terrible focal lengths, lengthening your faces and making you look more feminine. Since it slims the face, I suspect a lot of social media heavy women are letting themselves gain weight because in their selfie photos (taken from a slightly inclined angle, no less) they look significantly slimmer.
I don't change shit in what they write on their profiles. Because that pretty much doesn't matter. Believe me, I've experimented. But the hardcore truth is that you need to look more masculine, unless you can show through photos that you're a multimillionaire. And even that I wouldn't recommend as being your main attraction asset (but it does work).
Maybe I should release some independent studies sometime, I have some great datasets :)
Edit - nice downvotes but no rebuttals. Happy to have a discussion, these kind of things bothered me to no end as someone who wanted to bank on being a software engineer with a good personality in order to get dates.
But natural T increase? This sets off my quack-o-meter.
What's kind of scary is that the natural range is so large - like 300 - 1000 ng/dl for testosterone, but you will literally feel like a different person if you fix all of the above factors properly and go from say 350 to 600. And living in that higher T state for months and years at a time, you will start to look different as well (generally in a good way for men).