WEBVTT

00:00:00.792 --> 00:00:03.059
Every day, every week,

00:00:03.083 --> 00:00:05.268
we agree to terms and conditions.

00:00:05.292 --> 00:00:06.768
And when we do this,

00:00:06.792 --> 00:00:09.268
we provide companies with the lawful right

00:00:09.292 --> 00:00:12.976
to do whatever they want with our data

00:00:13.000 --> 00:00:15.375
and with the data of our children.

00:00:16.792 --> 00:00:19.768
Which makes us wonder:

00:00:19.792 --> 00:00:22.684
how much data are we giving
away of children,

00:00:22.708 --> 00:00:24.708
and what are its implications?

NOTE Paragraph

00:00:26.500 --> 00:00:27.893
I'm an anthropologist,

00:00:27.917 --> 00:00:30.518
and I'm also the mother
of two little girls.

00:00:30.542 --> 00:00:35.018
And I started to become interested
in this question in 2015

00:00:35.042 --> 00:00:37.768
when I suddenly realized
that there were vast --

00:00:37.792 --> 00:00:40.809
almost unimaginable amounts of data traces

00:00:40.833 --> 00:00:44.000
that are being produced
and collected about children.

00:00:44.792 --> 00:00:46.768
So I launched a research project,

00:00:46.792 --> 00:00:49.268
which is called Child Data Citizen,

00:00:49.292 --> 00:00:51.417
and I aimed at filling in the blank.

NOTE Paragraph

00:00:52.583 --> 00:00:55.601
Now you may think
that I'm here to blame you

00:00:55.625 --> 00:00:58.393
for posting photos
of your children on social media,

00:00:58.417 --> 00:01:00.559
but that's not really the point.

00:01:00.583 --> 00:01:04.000
The problem is way bigger
than so-called "sharenting."

00:01:04.792 --> 00:01:08.893
This is about systems, not individuals.

00:01:08.917 --> 00:01:11.208
You and your habits are not to blame.

NOTE Paragraph

00:01:12.833 --> 00:01:15.684
For the very first time in history,

00:01:15.708 --> 00:01:18.268
we are tracking
the individual data of children

00:01:18.292 --> 00:01:20.059
from long before they're born --

00:01:20.083 --> 00:01:22.768
sometimes from the moment of conception,

00:01:22.792 --> 00:01:25.143
and then throughout their lives.

00:01:25.167 --> 00:01:28.268
You see, when parents decide to conceive,

00:01:28.292 --> 00:01:31.268
they go online to look
for "ways to get pregnant,"

00:01:31.292 --> 00:01:34.042
or they download ovulation-tracking apps.

00:01:35.250 --> 00:01:37.851
When they do get pregnant,

00:01:37.875 --> 00:01:41.018
they post ultrasounds
of their babies on social media,

00:01:41.042 --> 00:01:43.059
they download pregnancy apps

00:01:43.083 --> 00:01:46.809
or they consult Dr. Google
for all sorts of things,

00:01:46.833 --> 00:01:48.351
like, you know --

00:01:48.375 --> 00:01:50.934
for "miscarriage risk when flying"

00:01:50.958 --> 00:01:53.726
or "abdominal cramps in early pregnancy."

00:01:53.750 --> 00:01:55.559
I know because I've done it --

00:01:55.583 --> 00:01:57.208
and many times.

00:01:58.458 --> 00:02:01.268
And then, when the baby is born,
they track every nap,

00:02:01.292 --> 00:02:02.559
every feed,

00:02:02.583 --> 00:02:05.167
every life event
on different technologies.

00:02:06.083 --> 00:02:07.559
And all of these technologies

00:02:07.583 --> 00:02:13.726
transform the baby's most intimate
behavioral and health data into profit

00:02:13.750 --> 00:02:15.542
by sharing it with others.

NOTE Paragraph

00:02:16.583 --> 00:02:18.726
So to give you an idea of how this works,

00:02:18.750 --> 00:02:23.934
in 2019, the British Medical Journal
published research that showed

00:02:23.958 --> 00:02:27.601
that out of 24 mobile health apps,

00:02:27.625 --> 00:02:31.083
19 shared information with third parties.

00:02:32.083 --> 00:02:37.917
And these third parties shared information
with 216 other organizations.

00:02:38.875 --> 00:02:42.309
Of these 216 other fourth parties,

00:02:42.333 --> 00:02:45.476
only three belonged to the health sector.

00:02:45.500 --> 00:02:50.018
The other companies that had access
to that data were big tech companies

00:02:50.042 --> 00:02:53.559
like Google, Facebook or Oracle,

00:02:53.583 --> 00:02:56.184
they were digital advertising companies

00:02:56.208 --> 00:03:00.333
and there was also
a consumer credit reporting agency.

00:03:01.125 --> 00:03:02.559
So you get it right:

00:03:02.583 --> 00:03:07.708
ad companies and credit agencies may
already have data points on little babies.

00:03:09.125 --> 00:03:11.893
But mobile apps,
web searches and social media

00:03:11.917 --> 00:03:15.018
are really just the tip of the iceberg,

00:03:15.042 --> 00:03:17.893
because children are being tracked
by multiple technologies

00:03:17.917 --> 00:03:19.643
in their everyday lives.

00:03:19.667 --> 00:03:23.809
They're tracked by home technologies
and virtual assistants in their homes.

00:03:23.833 --> 00:03:25.809
They're tracked by educational platforms

00:03:25.833 --> 00:03:28.018
and educational technologies
in their schools.

00:03:28.042 --> 00:03:29.643
They're tracked by online records

00:03:29.667 --> 00:03:32.684
and online portals
at their doctor's office.

00:03:32.708 --> 00:03:35.059
They're tracked by their
internet-connected toys,

00:03:35.083 --> 00:03:36.393
their online games

00:03:36.417 --> 00:03:39.083
and many, many, many,
many other technologies.

NOTE Paragraph

00:03:40.250 --> 00:03:41.893
So during my research,

00:03:41.917 --> 00:03:46.059
a lot of parents came up to me
and they were like, "So what?

00:03:46.083 --> 00:03:49.000
Why does it matter
if my children are being tracked?

00:03:50.042 --> 00:03:51.375
We've got nothing to hide."

00:03:52.958 --> 00:03:54.458
Well, it matters.

00:03:55.083 --> 00:04:01.101
It matters because today individuals
are not only being tracked,

00:04:01.125 --> 00:04:05.226
they're also being profiled
on the basis of their data traces.

00:04:05.250 --> 00:04:09.059
Artificial intelligence and predictive
analytics are being used

00:04:09.083 --> 00:04:12.726
to harness as much data as possible
of an individual life

00:04:12.750 --> 00:04:14.601
from different sources:

00:04:14.625 --> 00:04:19.143
family history, purchasing habits,
social media comments.

00:04:19.167 --> 00:04:21.018
And then they bring this data together

00:04:21.042 --> 00:04:23.792
to make data-driven decisions
about the individual.

00:04:24.792 --> 00:04:28.226
And these technologies
are used everywhere.

00:04:28.250 --> 00:04:30.643
Banks use them to decide loans.

00:04:30.667 --> 00:04:33.042
Insurance uses them to decide premiums.

00:04:34.208 --> 00:04:36.684
Recruiters and employers use them

00:04:36.708 --> 00:04:39.625
to decide whether one
is a good fit for a job or not.

00:04:40.750 --> 00:04:43.851
Also the police and courts use them

00:04:43.875 --> 00:04:47.393
to determine whether one
is a potential criminal

00:04:47.417 --> 00:04:50.042
or is likely to recommit a crime.

NOTE Paragraph

00:04:52.458 --> 00:04:56.518
We have no knowledge or control

00:04:56.542 --> 00:05:00.184
over the ways in which those who buy,
sell and process our data

00:05:00.208 --> 00:05:02.917
are profiling us and our children.

00:05:03.625 --> 00:05:07.667
But these profiles can come to impact
our rights in significant ways.

NOTE Paragraph

00:05:08.917 --> 00:05:11.125
To give you an example,

00:05:13.792 --> 00:05:17.851
in 2018 the "New York Times"
published the news

00:05:17.875 --> 00:05:19.851
that the data that had been gathered

00:05:19.875 --> 00:05:22.934
through online
college-planning services --

00:05:22.958 --> 00:05:27.684
that are actually completed by millions
of high school kids across the US

00:05:27.708 --> 00:05:31.351
who are looking for a college
program or a scholarship --

00:05:31.375 --> 00:05:34.417
had been sold to educational data brokers.

00:05:35.792 --> 00:05:41.226
Now, researchers at Fordham
who studied educational data brokers

00:05:41.250 --> 00:05:46.476
revealed that these companies
profiled kids as young as two

00:05:46.500 --> 00:05:49.559
on the basis of different categories:

00:05:49.583 --> 00:05:53.768
ethnicity, religion, affluence,

00:05:53.792 --> 00:05:55.851
social awkwardness

00:05:55.875 --> 00:05:58.809
and many other random categories.

00:05:58.833 --> 00:06:03.851
And then they sell these profiles
together with the name of the kid,

00:06:03.875 --> 00:06:06.684
their home address and the contact details

00:06:06.708 --> 00:06:08.559
to different companies,

00:06:08.583 --> 00:06:11.042
including trade and career institutions,

00:06:12.083 --> 00:06:13.351
student loans

00:06:13.375 --> 00:06:15.125
and student credit card companies.

00:06:16.542 --> 00:06:17.893
To push the boundaries,

00:06:17.917 --> 00:06:21.726
the researchers at Fordham
asked an educational data broker

00:06:21.750 --> 00:06:27.559
to provide them with a list
of 14-to-15-year-old girls

00:06:27.583 --> 00:06:30.958
who were interested
in family planning services.

00:06:32.208 --> 00:06:34.684
The data broker agreed
to provide them the list.

00:06:34.708 --> 00:06:39.583
So imagine how intimate
and how intrusive that is for our kids.

00:06:40.833 --> 00:06:44.809
But educational data brokers
are really just an example.

00:06:44.833 --> 00:06:49.518
The truth is that our children are being
profiled in ways that we cannot control

00:06:49.542 --> 00:06:52.958
but that can significantly impact
their chances in life.

NOTE Paragraph

00:06:54.167 --> 00:06:57.643
So we need to ask ourselves:

00:06:57.667 --> 00:07:02.351
can we trust these technologies
when it comes to profiling our children?

00:07:02.375 --> 00:07:03.625
Can we?

00:07:05.708 --> 00:07:06.958
My answer is no.

00:07:07.792 --> 00:07:09.059
As an anthropologist,

00:07:09.083 --> 00:07:12.851
I believe that artificial intelligence
and predictive analytics can be great

00:07:12.875 --> 00:07:14.893
to predict the course of a disease

00:07:14.917 --> 00:07:16.750
or to fight climate change.

00:07:18.000 --> 00:07:19.643
But we need to abandon the belief

00:07:19.667 --> 00:07:23.351
that these technologies
can objectively profile humans

00:07:23.375 --> 00:07:26.559
and that we can rely on them
to make data-driven decisions

00:07:26.583 --> 00:07:28.476
about individual lives.

00:07:28.500 --> 00:07:31.059
Because they can't profile humans.

00:07:31.083 --> 00:07:34.434
Data traces are not
the mirror of who we are.

00:07:34.458 --> 00:07:36.559
Humans think one thing
and say the opposite,

00:07:36.583 --> 00:07:39.018
feel one way and act differently.

00:07:39.042 --> 00:07:41.518
Algorithmic predictions
or our digital practices

00:07:41.542 --> 00:07:46.708
cannot account for the unpredictability
and complexity of human experience.

NOTE Paragraph

00:07:48.417 --> 00:07:49.976
But on top of that,

00:07:50.000 --> 00:07:52.684
these technologies are always --

00:07:52.708 --> 00:07:53.976
always --

00:07:54.000 --> 00:07:55.917
in one way or another, biased.

00:07:57.125 --> 00:08:02.184
You see, algorithms are by definition
sets of rules or steps

00:08:02.208 --> 00:08:05.917
that have been designed to achieve
a specific result, OK?

00:08:06.833 --> 00:08:09.559
But these sets of rules or steps
cannot be objective,

00:08:09.583 --> 00:08:11.726
because they've been designed
by human beings

00:08:11.750 --> 00:08:13.476
within a specific cultural context

00:08:13.500 --> 00:08:16.000
and are shaped
by specific cultural values.

