Jennifer Golbeck: The curly fry conundrum: Why social media "likes" say more than you might think

Translator: Reviewer: Anders Finn Jørgensen

If you remember that first decade of the web,

Hvis du kan huske det første årti af internettet

it was really a static place.

det var virkelig et stillestående sted.

You could go online, you could look at pages,

and they were put up either by organizations

og de var sat op af enten organisationer

who had teams to do it

som havde teams til at gøre det

or by individuals who were really tech-savvy

eller af individer der var super teknologihabile

for the time.

på den tid.

And with the rise of social media

Med fremkomsten af de sociale medier

and social networks in the early 2000s,

og sociale netværk i begyndelsen af 2000'erne

the web was completely changed

ændrede internettet sig fuldstændigt

to a place where now the vast majority of content

til et sted hvor nu størstedelen af indholdet

we interact with is put up by average users,

vi interagere med er lagt ud af almindelige brugere

enten som YouTubevideoer eller blogs

either in YouTube videos or blog posts

or product reviews or social media postings.

eller som produktanmeldelse eller opdateringer på sociale medier.

And it's also become a much more interactive place,

Det blev også et meget mere interaktivt sted,

where people are interacting with others,

hvor mennesker interagere med hinanden,

they're commenting, they're sharing,

de kommenterer, de deler

they're not just reading.

de læser ikke bare.

So Facebook is not the only place you can do this,

Så Facebook er ikke det eneste sted du kan gøre dette,

but it's the biggest,

men det er det største,

and it serves to illustrate the numbers.

og det tjener til at illustrere tallene.

Facebook has 1.2 billion users per month.

Facebook har 1.2 billioner brugere per måned.

So half the Earth's Internet population

Det er så halvdelen af Jordens internetbrugere

is using Facebook.

der benytter Facebook.

They are a site, along with others,

Det er en side, som sammen med andre,

that has allowed people to create an online persona

har givet personer mulighed for at opbygge en online personlighed

with very little technical skill,

med meget lidt teknologisk færdighed

and people responded by putting huge amounts

og mennesker reagerede ved at putte kæmpe mængder

of personal data online.

personlig data online.

So the result is that we have behavioral,

Så resultat er at vi har adfærdsmæssige,

preference, demographic data

præferencemæssige og demografisk data

for hundreds of millions of people,

for hundrede af millioner mennesker,

which is unprecedented in history.

Hvilket er hidtil uset i historien.

And as a computer scientist, what this means is that

Og som computerforsker, hvad det betyder er,

I've been able to build models

at jeg er blevet i stand til at bygge modeller

that can predict all sorts of hidden attributes

der kan forudse alle mulige skjulte egenskaber

for all of you that you don't even know

for alle jer, som jeg ikke engang kender,

you're sharing information about.

som i deler information omkring.

As scientists, we use that to help

Som forsker, bruger vi det til at hjælpe

the way people interact online,

den måde mennesker interagere online,

but there's less altruistic applications,

men der er mere egoistiske anvendelser,

and there's a problem in that users don't really

og der er et problem i at brugere ikke rigtigt

understand these techniques and how they work,

forstår disse teknikker og hvordan de virker,

and even if they did, they don't have a lot of control over it.

og selv hvis de gjorde, har de ikke meget kontrol over det.

So what I want to talk to you about today

Så hvad jeg gerne vil tale om i dag

is some of these things that we're able to do,

er nogle af de ting som vi er istand til at gøre

and then give us some ideas of how we might go forward

og så give jer nogle ideer om hvordan vi nok vil tage fat

to move some control back into the hands of users.

for at flytte noget af kontrollen tilbage til brugerne.

So this is Target, the company.

Det her er firmaet Target.

I didn't just put that logo

jeg har ikke blot sat dette logo

on this poor, pregnant woman's belly.

på denne fattige, gravide dames mave.

You may have seen this anecdote that was printed

Du har måske set denne anekdote der var trykket

in Forbes magazine where Target

i Forbes magasin hvor Target

sent a flyer to this 15-year-old girl

sendte en flyer til denne 15-årige pige

with advertisements and coupons

med reklamer og kuponer

for baby bottles and diapers and cribs

til babyflasker og bleer og vugger

two weeks before she told her parents

to uger før hun fortalte hendes forældre

that she was pregnant.

at hun var gravid.

Yeah, the dad was really upset.

Ja, faren var virkelig oprevet.

He said, "How did Target figure out

Han sagde, "hvordan fandt Target ud af

that this high school girl was pregnant

at denne gymnasiepige var gravid

before she told her parents?"

før hun fortalte det til hendes forældre?"

