[Applause]
comments are something we can predict in
1705 a time when many people thought
that comments were supernatural omens
Edmund Hali called his shots he had been
cataloging numerous comments and
concluded that they followed a natural
predictable pattern the same comet was
returning to Earth every 76 years and he
predicted that this comet would return
following a particular orbit in the sky
in 1758 when this prediction was
confirmed it demonstrated the awesome
predictive power that science can hold
earthquakes are something we can’t
predict the year is now 2017 and the
u.s. loses over 4 billion dollars a year
to earthquake damage and sure we know
generally where they’re likely to show
up that’s why this building has seismic
protection
but as Nate Silver discusses in his
recent book earthquakes are chaotic we
still don’t know when or where a
specific earthquake is going to occur
with any useful level of accuracy
what about romantic attraction is
attraction predictable like a comet or
certain combinations of people’s traits
and preferences predictably reliably
results in attraction or as a traction
chaotic like an earthquake where we just
don’t know when or where it’s going to
show up how many people here are
currently on the dating scene it kind of
sucks right
I mean some people really like dating in
the same way that some people really
like shopping you don’t even need to be
looking for anything in particular you
just like to browse but for those of us
type a people who just want to get in
find our perfect partner and get back
out dating can be exhausting you have
these experiences where you’re on a date
with someone who looks great on paper
and you’re turning down the dessert menu
and asking for the check because he’s
been talking about his collection of
fantasy swords
since the soup’s arrived and you’ve yet
to get a word in edgewise or she’s
literally swiping for new matches while
sitting right across from you and if
you’re type a like me you’re thinking
there has got to be a way to streamline
this process do I have to kiss all the
frogs can’t I just jump cut to the part
where I actually connect with someone
that’s the allure of matching algorithms
technology does the hard work for you
you just fill out some questionnaires
about who you are and what you like and
voila
here are some people you’ll actually
click with you get to get in find your
perfect partner and get back out it’s a
really alluring concept so much so that
people are willing to pay a lot of money
for the service but our matchmaking
algorithms actually feasible is
attraction like a comet well when we’re
talking about prediction we have to talk
about precision how much of attraction
can we predict because predicting all of
it with perfect accuracy is a really big
ask unreasonably big but predicting some
of attraction having at least a vague
sense of whether two people will like
each other or not it’s a much smaller
more reasonable ask all that’s required
is that people actually like the things
they say they like
what
[Applause]
what’s they I’m filling out an online
questionnaire and I say that I like men
who like puppies and are clean-shaven
and like to go surfing on the weekends
and an algorithm pairs me with this guy
am I actually more attracted to him then
I would be to this guy because if I am
that means that attraction is at least a
little bit predictable which means that
match making algorithms could be at
least a little bit helpful I’m a
relationships researcher and for a long
time my field has implicitly assumed
that this was a done deal
of course people like the things they
say they like and when you conduct
hypothetical studies say by presenting
people with dating profiles and asking
them to imagine who they’d pick they do
tend to pick the kinds of things they
say they’d pick if I say I like surfer
puppy guys and you show me these
profiles of course I’m gonna pick the
guy on the left
but then researchers started doing
studies with real potential partners
real matching and things started to get
a little weird because people would have
all these requirements for what they
wanted in a partner they’d have their
preferences and their deal-breakers and
then as soon as you presented them with
a real potential partner all of those
things would go right out the window I
conducted a study where I presented
people with a potential partner who had
at least two of their personal
deal-breakers things they had explicitly
said they did not want in a partner and
can you guess what percentage agreed to
go on the date anyway it was 74%
so studies like these begin to cast some
doubt but they don’t fully test our
question of whether matching is possible
they don’t fully answer the question of
whether attraction is like a comment or
an earthquake my colleagues Paul
Eastwick Eli Finkel and I tried to get
it this question in a more systematic
way recently using speed dating studies
single undergraduate students paired
with each other for a series of four
minute speed dates beforehand they
completed a large package of
questionnaires about their traits and
their preferences these are some
examples we had over a hundred traits
and preferences and then we use machine
learning to see if we could predict who
liked to specifically we used the random
forests method which randomly samples
different combinations of traits and
preferences thousands of times to try to
predict attraction as accurately as
possible so what did we find
well we found we could predict people’s
general tendencies to like others at up
to 18 percent accuracy
so some people are choosier than others
and our models could pick up on that we
could also predict people’s general
tendencies to be liked who’s hot and
who’s not up to 27 percent accuracy so
some people are more attractive than
others and again our models could detect
that then we get to relationship level
attraction and this is the part we
really care about can we predict which
pairs of individuals are particularly
well-suited are well matched and here
our models were completely useless
we could not predict attraction to a
