1
00:00:00,570 --> 00:00:04,933
MICHELLE: Apps with user-generated content
often rely on user policing.

2
00:00:04,933 --> 00:00:09,046
On Facebook, you can report
inappropriate or abusive content.

3
00:00:09,046 --> 00:00:12,630
The content review team reviews
the offending content and

4
00:00:12,630 --> 00:00:17,034
conducts disciplinary action if it
is deemed to violate the policies.

5
00:00:17,034 --> 00:00:21,768
The burden of finding harmful content
is shifted to users because Facebook

6
00:00:21,768 --> 00:00:26,593
doesn't want the responsibility of
proactively policing it themselves.

7
00:00:26,593 --> 00:00:29,836
But content moderation
is a colossal task and

8
00:00:29,836 --> 00:00:33,684
unsafe content can trigger
the people reporting it.

9
00:00:33,684 --> 00:00:39,265
In the book Behind the Screen, Sarah T.
Roberts explores the emotional toll that

10
00:00:39,265 --> 00:00:45,111
content moderators endure as they review
thousands of offensive material daily.

11
00:00:45,111 --> 00:00:49,890
According to a speech
Facebook CEO Mark Zuckerberg gave in 2019,

12
00:00:49,890 --> 00:00:52,326
there is some preemptive action.

13
00:00:52,326 --> 00:00:57,131
Facebook's AI systems identify
99% of the terrorist content

14
00:00:57,131 --> 00:01:00,109
which is blocked before anyone sees it.

15
00:01:00,109 --> 00:01:04,497
Per Facebook's reporting,
80% of all hate speech acted on

16
00:01:04,497 --> 00:01:08,573
by their platform was flagged
by algorithms in 2019.

17
00:01:08,573 --> 00:01:13,886
Granted, that doesn't include user flagged
content that results in no action,

18
00:01:13,886 --> 00:01:18,210
and the algorithms only cover
40 languages worldwide.

19
00:01:18,210 --> 00:01:22,776
The remaining languages only
have users reporting content and

20
00:01:22,776 --> 00:01:25,884
human moderators reviewing those flags.

21
00:01:25,884 --> 00:01:30,801
When asked why Facebook doesn't do more to
moderate its content, