Discriminating Data: Wendy Chun in Conversation with Lisa Nakamura
*transcript cleaned by Kaitlyn Gastineau


Introduction
0:00
[silence]
0:45
welcome everyone my name is uh Irina Aristarhova and on behalf of the Digital
0:51
Studies Institute and DISCO Network I'm really happy to present our
0:58
public event today the book talk Discriminating Data:
1:05
Wendy Chun in Conversation with Lisa Nakamura and Alex Barnett
1:11
um I am a professor and um
1:17
Acting Director of the Digital Studies Institute that is hosting this talk
1:22
and our co-sponsors for today's event are the following units at the University of
1:29
Michigan - Ann Arbor Asian Pacific Islander American Studies, English Language and Literature, American
1:37
Culture, Communications and Media, and the Stamps School of Art and
1:42
Design. Digital Studies Institute at the
1:47
University of Michigan provides space for scholars of the humanities and qualitative social sciences
1:54
who are experts in digital studies and supports their research and teaching
1:59
we're the nation's first one-stop place for all things digital including digital media studies, digital
2:07
humanities, digital pedagogy, digital art, aesthetic practice and design, and critical
2:14
thinking about the digital future as you can see the Digital Studies is a
2:20
big tent for us that has always included practitioners, critics, theorists, and
2:26
artists we are also a national leader in the study of digital media and race,
2:32
gender, and identity the next slide please
2:40
um before I go into the DISCO Network I would if Veronica if you could put the
2:46
previous slide in I forgot to mention here that we have a postdoctoral
2:51
fellowship applications open uh they are due on January 7
2:58
2022 and they are advertised currently at the chronicle of
3:04
higher education and I'm putting in the link right here so please for those
3:11
of you who are interested um and just gotten your doctorate in an area that's
3:17
related to the digital studies uh check it out the next slide please
3:23
the Digital Studies Institute is a proud uh home of the DISCO Network which is an
3:29
amazing Digital Inquiry Speculation Collaboration Optimism Network
3:35
led by Lisa Nakamura and you will hear more about it in a second
3:40
next slide please upcoming DISCO events include a Digital
3:45
Methodology Workshop with André Brock, Mad Frictions Workshop Series with
3:53
Remi Yergeau and from the Digital Studies Institute
3:59
we have an event on December the 9th Welcome to the Indigenous Future
4:06
and our Digital IDEAS summer institute will be devoted to critical access
4:12
technology and disability justice in summer of 2022 I welcome you also to wait for
4:19
applications and apply um and we will be happy to see you here in
4:25
Ann Arbor in the summer 2022 when the weather will be fabulous
4:31
uh the next slide please and here it's my pleasure to introduce
4:36
um a bios of our speakers today
4:42
Wendy Hui Kyong Chun is Simon Fraser University's Canada 150
4:49
Research Chair in New Media and leads the Digital Democracies Institute
4:55
she has been Professor and Chair of the Department of Modern Culture and Media at Brown University where she worked for
5:01
almost two decades she has also been a Visiting Scholar at the Annenberg School at the University
5:08
of Pennsylvania Member of the Institute for Advanced Study at Princeton and she
5:13
has held fellowships from the Guggenheim ACLS American Academy of Berlin
5:19
Radcliffe Institute for Advanced Study at Harvard Lisa Nakamura
5:25
is the Gwendoline Calvert Baker Collegiate Professor of American Culture and Digital Studies at the University of
5:32
Michigan she is the author of several books on race gender and the Internet
5:38
she is the founding Director of the Digital Studies Institute at the University of Michigan and the Lead P.I.
5:45
for the DISCO network next slide please
5:50
Alex Barnett is Group Leader for Numerical Analysis at the Center for Computational
5:56
Mathematics at the Flatiron Institute a division of the uh Simons Foundation in
6:03
New York City his Ph.D was in physics at Harvard University followed by postdoctoral work
6:09
in radiology at Massachusetts General Hospital and in applied mathematics at
6:15
the Courant Institute at NYU he then taught at the Department of Mathematics at Dartmouth College for 12
6:22
years becoming a full professor in 2017 his research interests include
6:28
computational partial differential equations, waves, fast algorithms integral
6:35
equations, neuroscience, imaging statistics, and inverse problems
6:40
he has published more than 50 papers received several grants from the National Science Foundation
6:46
won the XXI International Physics Olympiad and received Dartmouth's Karen E.
6:52
Wetterhahn Memorial Award for Distinguished Creative or Scholarly Achievement I personally look forward to
6:59
hearing today how Alex's work contributed to this Wendy's book
7:05
the next slide please and here I'm leaving you with hearing
7:11
more about the DISCO Network Digital Inquiry Speculation Collaboration
7:16
Optimist Network thank you and have a wonderful day
Discriminating Data
7:22
I can talk a little bit about the DISCO Network it stands for Digital Inquiry Collaboration Speculation and
7:28
Optimism and there are um six of us who are in this network and started it it's me
7:34
André Brock Catherine Knight Steele Rayvon Fouché
7:39
Stephanie Dinkins and Remi Yergeau and we have a three-year period during which we are hoping to
7:45
write teach and do some cool research on the intersection between disability race and
7:51
technology so I'm just going to dive right in I am so excited to be talking to Wendy and
7:58
later her co-author Alex about this new book Discriminating Data
8:03
from MIT Press and I'm hoping to
8:08
get at what this book is contributing to the field which is a lot so
8:14
in recent years we've seen quite a few critiques of the tech industry Wendy and I as we were saying before
8:20
both remember when there was a lot of utopian optimism and it seems that there's a kind of backlash to that
8:26
but I think what the field and what the conversation is missing is both nuance and history
8:32
around this question you know so this book Discriminating Data Correlation Neighborhoods and the New
8:38
Politics of Recognition provokes us to think about data in a new
8:44
way and it's beautifully written beautifully researched it has some fascinating
8:50
exercises that are very interactive which I've never seen in a critical book like this before which we'll talk about
8:55
a little more but I wanted to ask Wendy some things about the title
9:00
so Discriminating Data is a title that can go in lots of ways and that's one thing I love about it so
9:07
if you read it as being about data as the subject data playing a role in
9:13
creating new kinds of social discrimination against and systemic denial of resources
9:18
to marginalized groups which I feel now we know is true right that's not even that controversial but it was quite
9:24
controversial a while ago the data itself is an agent in creating discrimination against
9:30
people specifically people of color women and so on the data is an agent that makes
9:36
discrimination happen and that we know discrimination is wrong so that's one way to read it and I think
9:43
if we look at another way we can see discriminating data is something that we as humans can do
9:50
that data isn't all the same if we learn through critical
9:56
theory critical thinking maybe a little bit of math I'm hoping not too much math because that part I need some help with
10:02
um but I know we're going to have to get some help with um that if we see us as possibly being able to discriminate
10:09
between not only data which discriminates and data that maybe doesn't discriminate as much
10:15
but if we can learn to see somehow the inner workings and really important distinctions between different kinds of
10:20
data right personal data data about our face data about our name
10:26
data about our race and gender and understand what kinds of things people are doing when
10:33
they're making data can we see through data and what is being done with it can we have some
10:40
agency to know and I think that's what the tutorials and the book are about
10:45
to empower ourselves to know how data is discriminating not just against us possibly but creating new kinds of
10:52
discrimination maybe even new ways of being discriminatory that used to be invisible to us and something I love
10:59
about this book is that in our field we do a lot of gesturing towards
11:04
discrimination but only as an effect right that I know I was denied credit I know that I didn't get to see what my
11:10
friend saw because I'm a woman because I'm Black right or because I'm Asian
11:16
but though how is not always that easy for us to get at so in some ways we're
11:22
all grabbing different parts of the elephant you know we know that data is discriminatory but we don't know how we
11:28
can discriminate ourselves about how data is discriminatory and what it's doing not just to me but to
11:35
people I care about and as Wendy says people we care about are
11:40
in some ways the problem right that people we care about are people who are like us oftentimes
11:47
um they're our neighbors sometimes we care about our neighbors sometimes we don't sometimes we
11:52
recognize people we care about as being like us in a more ideal scenario we could recognize people who
11:59
aren't like us but are machines able to recognize that and what are the implications of that so I think in the
12:07
first case we could think of data as something which is doing things to us
12:12
and in the second we can hopefully see ourselves as people who can discriminate when we look at data if we get to look
12:19
at it and not just feel like objects so I'm going to invite Wendy to talk about method
12:26
because I was so compelled by the way this book draws its conclusions
12:33
what this book is doing to me is to create new origin stories for how we got
12:38
here so I think it's really easy to blame companies for being bad and that is not
12:46
even necessarily true but it doesn't tell us anything about the how so you know we're kind of swimming in
12:51
algorithmically curated news facial recognition systems social media um there's a lot of dissatisfaction with
12:58
those things but Wendy I think cuts right to the heart when she writes big data is the bastard child of
13:04
psychoanalysis and eugenics so I wanted to invite her to talk about this alternative genealogy
13:12
that in some ways pushes a lot against this focus on famous men and what they did wrong because that's
13:19
not really getting us anywhere so we used to think you know Norbert, Wiener, Vannevar, Bush all those
13:25
people we teach in intro classes were good and then we thought they were bad right and this is not a history which is
13:32
very fruitful so in this book Wendy you spent a lot of time looking in unfamiliar places
13:39
so not at MIT not at Silicon Valley but rather at housing
13:44
developments in the 50s you know different kinds of
13:51
racialized engendered places where we don't necessarily look so can you
13:57
talk more about how you found and selected um the kind of new origin story in
14:03
eugenics and physiognomy to understand say network science and facial recognition technology why it
14:09
matters where we get our histories from and what new light this sheds on people
14:15
doing work on say race but not necessarily technology or looking at say gender but not necessarily
14:22
the history of data science so that's a long question but I just wanted to invite you to talk about how you did
14:28
this and what you were what you were doing when you did it um so this is Wendy thank you so much
The Problem
14:34
Lisa for your um summary and your incredible question I'm so thrilled to
14:39
be here in conversation with you and um the DISCO Network
14:45
so I'm gonna this is a huge question um and so I'm gonna answer it in in several
14:51
different ways um the first is um one thing that was important for me in
14:56
terms of methodology and stories that we tell is to move away from the idea that
15:03
if we only added humanities and social sciences to technology then we would have ethical tech
15:09
so the problem is that we don't have social science or humanities interests in these that these are purely
15:16
technical and that's what's wrong because as I started reading more and more
15:22
about how these concepts became axiomatic notions like similarity breeds
15:27
connection or homophily which is what you mentioned earlier Lisa what became so key is that these were borrowings
15:34
from social sciences and the humanities so the issue isn't that the social
15:40
sciences and humanities aren't there but that concepts have been borrowed poorly or poor concepts have been borrowed um
15:48
and so part of what's needed is a richer more interdisciplinary
15:54
engagement where we're not just taking each other's methods poorly from each other but trying but understanding the
16:01
weight of what it means to work with these concepts and so what I always do and this comes I
16:09
think from my background both in engineering and in english literature is always to start with a basic concept um
16:16
and to ask myself how did this come to be and also to try to think through the the
16:22
very different and often conflicting meanings there are for any one concept so such as control freedom etc
16:30
um and what became so important for me when I was looking at these um axiomatic concepts and the first one I start off
16:37
is with correlation right so the beginning of the 21st century you heard about correlation all the time
16:43
right this was what was different about big data this was what was different about knowledge it was you know let us
16:50
get you know take over the future not the past um what I found really odd about that is
16:57
that the rhetoric around correlation was almost the same rhetoric that happened a
17:02
century earlier when eugenesis came up with the concept of correlation
17:09
um and why the rhetoric was so eerie is that both eugenicists and people talking
17:15
about big data were talking about correlation not as a way to understand or think statistically about the past
17:21
but rather to understand how to determine the future once you've determined things in the past and the
17:27
present that do not change um and so that for me was absolutely key
17:33
and trying to think through why and how is correlation as this future predictor
17:38
this way to close the future which is what eugenics thought to do um coming back but now in the name of disruption
17:46
so if what we're undergoing is disruption it's not because something radically new is
17:52
happening and the future can be anything but rather because the future has been closed through this logic and so for me
17:59
going back to the eugenicists and trying to understand what's different though about how they used correlation and
18:06
thought about correlation and society and us um was also key
18:12
because for eugenicists um they believed that their population was the nation as a
18:17
whole right um and so they had very disturbing notions of uplift
18:22
whereas now with a new correlation um it's being used to divide communities
18:27
and what's being touted are certain forms of exit rather than a communal exit rather than
18:36
this national uplift and that has everything to do with questions of race
18:41
and how race and difference is being thought through and so the ways in which homophily which
18:48
is the idea that similarity breeds connection has become this default concept and so the second chapter looks
18:54
at homophily and as you noted it comes from studies of housing projects
19:00
biracial housing projects and so in a bizarre way
19:05
by using the the notion homophily what we're doing is spreading segregation
19:11
through these networking concepts right so segregation isn't a
19:16
accident it's the default um and but what was also very key to me was
19:22
to go back to these studies and to see like even back then homophily wasn't the only choice
19:30
and that if you go back to the ways in which these concepts themselves were formulated you can think through other
19:36
possibilities and understand how the amplification of certain types of relations are destroying others so just
19:44
very briefly with homophily what was so key is that um this was um this came from a
19:49
housing study that came in the 1940s it was when there was incredible income
19:54
compression and also a real housing crisis and the question was you know what will
20:00
what should we do should we have more public housing and back then public housing was mainly
20:06
white or should we have private housing should we have mortgages
20:12
and we know what happened and we know the incredible wealth disparity that has
20:18
resulted from the decision that was made but back then they they you know said
20:24
okay so let's study these housing projects one housing project was an all-white co-op
20:30
um and the other one was a biracial housing project and they asked every
20:35
member of the families of these groups um questions like these were four-hour
20:40
interviews and so this was um blatantly forms of social engineering
20:46
and of um sociological surveillance they talked about this as a laboratory this is why housing projects were so
20:52
interesting because they were forms of social engineering so that idea of an enclosure in engineering which has now
20:59
become of course the way people describe social networks was there from the very concepts that
21:04
were drawn from and thinking through these enclosed communities therefore opens up these
21:11
historical narratives yeah something I love about the way you write this book is that it's very
Running Away
21:18
precise but there's also some very accessible kind of memorable writing in here and one of them is how do you show
21:24
a love of the same by running away when others show up so um I think as you say right the idea
21:31
add social science add humanities and stir doesn't get at the root issue here
21:36
which is the way these things including social science and the humanities are put together to create forms of
21:42
discrimination so how do we know when we run away how do we know when we are allowed to stay
21:49
right and I'm interested in how that maps on to say ethnic studies or women's studies
21:55
which are in some ways as you say structured around what you call the productive conflicts
22:01
that in some ways are antithetical to data science right because data science wants to fix something as having a
22:07
nature and then perform a bunch of operations based on that nature
22:12
which can be mistaken you know people think if we just clean the data maybe it won't be so racist but I think what
22:19
you're saying with this origin story it doesn't really matter how much you clean this data
22:25
right it then it's instead about the kind of premises of what data is
22:31
um and what conflict is and how conflicts should be dealt with in a network right so are we going to run away when
22:37
others show up and if not what are we going to do so this is a big question but I'm just
22:44
going to ask it since I think people would love to know from a kind of data science and also critical perspective
22:52
what you think given this is a kind of poisoned well to begin with but your history but certainly you know my
22:59
research when I think about what I've been trying to do too is to say how can something
23:05
based on say the labor of Indigenous women making our stuff not produce
23:10
discrimination you know it's invisible we don't know where that stuff came from but that is itself part of the problem
23:17
that the origins of things are mistaken you know the wrong kind of p- not the
23:22
wrong kind of people but certain bodies are getting a lot of credit good and bad for having created
23:29
the digital and others are totally totally excluded in many ways
23:34
so I'm going to transition to a question about dirty data
23:39
um because you write machine learning programs are not only trained on selected discriminatory and often dirty
23:44
data they are verified as true only if they reproduce these data
23:49
right in other words as you say they predict the past not the future
23:55
which makes sense that you're going to you know the eugenics is at the heart of these things so your Lombroso pictures
24:01
right of facial recognition I've been waiting for someone to do that for a long time and say you know this is
24:08
completely old this was dirty data too right these drawings of Jewish noses or
24:14
of Asian eyes are completely dirty data but we just didn't call them that back then
24:20
um so this is a theme throughout your book as you write social network create and spawn the reality they imagine they
24:26
become self-fulfilling prophecies so how do you interrupt this repetition of
24:33
the same um big question but I had to ask you
Deep Fakes
24:39
thank you Lisa um so Wendy here again I think that what's
24:45
absolutely key in the question that you asked is there's two parts to it right
24:50
one is uh the question of dirty data or what I would call deep fakes right so
24:56
what I argue is that deep fakes aren't just the product of these programs they're actually the ground
25:03
truth right these either corrupted police files
25:08
hollywood celebrities as the basis for facial recognition um training etc um what we
25:15
have is the problem that the deep fake has become ground truth so what's replicated is true is this deep fake
25:22
right but the other argument is even if we fix the deep fake you'll still have
25:29
reproduction of a certain ground truth as
25:34
what is true and part of this is because you know when these programs are trained
25:41
what they're verified against trying to see if they're they're predicting things correctly isn't future values but rather
25:47
certain values that have been held back during the training process and they're usually from the same data set or
25:53
they're out of sample data and so one thing that's fundamentally
25:59
that we need to think through is this definition of truth as repetition because if truth is repetition then only
26:06
racist and society is racist and the ground truth that is based on is this
26:12
deep fake racist past then these things will only be verified as as correct if they make racist
26:18
predictions so one of the largest things that I put out there is that we need other forms of verification validation
26:26
and truth and this is why we need to have an interdisciplinary dialogue because definitions of truth within the
26:33
humanities and philosophy for instance don't rely simply on repetition I mean
26:38
Hannah Arendt famously said if truth becomes um repetition then it becomes
26:43
nothing right and people always you know experience
26:48
things like films or novels that feel profoundly true even though they are fictional
26:55
so I think that these larger questions of truth are absolutely key but they also
27:00
therefore don't mean that you know truth is something absolutely unreproducible but it gets us to things even like first
27:06
principle and the relationship between first principles and truth so for example if you have this mindset that oh what is
27:13
true is what's reproducible if you think about global climate change you're absolutely screwed right because you
27:19
have to wait for global climate change to happen therefore you can verify it as
27:25
true um you've lost the battle and I think global climate change models make absolutely clear the difference between
27:32
what's true and what's correct yeah I think that's awesome because it really encourages other scholars to try
27:39
and find data that doesn't fit into this receive story right as you say in another section of
27:47
this book um the internet has become nothing but social experimentation to normalize the
27:52
non-normative so I love this kind of invitation to normalize
27:59
or to um to think about how the non-normative can be disruptive here so and that's the
28:05
kind of project of this alternative historiography I think that you and other people as well have been doing
28:12
um and I'm super fascinated by what you say is democracy equals offense
28:17
because I'm very offended all the time by the internet I feel like that's my job
28:23
and you know I hear what you're saying as well that data is doing a great job discriminating
28:29
against us but we're not doing a super great job discriminating data and this book is part of that project so
28:35
I wanted to invite Alex who is a collaborator of yours to speak to how his contribution is going to help
28:42
us do that I mean I'm a humanist I totally don't know anything about
28:48
math so I'm hopeful that you can speak more to how what you were thinking to help people who are like me
28:55
learn to do a better job hi Lisa
29:00
thank you it's Alex here um that's a tough question uh
29:06
so I think our goal of the collaboration with Wendy was to
29:12
put together tools to understand what's going on in a way that hasn't been done before like where you actually
29:19
see a basic sort of scientific and in
29:25
equations and graphs and pictures understanding of of the methods alongside well I mean I'm just adding
29:31
that onto these sort of in-depth history of where these ideas came from and it's it's
29:36
invariably fascinating like how uh tied up with the um
29:42
some of sort of scary social science history that the math is and and that you don't
29:49
even it's that's not taught uh if you go through a masters in data science and then you go out
29:55
into the real world and get hired by a million companies to uh to sell you know
30:01
to control their marketing algorithms or whatever I mean this is the story of
30:06
tens of thousands of graduates all trying to to get those jobs right now they are not taught
30:12
like I mean for instance the history of Linear Discriminant Analysis which i
30:17
sort of had was really shocked when I dug into that and learned that and we sort of put this
30:23
in the book so I'll hold I'll hold up the section just to cue those if you have the book but um
30:30
that this method was literally made in order to separate and segregate the rate
30:35
uh the different castes as the British uh uh Empire saw in India and that led
30:42
to this method that then is like now a totally standard method in statistics and uh
30:48
data analysis and it's the first thing you would learn if you want to classify pictures of cats and dogs and
30:54
all of the things that big data lets you do so part of it is just to
31:00
to to put the two together I mean normally you read a history book you'd read a
31:05
statistics book and they would rarely talk to each other um
31:11
so does that help solve the problems I think
31:16
I think you have to be informed on both sides about literally what the algorithms are doing
31:21
the sort of operations of formulae and then uh what they can lead to and and
31:28
where they came from uh and you have other
31:34
team members that are gonna can really help with this like journalists and then uh
31:40
computer scientists and whole teams of people that are questioning the ethics
31:46
of how these algorithms what they end up doing to people yeah something um I really
Machine Unlearning
31:53
resonated with reading this book is the idea of the machine the AI apocalypse because it was
31:59
longed for and then it was feared and people still talk about it I feel like it happened a while ago
32:04
and we're more or less just getting our head around that or starting to
32:09
recognize that it's already happened and now what are we gonna do so in the conclusion you urge the reader to
32:16
prevent this or rather not to prevent it but like try to engage it by um
32:22
doing machine unlearning so as you were saying Alex it might be learning more about the colonial logics
32:28
of these machines as a way to unlearn what the results might mean for us
32:34
in other words to create a different future not to repeat all the time so what are some specific things that we
32:41
might consider doing in our everyday life to enact machine unlearning
Ways to enact Machine Unlearning
32:52
that's a great question uh Lisa and there's so many different ways to think about it depending on who you are what
32:58
your everyday life looks like I think that the term unlearning of
33:03
course comes from Ariella Azoulay's work on potential history and what she
33:09
encourages us to do in terms of unlearning the archive is to move away from treating people as sources of
33:15
information but rather instead treat them as companions and so one thing that I argue in the
33:22
book is if you think through these communities that have been at the heart of this kind of analysis so
33:28
residents of biracial housing projects and Lisa as you and I have been looking at together residents of
33:35
Japanese American Internment camps who are central to the emergence of sentiment analysis right
33:42
what does it mean to treat these folks as companions rather than sources of information and what that looks like is
33:49
a serious dealing with the legacies of this and to think through questions of reparations and of living
33:57
with these histories and you also mentioned the apocalypse so what's so important
34:02
for me is the argument is um we the fact that we have this fear of discriminatory
34:08
algorithms and the fear of this um AI apocalypse at the same time coming up
34:14
isn't an accident and we need to actually think these two together because one way to one reason why we're
34:20
so afraid of of the uh AI apocalypse is because machines
34:26
are framed for us as servants um as slaves they're profoundly raced and
34:32
gendered right so (inaudible) has written about this of course you've talked about the labor practices as well so to take
34:38
on these issues we need to take on these questions of labor the framing of these machines and how we want to relate to
34:44
them and here this is why I think Indigenous Studies is so important and they've written texts like the
34:51
Making Kin with the Machines you know that's Jason Lewis et al's really important contribution as well as the
34:58
AI Indigenous protocol and what's so key about that in terms of thinking about apocalypses that have been I mean one
35:05
thing that's really clear within Indigenous Studies is that the apocalypse has already happened
35:10
and trying to engage with what remains and what sort of relationality is fundamental so I think
35:16
that's one thing let's start thinking through these entire relations that are part of
35:21
your engagement and what is your relationship to that is a fundamental relationship I think that the other
35:27
thing that um us as citizens and and as people need to think about is not simply you
35:34
know is this technology good or bad but what is it amplifying how is it amplifying and also what are the root
35:41
issues that are being um brought up by these so to ask ourselves
35:46
okay instead of saying okay this risk recidivism um
35:51
program is discriminates against racial minorities we need to fix it right we
35:57
have to realize actually that this thing is operating exactly as it should and what it's revealing is that the criminal
36:03
justice system is discriminatory um so one thing that we need to do is to take these results and then say like
36:10
global climate change models show us what's wrong with the world or what the future will look like if we keep doing x
36:15
y and z to take these results as cautionary tales and say okay then if we don't want this future then we've got to
36:22
change the way the criminal justice system operates we've got we've got to think through these questions
36:27
of injustice and just to be very engaged in what these these are doing
36:33
but Alex
36:38
um I all I would add is that yes the the apocalypse I think began for me
The Discount Card
36:46
sort of when you think about some uh a very simple corporate uh construction
36:54
namely the uh the discount card right in the old days right I think people
36:59
understood capitalism if you go to a store and you buy something well the store wants to sell
37:04
you things and and and they want you to buy as much as possible I mean that can be hyper
37:10
capitalistic but at least the everyday person understands what the
37:15
goal is of the company is to sell you some stuff and okay so then we get good at looking at ads and understand sort of
37:23
the media literacy at that level is important but when you get a
37:28
discount card that's sort of the first level where you where the consumer the consumer
37:34
uh starts to become a source of data and that even happened you know in the 80s before the internet
37:41
before we started hooking all the computers together and things went terribly wrong right so
37:47
now it's like that on steroids we have um the interaction you think you're having
37:52
with with a company is actually nothing like what is going on like you think with
37:57
Facebook you think you're getting a free service that it's kind of fun and you can share
38:02
information you know be social okay but what the company thinks is going on is and what is actually going on is that
38:09
you're generating data uh immense amount of personal information that is then sold to
38:14
advertisers and that kind of media literacy we need to sort of teach this to
38:20
everybody and in a sense then you have to understand the algorithms too as well as what they're doing and I don't know i
38:26
guess that's that's where I think that's why I think we have an apocalypse it's it's not capitalism it's that
38:32
people haven't that the companies have been very good at actually telling us a different story about what's happening
38:38
from what's really happening I sound paranoid right now but no that's such a good example you think you're
Free Bread
38:44
getting a free loaf of bread but you're really not you're instead you're giving something
38:49
I'm thinking of Greg Elmer's book Profiling Machines that came out in 2003 that's about discount cards in Canada
38:55
I remember reading this and thinking wow people should really be more concerned about free beer because [Alex: yes] there really
39:04
isn't such a thing
39:09
um okay I think we have a few more minutes before we want to open it up to questions from
39:16
people here did you want to expand on any any things that came up or
39:23
answer a question I didn't ask yet
39:34
well I hope we will bounce ideas around with what can be done concretely
39:39
for new either social platforms or um
39:44
laws you know on the legal standpoint uh things like Section 230 and then
39:50
making companies into admit that they are media companies not platforms this kind of stuff I don't
39:56
know it'd be interesting to see what maybe people will ask about that those kind of steps yeah can I ask what you're doing
40:05
me right no yeah well no no not this second but your practice is generally informed by what you know around how
40:12
data is being deployed well I mean I'm mostly a mathematics
40:19
researcher and working on algorithms that help solve science problems rather than social problems I would say so I
40:26
mean this is a this is a new venture for me uh but by
40:31
you know being inspired by Wendy's work and talking to her over the years and and
40:36
so I mean I've got I don't have like a personal project on this I just try and follow what's going on to discuss uh
40:44
and I mean I have been involved for instance I mean one one idea that that counters
40:51
the hegemony that the homophily concept that's so central to to this
40:57
book to Wendy's ideas uh is that you shouldn't group people together by what they like or who they
41:05
are or you know that you should randomize so the opposite of uh of segregation is randomization and I have
41:12
been involved in student housing where similar to this housing project work
41:18
uh for instance the fraternity system uh place I used to teach
41:24
is all about hierarchy and segregation and we really tried to fight to remove that
41:29
system and then you would have one based on people being placed with people who are different from them and then they
41:35
have to talk to each other I think that's one of the main ways progress can occur is okay
41:41
sharing ideas I should have framed my question differently I was thinking whether you thought we should quit facebook like that kind of thing because
Should I quit Facebook
41:48
people when they get a data scientist want to know should I quit Twitter should I quit Facebook should I use my
41:54
credit card like it becomes very kind of embodied of is it I always have
41:59
this feeling I'm doing something wrong when I post an Instagram picture when I tag somebody
42:05
um so I was wondering if you had any opinions about that I mean maybe as this book kind of sketches out that the it's
42:12
already such a systemic issue that they that may not matter very much in the big picture
42:21
i all I'll say is I think we have to educate in a sort of new form of media
Media literacy
42:26
literacy that teaches the kind of things Wendy discusses so that if you do post a picture you
42:33
know what's happening and why the company wants you to post a picture right uh that maybe maybe that
42:39
and you know I happen to deplatform from Facebook and and really enjoyed that extra
42:46
you know 20 minutes a day that I got but that's not I mean you know there's
42:52
going to be other things maybe we should have platforms that you actually pay a little money for a little like the telephone
42:58
you know in the old days I mean ATT was a big nasty monopoly but at least you were the customer
43:04
and you were paying for a service that was in order to make a call to somebody else now who is the customer well it's not
43:10
you anymore it's it's the advertising it's the businesses who advertise it's totally flipped around and and if we
43:17
could just if people could be aware of that when they use the platform I think that would help a lot right yeah I think looking at
Deepfakes
43:24
the um the kind of worksheet you know tutorial parts really helped me
43:30
understand why deep fakes is the kind of new stereotype right or the
43:36
new model that when things look like each other they're just much easier for machines to deal with them they
43:43
become preferred right so some of the pictures in here of faces that all look the same
43:49
and then the math kind of explaining why that's preferred really helped me understand why some faces aren't seen as
43:55
faces and some people aren't seen as people
Opening the chat
44:02
okay so I think it's time to open it up so let's see if there's any questions
44:08
from um people who are here I'm sure there are some awesome questions I look at the chat I'm like oh this is an amazing
44:15
group of people um so uh yeah I will
44:20
look at the chat and see what's going on here ah okay so we already have a really
44:28
interesting couple questions um what is a meaningful collaboration between computer scientists and
44:34
humanities scholars look like in practice uniquely qualified people to answer and
44:41
what does an ethical AI look like for me Wendy and Alex
What is a meaningful collaboration
44:49
so why don't I start with the first uh question this is Wendy um so I think that uh what a meaningful
44:56
collaboration looks like I'll start with an example that draws from my own experience um
45:02
and so at one point at Brown when the new president came in um she said we
45:08
want faculty-led initiatives um but of course part of this was to put the faculty
45:13
together who were going to then initiate projects and so I was put together with a head of computer scientist science and
45:20
somebody in biostats who worked on AIDS and we looked at each other and we said
45:26
well we're not going to work together unless there's a problem that we know we can't solve on our own so in other words
45:32
why do we all need to be at this table what is the question we can't solve and the question that we came up with
45:38
which was initiated from the person in biostats was I can show almost any correlation to be real I can show almost
45:46
anything to exist but I don't know what's true I don't know what matters and then the other question that was
45:51
initiated from some of my work on global climate change models is just because something is true doesn't mean that it
45:57
will inspire action so what is that relationship between truth and action and so we called it DIAL the data
46:04
inference action loop um it didn't it wasn't one of the ones that were chosen but it became a model for me of
46:10
interdisciplinary work that happens and it starts when we all have a question
46:16
that we that we're interested in that we want to solve and we know we can't do it on our own
46:22
and then we work with others because they're different which means respecting their methodology
46:27
but it also means questioning their first principles and having our first principles
46:33
questioned in turn and so that's what the digital democracy institute is all about is to take
46:40
the hard problems that face us from disinformation to abusive language to
46:45
network segregation bring together humanists, data scientists, network scientists, um people um in the arts, and
46:52
the social sciences and to say okay um this is what this is how we've approached this problem but it remains
46:59
why does it remain and so how can we use some of these roadblocks as opportunities for intervention and to be
47:06
as concrete as possible our project on misinformation and disinformation for example starts from
47:12
the fact that fact checking is absolutely important but it's not enough like over and over again people pointed
47:19
out that fact checking can actually spread stories to different audiences
47:24
um in fact some people spread stories they find compelling um whether or not they're
47:30
correct um and if you think through the 199- sorry the 2016 election what's so
47:36
key is that not only was it called the fake news election it was called the authenticity election and the person who
47:43
won scored high on both and so rather than saying okay this is a post-truth era and putting our hands up
47:50
we say well actually what's happening here um so instead of saying okay why is
47:55
NDP like how is this correct or incorrect ask ourselves why and how does anything come to feel true
48:02
irregardless of its facticity and what is that relationship between feeling true or authentic and facticity um and
48:10
here you know we've been drawing from theater performance studies and literary studies which has done you know a ton of
48:15
work around questions of authenticity um and trying to understand that relationship between authenticity and
48:21
authority truth and trust fact and things that were done
48:26
and so starting from what seems to be an absolute roadblock but then using every discipline and that where each
48:34
discipline is broken to draw from another discipline to push the question further
What is an ethical AI
48:46
did you want to add anything Alex um uh I'll just give an example of another I
48:53
mean other than DISCO and Wendy's uh wonderful institute there is amazing stuff happening at
48:59
Tufts University from a mathematician called Moon Duchin who has the redistricting project there
49:05
and she brings together train computer scientists lawyers social scientists and
49:11
uh you know political scientists who attack redistricting and really try and
49:17
create expert witnesses that then can testify and this so this really connects to the legal
49:23
system and I think so regulation is part of of uh what an ethical AI
49:28
has to be it has to the algorithms have to be available for examination
49:34
uh and um you know even if you have to sue it out of I mean the government has
49:41
to get squeeze it out of corporations it's going to be very difficult but they may
49:47
have to pass tests that maybe we could devise such tests uh
49:52
so I don't know I I'll probably just leave it at that awesome well here's a great question
Antiethical AI
49:59
about um truth before we do this Lisa they also asked you what you thought ethical AI
50:04
looked like [Lisa: oh gosh] I know that you've been thinking through the limitations of
50:09
this yeah well just to take a little piece of it um I realize I quoted you incorrectly
50:16
a little bit ago around democracy and offense because in this book you write how toxicity and
50:21
racism are baked into these social media platforms as proof of concept of debt work is diverse so if if people are
50:28
basically being really racist to each other and you know horrible and nothing is happening in in return that's kind of
50:36
proof of concept that democracy is working so you know there's some statistics
50:42
which I believe that only three percent of hate speech is ever called out in any way
50:48
right much less regulated meaning people actually losing access to accounts or even being
50:53
permanently banned three percent which seems high to me I forget where I get got this from is the paper or something
50:59
so you know it's incentive to make us this big racism machine
51:05
as evidence that democracy is working because democracy is about people
51:10
getting to do what they want and the evidence of doing what you want is doing something someone else doesn't want
51:15
so that to me is anti-ethical AI you know it's the basic principle you're writing about here that is only going to
51:22
make more of the same as people get more concerned about the apocalypse like democracy America we have to preserve it
51:28
we must have more racism as evidence that there is we're preserving it even more vigorously than before which
51:34
obviously I'm against um but having read this book I can see how that gets defended
51:40
you know and how as there was an article in the Post not long ago that Tarleton Gillespie cited
51:48
that shows that Facebook was actually regulating anti-white racism much more rigorously than anti-Black
51:55
racism right because they had decided that was made more sense and was probably you
52:02
know optimal I mean who knows um which is the opposite of ethical right
52:08
it's anti-ethical AI so seeing how these ways of
52:14
kind of I would say reproducing racism are
52:19
are obviously incentivized by companies like that has to happen that has to stop and I'm not sure exactly how it has to
52:26
stop but reading this book has made me see why it's happening that way right um and why there's so much disincentive
52:33
for companies to do anything about it um so there's a question here about
52:39
truth which I feel like I have to ask because it speaks to this question of authenticity which is core to this book
52:45
right so the the question of why we have to see people's personal really unpleasant details as you write
52:52
is evidence of something different from truth which is identity authenticity people being real often means
52:59
embarrassing themselves or others or suffering a lot of shame because of
53:04
personal details coming out right so um there's a request here for you to
53:10
speak about the differences between true correct
53:15
authentic and real and how these subtle terminologies are
53:20
part of a new model for media literacy
Authenticity
53:28
thanks that's a great question and I would also point that there's a lot of people working on this and so I think
53:34
that Sarah Banet-Weiser's work on authenticity has been fantastic I think that um the work that uh you know Alex
53:41
Juhasz and Ganaele Langlois and Nishant Shah have done around really fake is important and the beyond verification
53:48
project is a large international collaboration um so in terms of the relationship between what's authentic
53:55
true correct and what was the third term
54:01
um I think it was real real as in real ID
54:07
that's a real ID yeah those are so important and I think that one way to think through those and the differences
54:13
between those and the resonances they make is to go exactly with the example you started with Lisa which is how this
54:20
call for us to be authentic where the idea is we become transparent our insides and outsides coincide not
54:27
because we're being sincere not because we're being good people but rather because we're exposing something nasty
54:33
about ourselves or other people and therefore showing that we have nothing to hide right um and that this
54:39
concept of authenticity is so valuable because this open secret about ourselves
54:45
this nothing to hide about ourselves that we decide to parade as a way to say we're authentic is is often those
54:51
divisive likes that are really effective in terms of clustering people within network
54:57
neighborhoods right and so this is being weaponized and used against us and what's so key is that
55:03
authenticity shares the same root as author and as authoritarian
55:08
so it's always linked to this question of verification and certain notions of authority
55:15
against this or in relationship would be the question of what is true
55:22
where truth comes from the same word is trust and here I think that the work
55:28
that's been done in history of science by Steven Shapin and others has been actually crucial to
55:34
showing how it was only in the 18th century that this notion of truth emerged um that
55:42
that equaled what is quote unquote correct right and what he points out is it's actually a bunch of dudes in the
55:47
18th century within science trying to claim for claim certain modes of truth right um saying that these experiments
55:54
could be reproduced but they couldn't be right but the question was how does one learn to trust certain
55:59
infrastructures so that what is being produced is considered to be valid right
56:05
and so I think that thinking through the relationship between truth and trust not in a cynical way but to actually
56:11
understand that truth and trust are intimately intertwined is crucial to thinking through why it
56:18
is you trust something that this thing isn't like um so again what I think is so important is not to
56:24
say the issue is that people have become cynical or distrustful because democracy
56:30
is based on people questioning things the problem isn't that people question CNN it's that questioning of CNN that
56:37
leads to trust of epoch times so it's not the fact that people question is where they land
56:44
in this like trust of even more dubious sources that we need to think through and I think that the idea of constantly
56:51
questioning this literacy of constantly questioning so you don't land somewhere is absolutely key and tried to think
56:57
through structures of trust um in terms of what's real real of course
57:03
is race it's a thing right it's uh it gets us to all sorts of questions about what what how these things emerge and
57:11
how we consider them to relate to others and I think that starting from those
57:18
differences is really key because it moves us away from saying again this is post-truth this is post-modernism blah
57:24
blah blah as if this was new but rather getting us to think through these moments of trust and authority and
57:31
why it is that we expose these things when we do and how we do them
57:39
Alex did you want to add anything to it well I'm just reminded of the old Gary
57:45
Larson cartoon that shows that the dog typing on the computer saying nobody knows you're a
57:50
dog on the internet uh that that that sort of the es- you know this they knew that
57:57
(inaudible) there's a cat and mouse game not to fix too many animal metaphors going on
58:04
actually I sort of had fun illustrating this with the Bayesian spam filters which is a sort of
58:10
well-known uh even uh web 1.0 example of or even
58:15
pre-web right email hacking uh competition that's sort of going on can you get people to click
58:22
on um emails that sell you something and uh or to give away your money or just get
58:27
something in front of your eyeballs that you don't you didn't particularly think you wanted
58:33
to see so that's authenticity and the there is just this huge financial
58:40
incentive now we have this tool of instant duplication by you know you can
58:46
send billions of emails with with very little effort um
58:52
that's what leads to sort of the danger it's the scale that this is on you know in the old days you'd have to print
58:57
pamphlets and stick them on you know around town or something now you the scale in which you can uh fake
59:04
authenticity is huge and the amount of effort therefore that computer scientists are going to spend
59:10
meaning you know whatever we call them hackers right but maybe they're just people working for a company are going to spend
59:17
uh trying to trying to fake something to get in get inside your
59:23
your soul in some way is huge and that that's you know it's just not happened and
59:29
that's why democracy is at such stake as well truth is
59:34
very slippery to define even in academia right with these these studies the recent
59:39
uh repeats of like 100 classic social science papers and maybe a third of them actually hold up at the level that they
59:46
were originally claimed for businesses it's I mean they're they don't they
59:52
well it's kind of interesting they're constantly doing AV Testing as Wendy explains like the the multi-armed bandit they're trying things
59:58
out on us so they're actually the algorithms are learning truth about human behavior
1:00:04
but then it's not it's not for public uh discussion it's it's buried in an algorithm that you'll never find that
1:00:10
just means that this business is better at advertising at this one and nobody will ever really know why
1:00:17
yeah that's a very scary situation yeah as you were talking I thought about the corporate strategy of doubling down
1:00:24
which seems to be very mathematically oriented that something that's obviously a lie can become true if you simply
1:00:31
multiply it more yes you know um and that
1:00:37
is why I think these questions around authenticity truth and reality have become so loaded
1:00:42
that in some ways they come to mean their opposite you know the more strenuously someone asserts something is
1:00:48
true in some ways the more you ought to be concerned about it
1:00:54
so here's a great question um for people who are teaching and even for people who are just
1:00:59
not or taking classes which is what recommendations you have for teaching current and future undergrad students
1:01:05
who've grown up in the environment of pervasive algorithms and have no other comparison which is our students um
1:01:12
this just moved I'm going to find it again uh oh sorry this chat is it it's it's moved
1:01:20
to answered oh it has it okay thank you um so how do we uh
1:01:25
work with students who don't have any basis for comparison so it's not as if there was a before and after for them
1:01:31
around before all my data was gathered and you know I was misrecognized in these painful ways
1:01:38
versus the old ways how does things skeptically about the environment especially when it feels
1:01:44
useful to have personalized ads or personalized recommendations so I'm thinking especially about Spotify Wrapped or
1:01:50
things that are genuinely pretty joyful for people you know how do you put that into
1:01:56
context so even the things that produce pleasure and don't seem to be so bad for people
1:02:02
who have no other moment to remember but there's moments
1:02:10
I think what's key is you start with those moments I mean there's this view somehow that these undergrads are
1:02:15
completely naive that you know they're they need to be taught something before after
1:02:21
um as a way to get to them to think critically about what they're doing but I think we need to respect the fact that
1:02:27
they do think critically and that but to to move that and start with that in
1:02:33
different ways so I used to teach a very large digital media class at Brown
1:02:38
and I would start off it was it drew students from all disciplines it was it was really a fun course to teach um and
1:02:46
I would always start with uh the ways in which they're already using these
1:02:51
these technologies critically right and so back then Snap was huge and you know I asked them how many people you know
1:02:57
they always Snap how many of them actually think that these images disappear and they all put up their head and said they didn't think it
1:03:03
disappeared and I said how many of you regardless though would would not send me something on Snap that and
1:03:11
they're all like yeah so it was trying to get to think through their practices already I mean if you're talking about
1:03:17
Instagram I mean clearly people who are gramming understand that they're characters within this drama called big
1:03:24
data that what's being captured and manipulated and thought through isn't this sort of simply behaviorist
1:03:32
knee-jerk reaction but rather their engagement with it as character so I always say like let's think of ourselves
1:03:39
as characters rather than marionettes which is how you know the social drama
1:03:44
presents us because the social drama has only a behaviorist understanding of humans right
1:03:50
um to go to that that rich depth and gaps that we know are part of the process and say use that as the
1:03:56
beginning and we think ourselves as characters which students completely understand
1:04:01
immediately you also understand that if you're a character you have a script you have agency but there are ways in
1:04:08
which things are controlled also the ways in which your performance is limited there's also the
1:04:15
the question of finance and the whole performance etc etc um but then I say as well what's so key about it is that um
1:04:23
your character but your actions and your script are determined by everybody else
1:04:29
um so I get everyone then to open up their phones or think through their technology and I say well the way this thing works
1:04:35
is um and I have them open TCP/IP and do a open terminal do a TCP/IP dump is that
1:04:42
they then see the ways in which their card is actually sucking in everybody
1:04:47
else's traffic and then only actively deleting what's addressed to them so the ways in which your their computer's
1:04:54
default mode is actually what's called promiscuous mode right so part of what I do is get them to open up their
1:05:00
technology and to think through their own actions and what they understand about this
1:05:05
um and and the limitations and possibilities of it when we we start thinking about the ways in
1:05:11
which um correlation for instance means that we're always correlated to others so how could that be actually a place
1:05:17
from which to intervene rather than think of ourselves as only completely manipulated
1:05:31
should I haven't actually taught in such a class I have been a professor of mathematics for many years and try to
1:05:37
bring real-world examples in and the closest I've gotten is statistics
1:05:43
and you just get to see how the different presentation of statistical facts affects
1:05:48
your your feelings about them and that's shocking to people in a math class perhaps you know uh
1:05:55
there's some wonderful books like Kathy O'Neil's book on uh Weapons of Math Destruction that
1:06:01
should be mentioned that certainly eye-opener for me it's not a it's not a math book though it's a it's a popular book so I
1:06:08
would love to know what textbooks there are that sort of bridge the gap uh that
1:06:13
for the undergrads but I don't think this is fundamentally new I mean it's a bit like
1:06:18
uh you know why don't we teach the undergraduates about the meat industry or some other highly uh
1:06:25
processed and and intricate uh system that the capitalist system that's
1:06:30
set up for our benefit you know supposedly for our benefit right but uh so you can buy cheap meat at the
1:06:36
supermarket but no one ever questions it they're like meat comes from the supermarket that's the same level of engagement that many you know
1:06:43
you have you know with Facebook let's say you want to like understand how is the meat
1:06:48
grown in the animals and the whole packaging and the distribution and the er- and the
1:06:54
environmental effects all of these things you need to do the equivalent of that for the social media I think
1:07:02
but also Lisa I know that you've been teaching and thinking through these issues
1:07:10
oh wow I don't know I feel like um as you say
1:07:16
just because people don't remember a before doesn't mean they're super critical about now
1:07:23
you know I think if anything students are far more critical about now and the issue for me in teaching is to kind of
1:07:29
hold back the incredible anger and kind of rage and frustration around
1:07:34
how this happened and try to strategize a so work
1:07:40
backwards and we're forward so remember where these things came from in a way that isn't just about hating or loving
1:07:47
who we see is in charge right because that's a tendency that you see in a lot of histories which I don't think is
1:07:53
great because people are rarely in control of what they do so it doesn't make sense to
1:07:59
find victims and heroes and all that um but also thinking forward in terms of yes this is horrible but let's get
1:08:06
really creative or really off the map around what could be the difference because at this point there's nothing to lose by being
1:08:12
totally unrealistic right like I think the moment for trying to be measured and thoughtful and what
1:08:18
can we actually achieve which I've been asking you is a kind of red herring you know I think as you say Indigenous
1:08:24
philosophy Afrofuturism assumes that the the apocalypse has already happened so
1:08:29
if that's the case kind of incremental or conservative fixes to things
1:08:36
you know like me not doing Facebook or trying to get cleaner data or let's just optimize you can't optimize your way out
1:08:42
of this because it's a kind of root and branch problem right so I think that social media
1:08:47
companies are like we're gonna do a better job with moderation and hire more moderators it does not matter how many
1:08:53
moderators you hire at this point like if every human in the US became a moderator it would kind of
1:08:59
be like Borges' Map of the world that's the same size as the map as the world you know like in some ways everyone is
1:09:06
already doing the work of moderation but they're just not doing a great job
1:09:11
they're just not agreeing about what that work ought to be so I think this idea of common sense which I've been
1:09:17
very interested in for a long time the moderation policy was common sense
1:09:23
which is to say use your personal ideology and we will back you up um because we've you know section 230 we
1:09:29
are not responsible you are responsible um so I think you know really
1:09:36
unreal you know unfeasible solutions are very interesting
1:09:42
in a new way to me for that reason yeah yeah Iisa Wendy here I think that what is so key about what you brought up
1:09:48
too is that we need to refuse the terms that we're given right so moderation makes it seem as if suddenly we're going
1:09:55
into content when in fact I mean and this is Tarleton Gillespie this is Sarah T
1:10:01
Roberts etc they've shown that it's always been moderated it's always been shaped
1:10:07
so the question of moderation itself is so beside what we're trying to take on
1:10:15
right and having AI do more of it is the totally the opposite way
1:10:22
for the for many of the reasons you say in your book right like then you're really tripling down quadrupling down
1:10:28
right multiplying the problem and just making it more pervasive
1:10:33
so I want to get some more questions because these are such good questions so there's a
1:10:39
question here around where a person ought to start in this book if they want to learn more around
1:10:45
authenticity the real and truth and how it relates to data and everyday lives
1:10:52
so the question uh sorry this is Wendy uh is where to start yeah what's a good moment or chapter that you would
1:10:59
identify as a place for a person just starting the book to focus first
1:11:05
if they were interested in questions of authenticity then I would direct them to
1:11:10
the the chapter on authenticity which is the third chapter um which brings together studies of reality tv and
1:11:18
training in that sense right which is reality tv is all about training
1:11:23
and trying to think through algorithmic authenticity and the relationship between the two so that's the third
1:11:29
chapter I mean the book is rather strange just to talk about structure a little we have these sort of meaty
1:11:35
chapters and then each starts with a little theoretical interlude that goes into some of the questions that were
1:11:41
raised as well as Alex's incredible mathematical illustrations so I think
1:11:47
that also um having those interludes and conversation with that chapter around authenticity
1:11:54
would be a great place to begin
1:12:01
yeah I'm going to throw in a question about randomization that's inspired by a chat question which has to do with the
1:12:08
price of of diversity in a way so you were talking Alex about how you know
1:12:15
part of what data analytics does is to reduce randomization so everything gets matched
1:12:21
to something else which is legible to that thing which tends to reproduce different kinds of problems um this
1:12:27
question is about the price for people say who are a marginalized group and who are
1:12:34
mixed in with people who are unlike them so you know what's the human price I guess of
1:12:41
heterophily or mixing when we've already created a situation that's so toxic and painful for people
1:12:48
who are increasingly marginalized and discriminated against in everything they're doing right like
1:12:54
going to school and trying to live in a neighborhood and trying to buy stuff right or get a mortgage
1:13:02
how do you compensate or help the people who are then
1:13:08
having to be the random in the randomization that's a great question I think what it
1:13:14
brings out is part of the problem with homophily isn't simply that it talks about similarity breeding connection is
1:13:20
that it makes users and people responsible for segregation right so homophily becomes a
1:13:28
synonym for discrimination right but it erases questions of institutional discrimination it becomes a homonym
1:13:36
for a community that erases the real work that needs to be in place for communities to emerge so one thing
1:13:42
that's absolutely wrong about homophily is this notion that it's simply voluntary and produced
1:13:48
and so it's only it erases questions of institutional racism etc because those
1:13:54
questions of institutional racism are also there at the the question of these quote-unquote solutions which is throw
1:14:00
in a diverse person and then suddenly it's the work of the diverse person to make everything better right um and
1:14:07
completely ignoring those questions of institutions and the ways that that
1:14:12
relies and what I think is really important is if you look at these historical cases and ask yourself what
1:14:19
could have emerged um so again this the notion of homophily actually comes from the preferences of white residents
1:14:25
living in a biracial housing project for friends who were like them who were liberal like them or illiberal like
1:14:32
them where liberal meant that you thought that um housing projects should be biracial and you thought the
1:14:39
race has gone along illiberals thought the opposite but there was this other category called ambivalent which thought
1:14:45
the races shouldn't be mixed but people got along and then there was also the fact that
1:14:50
they interviewed all the black residents so to come up with the term homophily they got rid of most of their data they
1:14:56
got rid of like all the black residents responses and they got rid of the white ambivalence which was also the largest
1:15:02
group right because they wanted the sort of intense um decision-making
1:15:07
but if you go to some of the responses of the black residents they said we would you know we're here because we're
1:15:14
finally getting equal housing for equal money um that we're here because finally you
1:15:21
know the what's being um done is what's fair
1:15:27
um and it's and they said you know we would welcome everyone here but we're not out there to say oh we need a white
1:15:32
person here to make it diverse right so I think that thinking through for me what's key is indifference
1:15:39
which is the ways in which you start you build structures of care outside of this logic of love or hate um
1:15:46
but rather through this this notion that we will have to live in difference and we live in difference um
1:15:53
in ways that can be caring and so that for me it's absolutely key and I know that you and I Lisa at least have been
1:15:59
thinking through the limitations as well as this notion of heterophily that somehow if you simply just love the
1:16:05
opposite this will solve everything um and it's key that you know
1:16:10
this idea of simply loving what's different is also what drives corporate models
1:16:16
yeah or what drives certain ways to quote-unquote expand um recommender systems yeah it's so easy
1:16:23
to fetishize the different too because it's pleasurable to have difference it's stimulating and fun and
1:16:30
everybody likes Alex you wrote something can you read it or speak to it yeah go ahead I did yeah sorry I kind of hit
1:16:36
the same question on text that you were alluding to but
1:16:42
I mean I would love to know more of the social science literature I'm not an expert on that but
1:16:48
on the student housing I have read quite a bit and Elizabeth Armstrong in 2006 and her colleagues
1:16:55
embedded themselves in a sorority which with very homogeneous uh sort of
1:17:01
uh basically a white midwestern sorority and study like mechanisms where competition
1:17:08
leads to more dangerous behavior in student life and so I don't on the top of my head
1:17:15
have a paper that shows sort of the randomization how randomization works but you need small communities here you
1:17:21
can't really say we're going to take the whole internet and randomize everybody because then you will get the noisiest
1:17:27
most racist you know hate speech will still like come to the top because of the way our algorithms
1:17:34
you know work you have to build I think small communities that have accountability uh
1:17:41
online just like you do doing they work in real life right uh you know human beings have been
1:17:48
fighting the urge to form gangs that kill each other over the
1:17:53
years you know which which happens right but you've got
1:17:58
to try and form small groups of uh where there isn't a power imbalance right but that's
1:18:04
it that's the other thing you can't randomize people uh balanced in power
1:18:11
so I don't know how you do that online you know
1:18:16
your friend group should be chosen you know the people that it suggests should be
1:18:21
complementary and uh rather than you know just more of the same
1:18:27
but also um this is where you to bring up like ethnic studies and you asked me about ethnic studies and Asian American
1:18:33
Studies has done a lot of work precisely around this you know what does it mean that the wound is actually the also the
1:18:39
the formation of community like how can we think through the the legacy the psychic legacies of
1:18:45
of um abuse and how can we as well the ways in which care emerges in these communities um and think
1:18:52
rigorously about them in ways that that don't devalue this and I think that thinking through those those modes of
1:18:58
care is absolute care and powers as Alex put it are absolutely key as well as the
1:19:04
deeper sort of psychical legacies well thank you so much for writing this
1:19:09
book I think it's wonderful for teaching um it demystifies a lot of what
1:19:14
computers are doing which we can always use more of and it's just beautifully written so thank you both so much and
1:19:21
the book is available everywhere and right when did you want to talk more
1:19:27
about how people can get it yeah um so it's available everywhere there is a discount code as well
1:19:34
nice yes but it's only for 10% off and it only works in the U.S.
1:19:40
um and so I'm trying to find the discount
1:19:46
well um while you're looking I think we can also push it out through our social
1:19:51
media in DISCO and Digital Democracies and if you love this talk you should
1:19:59
become involved in DISCO which is a growing network and we want to include you know of course junior scholars
1:20:05
Jessica's going to say something about it great we just want to share a screen right now
1:20:11
[Music] hi everyone this is Jessica Hill Riggs
1:20:18
speaking and I'm the program manager of the DISCO Network which is a co-sponsor of this event
1:20:24
and we wanted to end this book talk by sharing more information on how to get involved in the DISCO Network
1:20:30
DISCO stands for Digital Inquiry Speculation Collaboration and Optimism
1:20:37
network and we're a new research initiative funded by the Mellon Foundation that
1:20:42
explores the intersection of identity and technology to envision a more just equitable
1:20:49
anti-racist and anti-ableist digital future we're a consortium of five universities
1:20:55
and research labs at the University of Michigan Purdue University Stony Brook University and the University of
1:21:01
Maryland-College Park and Georgia Tech and as a network we offer fellowship
1:21:06
opportunities scholarly training mentoring and publishing opportunities and public programming such as lecture
1:21:14
series like this one and conferences on cutting-edge digital topics regarding identity and
1:21:20
technology and we'd love for you to get involved please visit our website
1:21:27
disconetwork.org to join our listserv to learn more about our network's upcoming events
1:21:32
and opportunities this spring and summer 2022 we're also active at Twitter @disconetowk_
1:21:40
so please tweet at us or please feel free to reach out to us via email
1:21:46
at disconetwork@umich.edu and thank you for joining us at this event
1:21:52
have a great rest of your day and I also just wanted to say thank you to Alan our ASL interpreter and to Katie for
1:21:59
providing live captioning services and thank you
1:22:05
yeah thank you and I have to say DISCO is one of the most exciting things that are is happening right now so that
1:22:10
everyone should join and participate and that this is you all are going to change the world
1:22:17
well thank you Wendy and Alex this was amazing thank you thanks everybody thanks for the great questions thanks