specific person at all in either of our
two studies using over a hundred traits
and preferences so let’s come back to
this on the idea of online dating for a
minute now websites like eHarmony and
chemistry com claim that they can use
algorithms to match people together
scientific algorithms but they haven’t
actually let any scientists see these
algorithms their proprietary knowledge
we can’t share them they say just trust
us okay we can see what they measure the
questionnaires that go into the
algorithms but we don’t know how they
use those measures we don’t know what
the algorithms are is it my self esteem
combined with your outgoingness is it
your spirituality plus my neuroticism
my political views plus your allergy to
bees we don’t know the exciting thing
about machine learning is that it allows
us to test for the first time all of the
combinations of all the traits and
preferences we measured so whichever
combinations work to predict attraction
we should have picked up on it and we
found nothing now is this because this
was an undergraduate sample maybe the
models would have worked with a more
much first sample maybe is it because
people only met each other for a few
minutes and the models could have worked
if we had longer for the attraction to
unfold maybe or maybe attraction is not
like a comet it might not be the case
that we can take one person who has
Tradex and another person who likes
Tradex and boom attraction happens we
might just not know what we’re going to
be drawn to in another person and so we
actually experience that person until we
physically go out and meet that person
attraction might be chaotic like an
earthquake or we just don’t know when or
where it’s going to show up so if we’re
this confused about what we want how are
we ever supposed to find that ideal
person if the things we’re looking for
in the long term have virtually no
bearing on what we’re attracted to in
the short term how are we ever supposed
to select that perfect person well there
might be something a bit off about that
very question because dating websites
and apps have us approaching dating like
a kind of consumerism with the goal
being to select the ideal partner it
studies like mine and many others
suggest that what matters most in a
relationship isn’t so much or
combinations of traits and preferences
rather the experiences that you share
the ways that you behave around each
other and to word each other and so
maybe we shouldn’t be so laser focused
on who the person is worth choosing but
rather the relationship we’re building
are we building the kind of relationship
where we celebrate each other’s
successes support each other’s goals
treat each other with compassion
communicate honestly about our needs and
then strive to meet those needs what
steps are we taking to create the
environment that will nurture the kind
of relationship we want in the words of
my mentor Jeff McDonald maybe the key to
a successful relationship isn’t so much
finding the right partner as it is being
the right partner thank you
you
comments are something we can predict in
1705 a time when many people thought
that comments were supernatural omens
Edmund Hali called his shots he had been
cataloging numerous comments and
concluded that they followed a natural
predictable pattern the same comet was
returning to Earth every 76 years and he
predicted that this comet would return
following a particular orbit in the sky
in 1758 when this prediction was
confirmed it demonstrated the awesome
predictive power that science can hold
earthquakes are something we can’t
predict the year is now 2017 and the
u.s. loses over 4 billion dollars a year
to earthquake damage and sure we know
generally where they’re likely to show
up that’s why this building has seismic
protection
but as Nate Silver discusses in his
recent book earthquakes are chaotic we
still don’t know when or where a
specific earthquake is going to occur
with any useful level of accuracy
what about romantic attraction is
attraction predictable like a comet or
certain combinations of people’s traits
and preferences predictably reliably
results in attraction or as a traction
chaotic like an earthquake where we just
don’t know when or where it’s going to
show up how many people here are
currently on the dating scene it kind of
sucks right
I mean some people really like dating in
the same way that some people really
like shopping you don’t even need to be
looking for anything in particular you
just like to browse but for those of us
type a people who just want to get in
find our perfect partner and get back
out dating can be exhausting you have
these experiences where you’re on a date
with someone who looks great on paper
and you’re turning down the dessert menu
and asking for the check because he’s
been talking about his collection of
fantasy swords
since the soup’s arrived and you’ve yet
to get a word in edgewise or she’s
literally swiping for new matches while
sitting right across from you and if
you’re type a like me you’re thinking
there has got to be a way to streamline
this process do I have to kiss all the
frogs can’t I just jump cut to the part
where I actually connect with someone
that’s the allure of matching algorithms
technology does the hard work for you
you just fill out some questionnaires
about who you are and what you like and
voila
here are some people you’ll actually
click with you get to get in find your
perfect partner and get back out it’s a
really alluring concept so much so that
people are willing to pay a lot of money
for the service but our matchmaking
algorithms actually feasible is
attraction like a comet well when we’re
talking about prediction we have to talk
about precision how much of attraction
can we predict because predicting all of
it with perfect accuracy is a really big
ask unreasonably big but predicting some
of attraction having at least a vague
sense of whether two people will like
each other or not it’s a much smaller
more reasonable ask all that’s required
is that people actually like the things
they say they like
what
[Applause]
what’s they I’m filling out an online
