English-Video.net comment policy

The comment field is common to all languages

Let's write in your language and use "Google Translate" together

Please refer to informative community guidelines on TED.com

TEDxMidAtlantic 2013

Jennifer Golbeck: Your social media "likes" expose more than you think

Filmed
Views 2,046,206

Do you like curly fries? Have you Liked them on Facebook? Watch this talk to find out the surprising things Facebook (and others) can guess about you from your random Likes and Shares. Computer scientist Jennifer Golbeck explains how this came about, how some applications of the technology are not so cute -- and why she thinks we should return the control of information to its rightful owners.

- Computer scientist
As the director of the Human-Computer Interaction Lab at the University of Maryland, Jennifer Golbeck studies how people use social media -- and thinks about ways to improve their interactions. Full bio

If you remember that first decade of the web,
00:12
it was really a static place.
00:14
You could go online, you could look at pages,
00:16
and they were put up either by organizations
00:19
who had teams to do it
00:21
or by individuals who were really tech-savvy
00:23
for the time.
00:25
And with the rise of social media
00:27
and social networks in the early 2000s,
00:28
the web was completely changed
00:31
to a place where now the vast majority of content
00:33
we interact with is put up by average users,
00:36
either in YouTube videos or blog posts
00:40
or product reviews or social media postings.
00:42
And it's also become a much more interactive place,
00:46
where people are interacting with others,
00:48
they're commenting, they're sharing,
00:51
they're not just reading.
00:52
So Facebook is not the only place you can do this,
00:54
but it's the biggest,
00:56
and it serves to illustrate the numbers.
00:57
Facebook has 1.2 billion users per month.
00:59
So half the Earth's Internet population
01:02
is using Facebook.
01:04
They are a site, along with others,
01:06
that has allowed people to create an online persona
01:08
with very little technical skill,
01:11
and people responded by putting huge amounts
01:13
of personal data online.
01:15
So the result is that we have behavioral,
01:17
preference, demographic data
01:20
for hundreds of millions of people,
01:22
which is unprecedented in history.
01:24
And as a computer scientist,
what this means is that
01:26
I've been able to build models
01:29
that can predict all sorts of hidden attributes
01:30
for all of you that you don't even know
01:32
you're sharing information about.
01:35
As scientists, we use that to help
01:37
the way people interact online,
01:39
but there's less altruistic applications,
01:41
and there's a problem in that users don't really
01:44
understand these techniques and how they work,
01:46
and even if they did, they don't
have a lot of control over it.
01:49
So what I want to talk to you about today
01:52
is some of these things that we're able to do,
01:53
and then give us some ideas
of how we might go forward
01:56
to move some control back into the hands of users.
01:59
So this is Target, the company.
02:02
I didn't just put that logo
02:03
on this poor, pregnant woman's belly.
02:05
You may have seen this anecdote that was printed
02:07
in Forbes magazine where Target
02:09
sent a flyer to this 15-year-old girl
02:11
with advertisements and coupons
02:13
for baby bottles and diapers and cribs
02:15
two weeks before she told her parents
02:17
that she was pregnant.
02:19
Yeah, the dad was really upset.
02:21
He said, "How did Target figure out
02:24
that this high school girl was pregnant
02:25
before she told her parents?"
02:27
It turns out that they have the purchase history
02:29
for hundreds of thousands of customers
02:32
and they compute what they
call a pregnancy score,
02:34
which is not just whether or
not a woman's pregnant,
02:37
but what her due date is.
02:39
And they compute that
02:41
not by looking at the obvious things,
02:42
like, she's buying a crib or baby clothes,
02:44
but things like, she bought more vitamins
02:46
than she normally had,
02:49
or she bought a handbag
02:51
that's big enough to hold diapers.
02:52
And by themselves, those purchases don't seem
02:54
like they might reveal a lot,
02:56
but it's a pattern of behavior that,
02:59
when you take it in the context
of thousands of other people,
03:01
starts to actually reveal some insights.
03:04
So that's the kind of thing that we do
03:06
when we're predicting stuff
about you on social media.
03:08
We're looking for little
patterns of behavior that,
03:11
when you detect them among millions of people,
03:14
lets us find out all kinds of things.
03:16
So in my lab and with colleagues,
03:19
we've developed mechanisms where we can
03:21
quite accurately predict things
03:22
like your political preference,
03:24
your personality score, gender, sexual orientation,
03:26
religion, age, intelligence,
03:29
along with things like
03:32
how much you trust the people you know
03:34
and how strong those relationships are.
03:36
We can do all of this really well.
03:38
And again, it doesn't come from what you might
03:39
think of as obvious information.
03:41
So my favorite example is from this study
03:44
that was published this year
03:46
in the Proceedings of the National Academies.
03:47
If you Google this, you'll find it.
03:49
It's four pages, easy to read.
03:50
And they looked at just people's Facebook likes,
03:52
so just the things you like on Facebook,
03:55
and used that to predict all these attributes,
03:57
along with some other ones.
03:59
And in their paper they listed the five likes
04:01
that were most indicative of high intelligence.
04:04
And among those was liking a page
04:07
for curly fries. (Laughter)
04:09
Curly fries are delicious,
04:11
but liking them does not necessarily mean
04:13
that you're smarter than the average person.
04:15
So how is it that one of the strongest indicators
04:17
of your intelligence
04:21
is liking this page
04:22
when the content is totally irrelevant
04:24
to the attribute that's being predicted?
04:26
And it turns out that we have to look at
04:28
a whole bunch of underlying theories
04:30
to see why we're able to do this.
04:32
One of them is a sociological
theory called homophily,
04:34
which basically says people are
friends with people like them.
04:37
So if you're smart, you tend to
be friends with smart people,
04:40
and if you're young, you tend
to be friends with young people,
04:42
and this is well established
04:45
for hundreds of years.
04:46
We also know a lot
04:48
about how information spreads through networks.
04:49
It turns out things like viral videos
04:52
or Facebook likes or other information
04:54
spreads in exactly the same way
04:56
that diseases spread through social networks.
04:58
So this is something we've studied for a long time.
05:01
We have good models of it.
05:02
And so you can put those things together
05:04
and start seeing why things like this happen.
05:06
So if I were to give you a hypothesis,
05:09
it would be that a smart guy started this page,
05:11
or maybe one of the first people who liked it
05:14
would have scored high on that test.
05:16
And they liked it, and their friends saw it,
05:18
and by homophily, we know that
he probably had smart friends,
05:20
and so it spread to them,
and some of them liked it,
05:23
and they had smart friends,
05:26
and so it spread to them,
05:28
and so it propagated through the network
05:28
to a host of smart people,
05:30
so that by the end, the action
05:33
of liking the curly fries page
05:35
is indicative of high intelligence,
05:37
not because of the content,
05:39
but because the actual action of liking
05:41
reflects back the common attributes
05:43
of other people who have done it.
05:45
So this is pretty complicated stuff, right?
05:48
It's a hard thing to sit down and explain
05:51
to an average user, and even if you do,
05:53
what can the average user do about it?
05:56
How do you know that
you've liked something
05:58
that indicates a trait for you
06:00
that's totally irrelevant to the
content of what you've liked?
06:01
There's a lot of power that users don't have
06:05
to control how this data is used.
06:08
And I see that as a real
problem going forward.
06:10
So I think there's a couple paths
06:13
that we want to look at
06:15
if we want to give users some control
06:16
over how this data is used,
06:18
because it's not always going to be used
06:20
for their benefit.
06:21
An example I often give is that,
06:23
if I ever get bored being a professor,
06:24
I'm going to go start a company
06:26
that predicts all of these attributes
06:28
and things like how well you work in teams
06:29
and if you're a drug user, if you're an alcoholic.
06:31
We know how to predict all that.
06:33
And I'm going to sell reports
06:35
to H.R. companies and big businesses
06:36
that want to hire you.
06:39
We totally can do that now.
06:41
I could start that business tomorrow,
06:42
and you would have
absolutely no control
06:44
over me using your data like that.
06:46
That seems to me to be a problem.
06:48
So one of the paths we can go down
06:50
is the policy and law path.
06:52
And in some respects, I think
that that would be most effective,
06:54
but the problem is we'd
actually have to do it.
06:57
Observing our political process in action
07:00
makes me think it's highly unlikely
07:03
that we're going to get a bunch of representatives
07:05
to sit down, learn about this,
07:07
and then enact sweeping changes
07:09
to intellectual property law in the U.S.
07:11
so users control their data.
07:13
We could go the policy route,
07:16
where social media companies say,
07:17
you know what? You own your data.
07:18
You have total control over how it's used.
07:20
The problem is that the revenue models
07:22
for most social media companies
07:24
rely on sharing or exploiting
users' data in some way.
07:26
It's sometimes said of Facebook that the users
07:30
aren't the customer, they're the product.
07:32
And so how do you get a company
07:34
to cede control of their main asset
07:37
back to the users?
07:39
It's possible, but I don't think it's something
07:41
that we're going to see change quickly.
07:42
So I think the other path
07:45
that we can go down that's
going to be more effective
07:46
is one of more science.
07:48
It's doing science that allowed us to develop
07:50
all these mechanisms for computing
07:52
this personal data in the first place.
07:54
And it's actually very similar research
07:56
that we'd have to do
07:58
if we want to develop mechanisms
08:00
that can say to a user,
08:02
"Here's the risk of that action you just took."
08:04
By liking that Facebook page,
08:06
or by sharing this piece of personal information,
08:08
you've now improved my ability
08:10
to predict whether or not you're using drugs
08:12
or whether or not you get
along well in the workplace.
08:14
And that, I think, can affect whether or not
08:17
people want to share something,
08:19
keep it private, or just keep it offline altogether.
08:20
We can also look at things like
08:24
allowing people to encrypt data that they upload,
08:25
so it's kind of invisible and worthless
08:28
to sites like Facebook
08:30
or third party services that access it,
08:31
but that select users who the person who posted it
08:34
want to see it have access to see it.
08:37
This is all super exciting research
08:40
from an intellectual perspective,
08:42
and so scientists are going to be willing to do it.
08:43
So that gives us an advantage over the law side.
08:45
One of the problems that people bring up
08:49
when I talk about this is, they say,
08:51
you know, if people start
keeping all this data private,
08:52
all those methods that you've been developing
08:55
to predict their traits are going to fail.
08:57
And I say, absolutely, and for me, that's success,
09:00
because as a scientist,
09:03
my goal is not to infer information about users,
09:05
it's to improve the way people interact online.
09:09
And sometimes that involves
inferring things about them,
09:11
but if users don't want me to use that data,
09:15
I think they should have the right to do that.
09:18
I want users to be informed and consenting
09:20
users of the tools that we develop.
09:22
And so I think encouraging this kind of science
09:24
and supporting researchers
09:27
who want to cede some of that control back to users
09:29
and away from the social media companies
09:32
means that going forward, as these tools evolve
09:34
and advance,
09:37
means that we're going to have an educated
09:38
and empowered user base,
09:40
and I think all of us can agree
09:41
that that's a pretty ideal way to go forward.
09:42
Thank you.
09:45
(Applause)
09:47

▲Back to top

About the speaker:

Jennifer Golbeck - Computer scientist
As the director of the Human-Computer Interaction Lab at the University of Maryland, Jennifer Golbeck studies how people use social media -- and thinks about ways to improve their interactions.

Why you should listen

Jennifer Golbeck is an associate professor in the College of Information Studies at the University of Maryland, where she also moonlights in the department of computer science. Her work invariably focuses on how to enhance and improve the way that people interact with their own information online. "I approach this from a computer science perspective and my general research hits social networks, trust, web science, artificial intelligence, and human-computer interaction," she writes.

Author of the 2013 book, Analyzing the Social Web, Golbeck likes nothing more than to immerse herself in the inner workings of the Internet tools so many millions of people use daily, to understand the implications of our choices and actions. Recently, she has also been working to bring human-computer interaction ideas to the world of security and privacy systems.

More profile about the speaker
Jennifer Golbeck | Speaker | TED.com