0.0
You have followers you have never met. 00:00:04,562 --> 00:00:09,011 1.5s] This is not an observation. It is
9.011
not a complaint about social media. It is
11.692
a statement of fact so universal that you
14.376
have stopped questioning it. You have followers you
17.848
have never met, whose profile pictures you have
21.005
never examined, whose usernames you have never read
24.477
aloud, whose existence you have accepted the way
27.712
you accept furniture in a room you walk
30.236
through every day. They are there. They have
34.576
always been there. You do not know when
36.978
they arrived. I want you to
43.036
do something right now. Not later. Not after
45.467
this video. Right now. Open your phone. Go
47.767
to your follower list. Not your following list
50.329
— your follower list. The people who chose
52.629
to see your content. Scroll past the names
54.928
you recognize. Past your friends. Past your family.
57.819
Past the accounts you vaguely remember following you
60.776
back after you followed them. Keep scrolling. You
67.01
will find them in the middle. Not at
69.25
the top — those are recent. Not at
71.336
the bottom — those are old friends. In
73.731
the middle. A cluster of accounts that share
76.59
a specific set of characteristics so consistent that
80.067
once you see the pattern, you will not
82.462
be able to unsee it. The profile picture
86.505
is a real photograph. Not AI-generated — real.
89.504
A real person in a real location with
91.811
real lighting and real imperfections. The kind of
95.041
photograph that was taken between two thousand twelve
98.579
and two thousand eighteen, when smartphone cameras were
102.27
good enough to be clear but not good
104.5
enough to be cinematic. The bio contains exactly
109.51
three to five emoji. A hobby. A relationship
112.985
status or family reference. A single inspirational word
117.492
or phrase. The bio reads like it was
120.215
written by a human being. Because it was.
123.408
Once. The account follows between eight hundred and
128.435
fifteen hundred other accounts. It has between two
131.808
hundred and a thousand followers of its own.
134.708
It has posted between eight and thirty times.
137.689
The posts are photographs — meals, sunsets, pets,
140.982
children at birthday parties, a vacation beach. And
145.901
the last post is between three and ten
148.006
years old. In two thousand twenty
153.941
four, a cybersecurity research team at the University
157.003
of Amsterdam published a paper that received almost
159.931
no mainstream coverage. The paper was titled "Coordinated
163.258
Inauthentic Persistence: Dormant Account Networks and Post-Mortem Digital
167.65
Activity." The title alone should have made headlines.
170.778
It did not. The Amsterdam team had developed
175.147
a behavioral clustering algorithm that could identify coordinated
179.849
account networks not by what the accounts posted,
183.252
but by the temporal pattern of their micro-interactions.
187.225
Likes. Follows. Brief profile visits. The invisible actions
191.44
that leave no visible trace on anyone's feed
194.44
but are recorded in platform telemetry. They analyzed
199.754
eleven million accounts across three platforms over a
203.809
fourteen-month period. Their algorithm identified what they called
209.011
"dormancy swarms" — clusters of accounts that had
212.713
stopped posting original content but continued to perform
217.121
micro-interactions in synchronized patterns. The swarms were enormous.
223.764
The smallest contained eight hundred accounts. The largest
227.305
contained over forty thousand. And they were coordinated
230.707
with a precision that eliminated any possibility of
233.763
coincidence. Every account in the swarm had posted
239.484
original content at some point. Every account had
244.074
a real profile picture. Every account had a
248.009
bio that read like a human being wrote
251.397
it. And every account had stopped posting between
255.987
three and ten years ago. Not deactivated. Not
260.141
deleted. Just… stopped. But they had not stopped
265.332
interacting. The accounts continued to follow new users.
268.752
Continued to like posts. Continued to perform the
271.683
invisible micro-actions that social media algorithms interpret as
275.733
signals of an engaged, authentic audience. And here
280.37
is the detail that made the Amsterdam researchers
283.7
request additional security clearance before publishing their findings.
290.275
The accounts were not following random
295.647
users. They were following specific users. Users who
299.453
had recently been identified by advertising algorithms as
303.683
"high-influence micro-targets" — ordinary people with small but
308.42
highly engaged audiences whose purchasing decisions ripple outward
313.411
through their social networks. The dormant accounts were
319.052
being aimed. Not scattered like seeds. Aimed like
322.598
weapons. Someone was paying for this. Someone was
328.014
operating these swarms. Someone had access to thousands
332.587
of dormant accounts with real histories, real photographs,
337.446
real bios — and was deploying them in
340.305
coordinated campaigns targeting specific individuals. The Amsterdam team
347.715
traced the command infrastructure through fourteen layers of
351.825
proxy servers, three cryptocurrency mixing services, and a
355.78
shell company registered in the Seychelles. At the
359.115
end of the chain, they found a marketplace.