questionnaire and I say that I like men
who like puppies and are clean-shaven
and like to go surfing on the weekends
and an algorithm pairs me with this guy
am I actually more attracted to him then
I would be to this guy because if I am
that means that attraction is at least a
little bit predictable which means that
match making algorithms could be at
least a little bit helpful I’m a
relationships researcher and for a long
time my field has implicitly assumed
that this was a done deal
of course people like the things they
say they like and when you conduct
hypothetical studies say by presenting
people with dating profiles and asking
them to imagine who they’d pick they do
tend to pick the kinds of things they
say they’d pick if I say I like surfer
puppy guys and you show me these
profiles of course I’m gonna pick the
guy on the left
but then researchers started doing
studies with real potential partners
real matching and things started to get
a little weird because people would have
all these requirements for what they
wanted in a partner they’d have their
preferences and their deal-breakers and
then as soon as you presented them with
a real potential partner all of those
things would go right out the window I
conducted a study where I presented
people with a potential partner who had
at least two of their personal
deal-breakers things they had explicitly
said they did not want in a partner and
can you guess what percentage agreed to
go on the date anyway it was 74%
so studies like these begin to cast some
doubt but they don’t fully test our
question of whether matching is possible
they don’t fully answer the question of
whether attraction is like a comment or
an earthquake my colleagues Paul
Eastwick Eli Finkel and I tried to get
it this question in a more systematic
way recently using speed dating studies
single undergraduate students paired
with each other for a series of four
minute speed dates beforehand they
completed a large package of
questionnaires about their traits and
their preferences these are some
examples we had over a hundred traits
and preferences and then we use machine
learning to see if we could predict who
liked to specifically we used the random
forests method which randomly samples
different combinations of traits and
preferences thousands of times to try to
predict attraction as accurately as
possible so what did we find
well we found we could predict people’s
general tendencies to like others at up
to 18 percent accuracy
so some people are choosier than others
and our models could pick up on that we
could also predict people’s general
tendencies to be liked who’s hot and
who’s not up to 27 percent accuracy so
some people are more attractive than
others and again our models could detect
that then we get to relationship level
attraction and this is the part we
really care about can we predict which
pairs of individuals are particularly
well-suited are well matched and here
our models were completely useless
we could not predict attraction to a
specific person at all in either of our
two studies using over a hundred traits
and preferences so let’s come back to
this on the idea of online dating for a
minute now websites like eHarmony and
chemistry com claim that they can use
algorithms to match people together
scientific algorithms but they haven’t
actually let any scientists see these
algorithms their proprietary knowledge
we can’t share them they say just trust
us okay we can see what they measure the
questionnaires that go into the
algorithms but we don’t know how they
use those measures we don’t know what
the algorithms are is it my self esteem
combined with your outgoingness is it
your spirituality plus my neuroticism
my political views plus your allergy to
bees we don’t know the exciting thing
about machine learning is that it allows
us to test for the first time all of the
combinations of all the traits and
preferences we measured so whichever
combinations work to predict attraction
we should have picked up on it and we
found nothing now is this because this
was an undergraduate sample maybe the
models would have worked with a more
much first sample maybe is it because
people only met each other for a few
minutes and the models could have worked
if we had longer for the attraction to
unfold maybe or maybe attraction is not
like a comet it might not be the case
that we can take one person who has
Tradex and another person who likes
Tradex and boom attraction happens we
might just not know what we’re going to
be drawn to in another person and so we
actually experience that person until we
physically go out and meet that person
attraction might be chaotic like an
earthquake or we just don’t know when or
where it’s going to show up so if we’re
this confused about what we want how are
we ever supposed to find that ideal
person if the things we’re looking for
in the long term have virtually no
bearing on what we’re attracted to in
the short term how are we ever supposed
to select that perfect person well there
might be something a bit off about that
very question because dating websites
and apps have us approaching dating like
a kind of consumerism with the goal
being to select the ideal partner it
studies like mine and many others
suggest that what matters most in a
relationship isn’t so much or
combinations of traits and preferences
rather the experiences that you share
the ways that you behave around each
other and to word each other and so
maybe we shouldn’t be so laser focused
on who the person is worth choosing but
rather the relationship we’re building
are we building the kind of relationship
where we celebrate each other’s
successes support each other’s goals
treat each other with compassion
communicate honestly about our needs and
then strive to meet those needs what
steps are we taking to create the
environment that will nurture the kind
of relationship we want in the words of
my mentor Jeff McDonald maybe the key to
a successful relationship isn’t so much
finding the right partner as it is being
the right partner thank you
you
Please follow and like us: