WEBVTT
00:00:03.082 --> 00:00:06.455
Hi, my name is Safiya Umoja Noble,
and I'm an assistant professor
00:00:06.455 --> 00:00:09.481
in the Annenberg School
of Communication and Journalism.
00:00:09.481 --> 00:00:12.861
My research looks at racist
and sexist algorithmic bias
00:00:12.861 --> 00:00:16.473
and the way in which people
are marginalized and oppressed
00:00:16.473 --> 00:00:19.552
by digital media platforms.
00:00:19.552 --> 00:00:23.271
I spent 15 years in corporate
marketing and advertising,
00:00:23.271 --> 00:00:26.512
working for some of
the largest Fortune 100 brands
00:00:26.512 --> 00:00:27.561
in the United States.
00:00:27.561 --> 00:00:30.499
We were starting to redirect
significant portions
00:00:30.499 --> 00:00:33.861
of our advertising media
buying dollars online
00:00:33.861 --> 00:00:36.742
and thinking about, in fact,
how to game Google search
00:00:36.742 --> 00:00:40.182
and Yahoo! to elevate the brands
and amplify the messages.
00:00:40.182 --> 00:00:43.681
And so at the moment that
I was leaving corporate America
00:00:43.681 --> 00:00:47.130
and moving into academia,
the public was increasingly
00:00:47.130 --> 00:00:48.641
falling in love with Google.
00:00:48.641 --> 00:00:52.610
And this lead me to thinking
that this was a space and a place
00:00:52.610 --> 00:00:55.048
that needed to be looked at more closely.
00:00:55.048 --> 00:00:57.548
It was interesting to see
this total diversion
00:00:57.548 --> 00:00:59.748
of public goods,
00:00:59.748 --> 00:01:01.255
public knowledge,
00:01:01.255 --> 00:01:04.298
and libraries being shifted into
00:01:04.298 --> 00:01:07.928
a corporate, privately-held company.
00:01:09.078 --> 00:01:11.707
When we go to places like Google search,
00:01:11.707 --> 00:01:13.968
the public generally thinks
that what they'll find there
00:01:13.968 --> 00:01:18.748
will be credible and fairly representing
different kinds of ideas, people,
00:01:18.748 --> 00:01:20.256
and spheres of knowledge.
00:01:20.256 --> 00:01:25.597
And so this is what really prompted
a 6 year inquiry into this phenomenon
00:01:25.597 --> 00:01:29.538
of thinking about misrepresentation
on the internet, particularly
00:01:29.538 --> 00:01:34.537
when people are using search engines,
and that culminated in my new book,
00:01:34.537 --> 00:01:38.729
Algorithms of Oppression:
How Search Engines Reinforce Racism.
00:01:38.729 --> 00:01:42.977
People think of algorithms
as simply a mathematical formulation.
00:01:42.977 --> 00:01:47.249
But in fact algorithms
are really about automated decisions.
00:01:48.199 --> 00:01:52.491
In 2009, I was kind of joking around,
in fact, with a colleague,
00:01:52.491 --> 00:01:54.900
and I was telling him
that I was really interested
00:01:54.900 --> 00:01:56.810
in what's happening with Google.
00:01:56.810 --> 00:01:59.181
And just kind of offhand he said to me,
00:01:59.181 --> 00:02:02.291
"Oh yeah, you should see what happens
when you google 'Black girls'."
00:02:02.291 --> 00:02:06.611
Of course I immediately did the search,
found that pornography was the primary way
00:02:06.611 --> 00:02:10.911
that Black girls, Latina girls,
Asian girls were represented.
00:02:10.911 --> 00:02:13.580
That started a whole
deeper line of inquiry
00:02:13.580 --> 00:02:17.042
about the way in which
misrepresentation happens
00:02:17.042 --> 00:02:19.002
for women of color on the internet
00:02:19.002 --> 00:02:22.170
and what some of the broader
social consequences of that are.
00:02:22.170 --> 00:02:25.369
In my work, I look at the way
that these platforms are designed
00:02:25.369 --> 00:02:29.271
to amplify certain voices
and silence other voices.
00:02:29.271 --> 00:02:30.591
How does that come about?
00:02:30.591 --> 00:02:32.350
What is that phenomena about?
00:02:32.350 --> 00:02:36.051
What's the role of capital
or advertising dollars
00:02:36.051 --> 00:02:38.940
in driving certain results
to the first page?
00:02:38.940 --> 00:02:42.062
What do the results mean
in kind of a broader social,
00:02:42.062 --> 00:02:45.271
historical, economic context?
00:02:45.271 --> 00:02:49.300
So I contextualize the results
that I find to show
00:02:49.300 --> 00:02:53.341
how incredibly problematic this is
because it further marginalizes
00:02:53.341 --> 00:02:55.271
people who are already living in a margin,
00:02:55.271 --> 00:02:58.999
people who are already suffering
from systemic oppression,
00:02:58.999 --> 00:03:02.821
and yet again, these results
show up in these platforms
00:03:02.821 --> 00:03:07.330
as if they are credible, fair,
objective, neutral ideas.
00:03:07.330 --> 00:03:10.401
In the end, I call for alternatives.
00:03:10.401 --> 00:03:13.281
And I argue strongly that we need
00:03:13.281 --> 00:03:16.422
to have things like public interest search
00:03:16.422 --> 00:03:19.130
that are not driven by commercial biases.
00:03:19.130 --> 00:03:23.110
And I put out some ideas about
what it means to imagine
00:03:23.110 --> 00:03:27.572
and create alternatives
in our public information sphere
00:03:27.572 --> 00:03:29.670
that are based on
a different set of ethics.
00:03:29.670 --> 00:03:33.460
If anything, I think that this book
is the kind of book that will help us
00:03:33.460 --> 00:03:36.849
re-frame the idea that,
"We should just google it"
00:03:36.849 --> 00:03:39.201
and everything will be fine.
00:03:39.201 --> 00:03:41.382
♪ (music) ♪