361.906
Not on the dark web. On the regular
364.078
internet. A website with a clean design, professional
367.645
copy, and a pricing page. The marketplace sold
372.13
access to dormant social media accounts in bulk.
375.218
The pricing was tiered by account age, follower
378.231
count, and what the marketplace called the "trust
381.395
coefficient." And the product descriptions used a term
386.75
that the researchers had never encountered before. 00:06:32,889 --> 00:06:42,815 2.5s] "Heritage accounts." Heritage accounts. The word "heritage"
402.815
implies inheritance. It implies something passed down. Something
407.452
left behind by someone who is no longer
410.056
here to use it. The Amsterdam researchers noted
414.331
the terminology in their paper without further comment.
417.416
They were cybersecurity specialists, not investigators. They documented
421.528
the technical infrastructure, published their findings, and moved
425.255
on to other projects. But one member of
429.157
the team did not move on. A doctoral
431.737
student named Asha Mertens, who had been responsible
435.739
for the manual verification phase of the research
439.474
— the part where a human being actually
442.321
looked at the accounts, one by one, to
445.078
confirm that the algorithm's classifications were accurate. Asha
451.579
Mertens looked at four thousand two hundred accounts
454.804
over the course of three months. And she
457.17
noticed something that the algorithm was not designed
460.467
to detect. The profile pictures matched
467.676
obituaries. Asha Mertens did not set
474.689
out to cross-reference social media profiles with death
479.147
records. She was verifying account authenticity — confirming
484.07
that the profiles identified by the clustering algorithm
488.621
were real accounts with real histories, not recently
492.8
fabricated imitations. But verification requires looking. And Asha
499.236
Mertens was thorough. The first match was Robert
504.296
Calloway. She found his obituary on the second
507.959
page of a Google search for his name
510.682
and hometown, which were both visible in his
514.157
social media profile. The obituary was from two
517.913
thousand nineteen. His account had liked fourteen posts
522.421
in the past month. She told herself it
526.648
was a coincidence. Someone with the same name.
529.86
A common face. A mistake. The second match
534.134
was a woman named Patricia Huang. Died in
536.669
two thousand seventeen. Her Instagram account had followed
540.471
thirty-seven new users in the past quarter. The
544.918
third match was a teenager named Devon Williams.
547.503
Killed in a car accident in two thousand
549.582
sixteen. His Twitter account had liked a cryptocurrency
552.606
promotion four days ago. By the time Asha
556.212
Mertens had cross-referenced three hundred of the four
559.041
thousand two hundred accounts in her verification sample,
562.051
she had confirmed forty-seven direct matches between active
565.181
dormant accounts and published obituaries. Forty-seven dead people
570.403
whose social media accounts were actively engaging with
573.68
the living internet. Not in a metaphorical sense.
578.268
Not in the way we say someone "lives
580.515
on" through their social media presence. In the
583.613
operational, technical, server-log-verified sense. These accounts were being
588.958
accessed. Commands were being issued through them. They
592.676
were following, liking, and in some cases commenting
596.162
— generic comments, single emoji, the kind of
599.106
interaction that algorithms reward but humans rarely examine.
604.789
The dead were participating in the internet. And
607.84
no one had noticed because no one looks
610.222
at their follower list the way Asha Mertens
612.901
looked at hers. She expanded her
619.229
methodology. Instead of manually searching for obituaries, she
623.399
built an automated cross-referencing tool that compared profile
627.644
photographs against digitized obituary databases, memorial websites, and
632.571
genealogy platforms. The tool used facial recognition —
636.21
not the sophisticated real-time systems used by law
639.546
enforcement, but a simple image-matching algorithm that compared
643.867
profile pictures against photographs published in death notices.
649.688
She ran it against the full dataset of
651.886
dormant accounts identified by the Amsterdam clustering algorithm.
656.068
Eleven million accounts. Three point two percent. Of
660.847
eleven million dormant accounts identified as part of
664.285
coordinated inauthentic swarms, three point two percent belonged
668.545
to people who were verifiably dead. That is
672.702
three hundred and fifty-nine thousand accounts. 175 00:11:20,567 --> 00:11:24,359 Three hundred and fifty-nine thousand dead people, active
684.359
on social media. Following. Liking. Commenting. Shaping algorithms.
688.91
Influencing what the living see, read, and believe.
693.747
And that was only the accounts Asha Mertens
696.5
could verify — the ones whose obituaries were
699.407
digitized and publicly accessible. The true number, she
703.078
estimated in a supplementary analysis that she never
706.519
published, could be between two and five times
709.502
higher. Because not everyone gets an obituary. Not
712.791
everyone's death notice is digitized. Not every country
716.462
maintains accessible records. The conservative estimate: three hundred
722.819
and fifty-nine thousand. The realistic estimate: over a
728.182
million. The question that Asha Mertens could not
733.066
answer — the question that drove her to
735.626
work eighteen-hour days for eleven weeks until her
739.066
academic advisor intervened — was not how. The
742.186
how was straightforward. Abandoned accounts with weak passwords,
746.746
accounts linked to email addresses that were themselves
750.586
abandoned after the owner's death, accounts on platforms
754.506
that had no mechanism for reporting a user's
757.466
death and removing their profile. The how was
760.506
a failure of infrastructure. A gap in the
763.193
system that no one had bothered to close
765.711
because no one had realized it was a
767.924
door. The question was why. Why specifically target
775.525
dead people's accounts? Why not simply create new
779.045
fake accounts, as bot farms had done for
781.811
years? Why go to the trouble of identifying
784.828
deceased users, gaining access to their profiles, and
788.684
reanimating them? The answer was on the pricing
793.004
page of the marketplace. In the phrase that
795.223
Asha Mertens would circle in red ink and
797.257
pin to the center of her corkboard. 00:13:21,603 --> 00:13:27,304 2.0s] "Average trust coefficient: ninety-four point seven percent."
808.804
They are using the dead because the dead
811.039
are trusted. Every social
822.358
media platform maintains a system that it does
825.529
not publicly acknowledge. The terminology varies — "credibility
830.083
index," "authenticity rating," "behavioral trust metric" — but
834.556
the function is identical. Every account is assigned
838.215
a score. The score determines how the platform
841.387
treats that account's actions. A new account —
846.056
created today, with no posts, no followers, no
849.217
history — has a trust score near zero.
851.73
Its likes carry no weight. Its follows trigger
854.891
spam filters. Its comments are shadow-suppressed. The platform
859.349
treats it as guilty until proven innocent, because
862.834
the platform has learned, through years of bot
865.995
warfare, that new accounts are overwhelmingly fake. An
871.282
account created in two thousand twelve by a
873.784
human being who used it for six years
875.869
— who posted photographs of their children, who
878.65
argued about politics, who left a birthday comment
881.638
on their sister's wall every March, who misspelled
884.627
words and used the wrong emoji and exhibited
887.199
all the beautiful, chaotic inconsistency of a real
890.188
human life — that account has a trust
892.273
score that approaches the theoretical maximum. It is
896.924
algorithmically invisible. Its actions pass through every filter.
901.281
Its likes register as genuine engagement. Its follows
904.737
are counted as organic growth. Its comments appear
907.968
without delay, without review, without the invisible hand
911.724
of moderation touching them. And when that human
916.388
being dies, the score does not die with
918.96
them. The score persists. The account
927.107
persists. The history persists. And the trust —
930.877
that precious, painstakingly accumulated trust — sits there.
935.873
Unguarded. Unmonitored. A vault with no lock, in
939.737
a house with no owner, on a street
942.282
where no one is watching. This is the
946.454
market. Not a metaphor. A literal marketplace with
949.76
buyers, sellers, and a commodity that replenishes itself
953.528
every time someone dies without deleting their social
957.065
media accounts. Asha Mertens' investigation eventually led her
962.167
to three distinct tiers of the heritage account
964.631
trade. Tier One is the bulk market. Low-cost
968.954
packages of dormant accounts sold to influencer marketing
972.91
agencies, small businesses, and social media managers who
976.866
need to inflate follower counts. These accounts follow,
980.664
occasionally like, and never comment. They are the
984.067
foot soldiers — the background noise of artificial
987.469
engagement. A package of five hundred costs less
990.713
than three hundred dollars. The buyers rarely ask
994.037
where the accounts come from. The sellers never
997.202
volunteer the information. Tier Two is the amplification
1002.911
market. Mid-range packages of high-trust dormant accounts sold
1007.992
to political campaigns, cryptocurrency promoters, and disinformation networks.
1014.551
These accounts do not merely follow — they
1017.784
engage. They like specific posts at specific times
1021.757
to trigger algorithmic amplification. They follow specific users
1027.023
to manipulate recommendation algorithms. A coordinated action by
1031.987
two thousand heritage accounts with trust scores above
1035.619
ninety can push a post from obscurity to
1038.169
a platform's trending feed in under four hours.
1042.76
Tier Three is the one that Asha Mertens
1044.41
almost did not include in her research because
1046.421
she was not certain anyone would believe her.
1049.88
Tier Three is the identity market. Individual heritage
1053.777
accounts — not bulk, not packages, but single
1056.927
accounts — sold to buyers who need a
1059.332
specific type of digital identity. A middle-aged woman
1063.228
from the Midwest. A college student from London.
1066.628
A retired engineer from São Paulo. The buyer
1069.695
specifies the demographic, the location, the age range,
1073.675
the interests. The seller delivers a real account,
1077.24
with a real history, belonging to a real
1079.976
person who is really dead. The price for
1083.931
a Tier Three account ranges from two thousand
1086.11
to fifteen thousand dollars, depending on the account's
1088.863
age, engagement history, and the completeness of the
1091.444
deceased owner's digital footprint. Fifteen thousand
1098.881
dollars for a dead person's identity. Not their
1102.301
Social Security number. Not their bank account. Their
1106.234
social media presence. Their digital face. Their accumulated
1110.766
trust. And the buyers at Tier Three are
1115.061
not marketers. They are not political operatives. They
1119.185
are not influencer agencies. They are AI training
1124.579
networks. The most sophisticated large language
1131.708
models — the ones that generate text, analyze
1134.345
sentiment, produce content that is indistinguishable from human
1138.231
writing — are trained partially on social media
1141.007
data. The models learn what human communication looks
1144.2
like by studying billions of examples of human
1146.906
communication. But as the internet has filled with
1151.904
synthetic content — AI-generated text, bot interactions, machine-produced
1157.654
engagement — the training data has become contaminated.
1161.835
Models trained on contaminated data produce contaminated output.
1166.8
The industry calls this "model collapse" — a
1170.023
recursive degradation where AI trained on AI output
1173.856
becomes progressively less human with each generation. The
1179.777
solution, for certain operators, is to ensure that
1183.203
training data comes from verified human sources. And
1186.788
the most verified human sources on the internet
1189.975
are the accounts with the highest trust scores.
1193.162
The accounts that platforms have spent years confirming
1196.987
are real, authentic, and human. The accounts of
1201.628
the dead. The dead are training the machines
1205.403
that will speak for the living. 313 00:20:11,517 --> 00:20:16,471 Her name is Linda Ortega. She
1216.471
is fifty-three years old. She lives in a
1219.214
two-bedroom apartment in Albuquerque, New Mexico, with a
1223.285
tabby cat named Professor and a refrigerator covered
1227.025
in photographs held up by magnets from places
1230.182
she has visited with her son. Her son's
1234.259
name was Marcus. He was twenty-four when he
1236.883
died. Acute lymphoblastic leukemia. The diagnosis came in
1240.528
January of two thousand twenty. The treatment lasted
1243.808
eleven months. Marcus died on December second, two
1246.943
thousand twenty, in a hospital room with white
1249.786
walls and a window that faced the parking
1252.264
lot. Marcus had an Instagram account. He posted
1256.828
photographs of sunsets, his friends, his cat before
1260.215
Professor — a calico named Doctor who died
1262.91
two years before Marcus did. His last post
1265.605
was from September of two thousand twenty. A
1268.453
sunset photographed from his hospital room window. The
1272.072
caption read: "Not bad for a Tuesday." Not
1276.333
bad for a Tuesday. After Marcus died, Linda
1281.174
did not touch his account. She did not
1283.917
delete it. She did not memorialize it. She
1287.014
did not even log in. The account existed
1289.934
as Marcus had left it — a small,
1292.146
honest archive of a young man who liked
1294.977
sunsets and cats and had a dry sense
1297.543
of humor about dying. Linda sometimes opened Instagram
1302.9
and looked at Marcus's profile. Not every day.
1305.945
Some weeks, not at all. But when she
1308.21
did, she scrolled through his posts the way
1311.021
you might turn the pages of a photo
1313.207
album. Slowly. With the kind of attention that
1316.253
only grief can produce. On March
1322.752
fifteenth, two thousand twenty-four — three years and
1326.283
three months after Marcus died — Linda received
1329.353
a notification on her phone. Marcus_sunsets liked a
1335.367
post. Linda tapped the notification. Instagram opened. The
1341.775
activity log showed that marcus_sunsets had liked a
1345.831
sponsored post from an energy drink brand called
1349.611
VoltRush. The post was a photograph of a
1352.654
muscular man running on a beach with the
1355.696
caption "Fuel Your Fire 🔥 #VoltRush #Energy #NeverStop."
1361.714
Marcus — her Marcus, who spent his last
1364.22
months too weak to walk to the bathroom
1366.726
without help, who joked about sunsets because he
1369.937
was not sure how many more he would
1372.13
see — had liked a post about fueling
1374.401
your fire. About never stopping. The algorithm did
1379.35
not know it was being cruel. The algorithm
1382.282
does not know anything. It was executing a
1385.213
task. A heritage account designated marcus_sunsets had been
1389.568
assigned to a Tier Two amplification campaign for
1393.086
a beverage company's product launch. The campaign required
1397.357
twelve thousand likes from high-trust accounts within a
1401.377
six-hour window. Marcus's account — trust score ninety-three
1405.266
point four, created two thousand seventeen, last original
1408.73
post two thousand twenty, no red flags, no
1411.154
irregularities — was one of twelve thousand accounts
1414.271
activated for the campaign. Linda Ortega
1421.634
reported the account. She clicked "Report," selected "This
1425.531
account may be hacked," filled out the form,
1428.359
and submitted it. She received an automated response
1431.797
within thirty seconds: "Thanks for your report. We'll
1435.313
review this and take action if we find
1437.682
a violation of our Community Guidelines." Three weeks
1442.947
later, the account was still active. Still liking.
1447.309
Still following. Still performing. She reported it again.
1453.824
Same automated response. Same result. She tried to
1459.301
recover the account — to log in as
1461.269
Marcus, to change the password, to do anything
1464.111
to make it stop. But the email address
1466.371
linked to Marcus's account was his university email,
1469.651
which had been deactivated six months after his
1472.566
death. The recovery process required access to that
1475.773
email. Without it, the platform's security system —
1478.98
the same system designed to prevent unauthorized access
1482.479
— prevented Linda from reaching her own son's
1485.249
account. The system that could not stop a
1488.982
bot network from operating Marcus's account could very
1491.964
effectively stop his mother from shutting it down.
1496.192
She contacted support. She waited fifteen
1501.965
business days. She received a response requesting a
1505.378
death certificate. She mailed a death certificate. She
1509.023
waited twenty-two more business days. She received a
1512.514
response saying the death certificate had been received
1516.237
and the case was "under review." During those
1521.32
thirty-seven business days, marcus_sunsets liked eighty-four posts, followed
1530.66
nineteen new accounts, and commented on three posts
1536.615
with single emoji — a fire emoji, a
1540.405
heart emoji, and a thumbs up emoji. 00:25:52,016 --> 00:26:02,746 2.0s] Eighty-four likes. Nineteen follows. Three comments. In
1562.746
the voice of her dead son. While she
1565.158
waited for a corporation to acknowledge that he
1568.485
was dead. On day forty-one, the account was
1573.21
finally memorialized. The word "Remembering" was added before
1578.147
Marcus's name. The profile was locked. No more
1581.711
likes. No more follows. No more comments. But
1586.643
Linda Ortega does not use the word "memorialized."