It turns out that they have the purchase history

Det viser sig, at de har købshistorikken

for hundreds of thousands of customers

for hundredetusinde kunder

and they compute what they call a pregnancy score,

og de beregner hvad de kalder en graviditetsscore,

which is not just whether or not a woman's pregnant,

der ikke bare er hvorvidt en kvinde er gravid eller ej,

but what her due date is.

men hvornår hendes termin er.

And they compute that

De beregner dette

not by looking at the obvious things,

ikke ved at kigge på de åbenlyse ting,

like, she's buying a crib or baby clothes,

som, hun køber en vugge eller baby tøj,

but things like, she bought more vitamins

men ting som, hun købte flere vitaminer

than she normally had,

end hun normalt gjorde,

or she bought a handbag

eller hun købte en håndtaske

that's big enough to hold diapers.

der er stor nok til at indeholde bleer.

And by themselves, those purchases don't seem

For i sig selv, virker disse køb ikke til

like they might reveal a lot,

at de ville afsløre en masse,

but it's a pattern of behavior that,

men der er et adfærdsmønster som,

when you take it in the context of thousands of other people,

når du tager det i konteksten af flere tusinde af mennesker,

starts to actually reveal some insights.

rent faktisk begynder at give noget indsigt.

So that's the kind of thing that we do

Så det er den slags ting vi gør

when we're predicting stuff about you on social media.

når vi forudsiger ting omkring dig på de sociale medier.

We're looking for little patterns of behavior that,

Vi leder efter små adfærdsmønster der,

when you detect them among millions of people,

når du opdager dem imellem millioner af mennesker,

lets us find out all kinds of things.

lader os finde ud af alle mulige ting

So in my lab and with colleagues,

Så i mit laboratorie med kollegaer,

we've developed mechanisms where we can

har vi udviklet mekanismer hvor vi

quite accurately predict things

rimelig præcist kan forudsige ting

like your political preference,

som din politiske præference,

your personality score, gender, sexual orientation,

din personlighedscore, køn, seksuelle orientering,

religion, age, intelligence,

religion, alder, intelligens,

along with things like

sammen med ting som

how much you trust the people you know

hvor meget du stoler på de mennesker du kender

and how strong those relationships are.

og hvor stærke disse forhold er.

We can do all of this really well.

Vi kan gøre alt dette ret godt.

And again, it doesn't come from what you might

Og igen, det kommer ikke fra hvad du måske

think of as obvious information.

tror er den åbenlyse information.

So my favorite example is from this study

Så mit favorit eksempel fra en undersøgelse

that was published this year

der blev udgivet dette år

in the Proceedings of the National Academies.

i "Proceedings of the National Academies"

If you Google this, you'll find it.

Hvis du Googler dette, finder du det.

It's four pages, easy to read.

Det er fire sider, let at læse.

And they looked at just people's Facebook likes,

og de kiggede bare på folks Facebook likes,

so just the things you like on Facebook,

så bare de ting du kan lide på Facebook,

and used that to predict all these attributes,

og brugte dette til at forudsige disse attributter.

along with some other ones.

sammen med nogle yderligere.

And in their paper they listed the five likes

I deres artikel oplistede de fem likes

that were most indicative of high intelligence.

der mest indikerede en høj intelligens.

And among those was liking a page

Iblandt dem var at like en side

for curly fries. (Laughter)

for curly fries. (latter)

Curly fries are delicious,

Curly fries er lækre,

but liking them does not necessarily mean

men at kunne lide dem betyder ikke nødvendigvis,

that you're smarter than the average person.

at du er klogere end den gennemsnitlige person.

So how is it that one of the strongest indicators

Så hvordan kan det være at en af de stærkeste indikatorer

of your intelligence

på din intelligens

is liking this page

er at like denne side

when the content is totally irrelevant

når indholdet er totalt irrelevant

to the attribute that's being predicted?

for de attributter der bliver forudsagt?

And it turns out that we have to look at

Og det viser sig, at vi er nødt til at se på

a whole bunch of underlying theories

en hel bunke underliggende teorier

to see why we're able to do this.

for at se hvorfor vi er i stand til det.

One of them is a sociological theory called homophily,

En af dem er en sociologisk teori kaldet homofili

which basically says people are friends with people like them.

som dybest set siger, at folk er venner med folk som ligner dem.

So if you're smart, you tend to be friends with smart people,

Så hvis du er klog, har du en tendens til, at være venner med kloge mennesker

and if you're young, you tend to be friends with young people,

og hvis du er ung, tendere du til at være venner med unge mennesker,

and this is well established

og dette er veletableret

for hundreds of years.

i hundrede af år.

We also know a lot

Vi ved også en masse

about how information spreads through networks.

omkring hvordan information spreder sig igennem netværk.

It turns out things like viral videos

Det viser sig at ting som virale videoer

or Facebook likes or other information

eller Facebook likes eller anden information

spreads in exactly the same way

spreder sig på nøjagtig samme måde

that diseases spread through social networks.

som sygdom spredes gennem sociale netværk.

So this is something we've studied for a long time.

Dette er noget vi har studeret i lang tid.

We have good models of it.

Vi har gode modeller af det.

And so you can put those things together

og så kan du sætte de ting sammen

and start seeing why things like this happen.

og begynde at se hvorfor ting som dette sker

So if I were to give you a hypothesis,

Hvis jeg skulle give jer en hypotese,

it would be that a smart guy started this page,

ville det være at en klog fyr startede denne side,

or maybe one of the first people who liked it

eller måske en af de første personer der likede den

would have scored high on that test.

ville have scoret højere på den test.

And they liked it, and their friends saw it,

De likede den, og deres venner så det,

and by homophily, we know that he probably had smart friends,

og via homofili, ved vi, at han højst sandsynlig havde kloge venner,

and so it spread to them, and some of them liked it,

så det spreder sig til dem, og nogle af dem likede det,

and they had smart friends,

og de havde kloge venner,

and so it spread to them,

and so it propagated through the network

og så spredte det sig til dem,

og sådan forplantede det sig igennem netværket

to a host of smart people,

til et væld af kloge personer,

so that by the end, the action

så i sidste ende, handlingen

of liking the curly fries page

af at like curly fries siden

is indicative of high intelligence,

er indikator af høj intelligens

not because of the content,

ikke pga. indholdet

but because the actual action of liking

men fordi den faktiske handling af at like

reflects back the common attributes

afspejler den fælles attribut

of other people who have done it.

af de mennesker som har gjort det.

So this is pretty complicated stuff, right?

Det er en rimeligt kompliceret ting, ikke?

It's a hard thing to sit down and explain

det er svært at sætte sig ned og forklare

to an average user, and even if you do,

til en gennemsnitlig bruger, og hvis du gør det,

what can the average user do about it?

hvad kan brugeren gøre ved det?

How do you know that you've liked something

Hvordan ved du at du likede noget

that indicates a trait for you

der indikere et træk ved dig

that's totally irrelevant to the content of what you've liked?

der er totalt irrelevant for det indhold af det du likede?

There's a lot of power that users don't have

Der er en stor magt som brugere ikke

to control how this data is used.

har til at kontrollere hvordan denne data er brugt.

And I see that as a real problem going forward.

Og jeg ser det som et reelt problem fremadrettet

So I think there's a couple paths

Så jeg tror der er et par veje

that we want to look at

som vi skal se på

if we want to give users some control

hvis vi vil give brugerne noget kontrol

over how this data is used,

omkring hvordan denne data er brugt,

because it's not always going to be used

fordi det ikke altid bliver brugt

for their benefit.

til deres fordel.

An example I often give is that,

Et eksempel jeg tit giver er, at

if I ever get bored being a professor,

hvis jeg nogensinde blev træt af at være professor

I'm going to go start a company

vil jeg starte et firma

that predicts all of these attributes

der forudsiger alle disse attributter

and things like how well you work in teams

og ting som hvor godt du arbejder i teams

and if you're a drug user, if you're an alcoholic.

og om du er stofmisbruger, om du er alkoholiker

We know how to predict all that.

Vi ved hvordan vi kan forudsige alt dette.

And I'm going to sell reports

og jeg vil sælge nogle rapporter

to H.R. companies and big businesses

til HR firmaer og store virksomheder

that want to hire you.

der vil hyre dig.

We totally can do that now.

Det kan vi altså gøre nu.

I could start that business tomorrow,

Jeg kunne starte dette firma imorgen,

and you would have absolutely no control

og du ville have absolut ingen kontrol

over me using your data like that.

over at jeg bruger dine data som det.

That seems to me to be a problem.

Det virker for mig som et problem.

So one of the paths we can go down

Så en af de veje vi kan gå

is the policy and law path.

er politik og lovvejen

And in some respects, I think that that would be most effective,

Og i nogle henseende, tror jeg at dette ville være mest effektivt,

but the problem is we'd actually have to do it.

men problemet er at vi faktisk ville skulle gøre det.

Observing our political process in action

Ved observering af vores aktive politiske processer

makes me think it's highly unlikely

får det mig til at tro at det er meget usandsynligt

that we're going to get a bunch of representatives

at vi kan få en bunke repræsentanter

to sit down, learn about this,

til at sidde sig ned, lære om dette,

and then enact sweeping changes

og så vedtage gennemgribende ændringer

to intellectual property law in the U.S.

for den intellektuelle ejendomsrets lov i USA

so users control their data.

så brugerne kan kontrollere deres data.

We could go the policy route,

Vi kunne gå politik vejen,

where social media companies say,

hvor sociale medie virksomheder siger,

you know what? You own your data.

ved du hvad? Du ejer din egne data.

You have total control over how it's used.

Du har total kontrol over hvordan det bliver brugt.

The problem is that the revenue models

Problemet er, at indtægtsmodellerne

for most social media companies

for de fleste sociale medier

rely on sharing or exploiting users' data in some way.

afhænger af deling eller udnyttelse af brugers data.

It's sometimes said of Facebook that the users

Det er nogle gange sagt af Facebook at brugerne

aren't the customer, they're the product.

ikke er forbrugerne, de er produktet.

And so how do you get a company

Så hvordan får du et firma

to cede control of their main asset

til at give kontrollen af deres primære aktiv

back to the users?

tilbage til brugerne?

It's possible, but I don't think it's something

Det er muligt, men jeg tror ikke det er noget

that we're going to see change quickly.

som vi vil se ske hurtigt.

So I think the other path

Så jeg tror den anden vej

that we can go down that's going to be more effective

som vi kan gå ned af som vil være mere effektiv

is one of more science.

er en af mere videnskab.

It's doing science that allowed us to develop

Det er at lave forsøg der tillod os at udvikle

all these mechanisms for computing

de mekanismer til beregning

this personal data in the first place.

af personlig data i første omgang.

And it's actually very similar research

Og det er faktisk meget lignende forskning

that we'd have to do

som vi ville skulle lave

if we want to develop mechanisms

hvis vi ville udvikle mekanismer

that can say to a user,

der kan fortælle en bruger,

"Here's the risk of that action you just took."

"her er risikoen ved det du lige har foretaget dig"

By liking that Facebook page,

Ved at like den Facebook sige,

or by sharing this piece of personal information,

eller ved at dele dette stykke personlige information,

you've now improved my ability

har du forbedret min evne

to predict whether or not you're using drugs

til at forudsige hvorvidt du bruger stoffer

or whether or not you get along well in the workplace.

eller om du ikke kommer godt ud af det på arbejdspladsen.

And that, I think, can affect whether or not

og det, tror jeg, kan påvirker hvorvidt

people want to share something,

folk vil dele noget,

keep it private, or just keep it offline altogether.

holde det privat, eller bare holde det offline.

We can also look at things like

Vi kan også se på ting som

allowing people to encrypt data that they upload,

give folk lov til at kryptere data de uploader,

so it's kind of invisible and worthless

så det bliver usynligt og værdiløst

to sites like Facebook

for sites som Facebook

or third party services that access it,

eller tredjeparts servicer der har adgang til det

but that select users who the person who posted it

men som udvalgte brugere som personen der postede det

want to see it have access to see it.

vil have til at se det - kan få adgang til det.

This is all super exciting research

Dette er superspændende forskning

from an intellectual perspective,

fra et intellektuelt perspektiv,

and so scientists are going to be willing to do it.

så forskere vil være villige til at gøre det.

So that gives us an advantage over the law side.

Så det giver os en fordel overfor den lovmæssige side.

One of the problems that people bring up

Et af de problemer folk bringer på banen

when I talk about this is, they say,

når jeg taler om dette er, siger de

you know, if people start keeping all this data private,

du ved, hvis folk begynder at holde alle disse data private,

all those methods that you've been developing

alle disse metoder som du har udviklet

to predict their traits are going to fail.

til at forudsige deres evner vil mislykkes.

And I say, absolutely, and for me, that's success,

og jeg siger, absolut, og for mig, er det succes.

because as a scientist,

Fordi som forsker,

my goal is not to infer information about users,

er mit mål ikke at udlede information om brugere,

it's to improve the way people interact online.

det er at forbedre den måde folk interagere online.

And sometimes that involves inferring things about them,

og nogle gange involvere det udledning af ting om dem

but if users don't want me to use that data,

men hvis brugere ikke vil have mig til at bruge de data

I think they should have the right to do that.

tror jeg de skulle have retten til at gøre det.

I want users to be informed and consenting

Jeg vil have informerede og samtykkende

users of the tools that we develop.

brugerne af de værktøjer vi udvikler.

And so I think encouraging this kind of science

Så jeg tror at opfordre til denne form for videnskab

and supporting researchers

og støtte forskere

who want to cede some of that control back to users

som vil tage noget af denne kontrol tilbage til brugerne

and away from the social media companies

og væk fra de sociale medie virksomheder

means that going forward, as these tools evolve

betyder at gå fremad, som disse værktøjer udvikles

and advance,

og bliver bedre,

means that we're going to have an educated

vil vi være nødt til at have en veluddannet

and empowered user base,

og selvstændig brugerbase

and I think all of us can agree

og jeg tror at vi alle kan blive enige om

that that's a pretty ideal way to go forward.

at det er en ret ideel måde at gå fremad på.

Thank you.

Tak.

(Bifald)

(Applause)