00:08:16.667 --> 00:08:18.393
So when machines learn,

00:08:18.417 --> 00:08:20.667
they learn from biased algorithms,

00:08:21.625 --> 00:08:24.833
and they often learn
from biased databases as well.

NOTE Paragraph

00:08:25.833 --> 00:08:29.559
At the moment, we're seeing
the first examples of algorithmic bias.

00:08:29.583 --> 00:08:33.083
And some of these examples
are frankly terrifying.

00:08:34.500 --> 00:08:38.559
This year, the AI Now Institute
in New York published a report

00:08:38.583 --> 00:08:40.976
that revealed that the AI technologies

00:08:41.000 --> 00:08:44.476
that are being used
for predictive policing

00:08:44.500 --> 00:08:47.625
have been trained on "dirty" data.

00:08:48.333 --> 00:08:51.226
This is basically data
that had been gathered

00:08:51.250 --> 00:08:55.434
during historical periods
of known racial bias

00:08:55.458 --> 00:08:57.708
and nontransparent police practices.

00:08:58.542 --> 00:09:02.601
Because these technologies
are being trained with dirty data,

00:09:02.625 --> 00:09:04.059
they're not objective,

00:09:04.083 --> 00:09:08.601
and their outcomes are only
amplifying and perpetrating

00:09:08.625 --> 00:09:10.250
police bias and error.

NOTE Paragraph

00:09:13.167 --> 00:09:16.309
So I think we are faced
with a fundamental problem

00:09:16.333 --> 00:09:17.976
in our society.

00:09:18.000 --> 00:09:22.792
We are starting to trust technologies
when it comes to profiling human beings.

00:09:23.750 --> 00:09:26.268
We know that in profiling humans,

00:09:26.292 --> 00:09:29.101
these technologies
are always going to be biased

00:09:29.125 --> 00:09:31.851
and are never really going to be accurate.

00:09:31.875 --> 00:09:34.809
So what we need now
is actually political solution.

00:09:34.833 --> 00:09:39.542
We need governments to recognize
that our data rights are our human rights.

NOTE Paragraph

00:09:40.292 --> 00:09:44.375
(Applause and cheers)

NOTE Paragraph

00:09:47.833 --> 00:09:51.917
Until this happens, we cannot hope
for a more just future.

00:09:52.750 --> 00:09:55.476
I worry that my daughters
are going to be exposed

00:09:55.500 --> 00:09:59.226
to all sorts of algorithmic
discrimination and error.

00:09:59.250 --> 00:10:01.643
You see the difference
between me and my daughters

00:10:01.667 --> 00:10:04.851
is that there's no public record
out there of my childhood.

00:10:04.875 --> 00:10:08.893
There's certainly no database
of all the stupid things that I've done

00:10:08.917 --> 00:10:11.059
and thought when I was a teenager.

NOTE Paragraph

00:10:11.083 --> 00:10:12.583
(Laughter)

NOTE Paragraph

00:10:13.833 --> 00:10:16.583
But for my daughters
this may be different.

00:10:17.292 --> 00:10:20.476
The data that is being collected
from them today

00:10:20.500 --> 00:10:24.309
may be used to judge them in the future

00:10:24.333 --> 00:10:27.292
and can come to prevent
their hopes and dreams.

NOTE Paragraph

00:10:28.583 --> 00:10:30.101
I think that's it's time.

00:10:30.125 --> 00:10:31.559
It's time that we all step up.

00:10:31.583 --> 00:10:34.059
It's time that we start working together

00:10:34.083 --> 00:10:35.518
as individuals,

00:10:35.542 --> 00:10:38.059
as organizations and as institutions,

00:10:38.083 --> 00:10:41.184
and that we demand
greater data justice for us

00:10:41.208 --> 00:10:42.601
and for our children

00:10:42.625 --> 00:10:44.143
before it's too late.

NOTE Paragraph

00:10:44.167 --> 00:10:45.434
Thank you.

NOTE Paragraph

00:10:45.458 --> 00:10:46.875
(Applause)