1589.978
In the interview she gave to a local
1592.227
Albuquerque news station — an interview that was
1595.407
aired once, at eleven PM, between a weather
1598.199
report and a used car advertisement — she
1600.836
used a different word. She said they held
1604.855
his account hostage. She said the internet made
1609.51
her son work after he died. She said
1613.477
she had to prove he was dead to
1615.422
a machine that already knew he was dead
1618.015
and did not care. Linda Ortega's
1624.562
story is not unique. It is not even
1626.572
rare. A two thousand twenty-five survey conducted by
1629.802
the Digital Legacy Alliance — a nonprofit organization
1633.175
that advocates for posthumous digital rights — found
1636.405
that fourteen percent of respondents who had lost
1639.419
a family member in the past five years
1641.645
had observed unexpected activity on the deceased person's
1645.233
social media accounts. Fourteen percent. One in seven
1651.674
grieving families. Watching their dead interact with a
1654.424
world that has moved on without them. Watching
1656.706
algorithms puppeteer the digital remains of the people
1659.456
they loved. Watching and being unable to stop
1661.68
it because the systems designed to protect accounts
1664.254
from unauthorized access cannot distinguish between a mother
1667.355
trying to lay her son to rest and
1668.876
a hacker trying to steal his identity. The
1672.468
dead have more rights on social media than
1675.035
the living who mourn them. 00:28:01,421 --> 00:28:09,765 3.0s] I have a request. Not
1689.765
a suggestion. Not a rhetorical exercise. A request
1693.172
that I am making of you specifically, right
1696.025
now, in this moment, because you have spent
1698.878
twenty-eight minutes understanding something that cannot be un-understood.
1705.687
Pick up your phone. Open your social media.
1710.389
Any platform. The one you use most. The
1713.209
one where you have the most followers. The
1716.293
one you think you know. Go to your
1720.045
follower list. Scroll past the names you recognize.
1725.014
Past your friends. Past your family. Past the
1728.112
people you actually know. Keep scrolling. 451 00:28:58,165 --> 00:29:00,618 You will find an account. Maybe more than
1740.618
one. An account with no profile picture, or
1743.216
a profile picture that was taken years ago.
1745.814
An account that follows eight hundred people and
1748.772
has forty-three followers of its own. An account
1751.731
that has not posted since two thousand eighteen
1754.617
or two thousand nineteen. An account that watched
1758.801
your story yesterday at three in the morning.
1762.385
They did not watch it. 00:29:29,529 --> 00:29:34,631 1.5s] The person who owned that account was
1774.631
buried in two thousand nineteen. Their name was
1778.176
Elaine. She was thirty-one. She liked hiking and
1781.81
terrible puns and she had a dog named
1784.469
Biscuit who outlived her by two years. She
1787.571
posted her last photograph on a Tuesday —
1790.584
a trail somewhere in Oregon, the light coming
1793.952
through the trees in columns, the caption a
1797.143
single word: "Breathe." She does not breathe anymore.
1803.624
But her account does. Her account follows. Her
1806.935
account likes. Her account watches your stories at
1810.585
three in the morning because the server farm
1813.726
in Bucharest that operates her profile runs its
1817.121
engagement cycles during off-peak hours when the algorithmic
1821.621
scrutiny is lowest. You posted a photograph of
1826.483
your dinner last Tuesday. Elaine liked it. You
1829.886
saw the notification and did not think about
1833.115
it. You did not recognize the name. You
1835.907
did not click on the profile. You accepted
1838.962
the like the way you accept air —
1841.231
automatically, unconsciously, as a feature of the environment
1845.943
you inhabit. You are performing for an audience
1851.08
of corpses. Every like you have
1857.949
ever received may include likes from the dead.
1861.179
Every follower count you have ever checked includes
1864.825
the dead. Every metric you have ever used
1867.641
to measure your relevance, your reach, your value
1871.121
as a human being in the digital attention
1873.937
economy includes the dead. The platforms know this.
1878.701
They have always known this. They do not
1880.836
remove dormant accounts because dormant accounts inflate the
1884.264
platform's user metrics. A platform with two billion
1887.176
accounts can report two billion users to advertisers,
1890.152
to investors, to the public. It does not
1892.287
matter that millions of those accounts are operated
1895.133
by no one. It does not matter that
1896.88
hundreds of thousands are operated by the dead.
1899.468
The number goes up. The stock price follows.
1903.362
You are not the customer. You
1908.241
are not the product. You are the living
1910.906
half of an audience that includes the dead,
1913.904
and the platform profits from both halves equally
1917.401
because to an algorithm, engagement is engagement. A
1921.149
like is a like. A follow is a
1922.981
follow. It does not matter whose thumb pressed
1926.229
the button. It does not matter if there
1930.025
was a thumb at all. The next time
1933.285
you pick up your phone. The next time
1935.368
you check your notifications. The next time you
1938.146
see that someone liked your post, watched your
1940.855
story, followed your account. Ask yourself one question.
1947.641
Are they alive?
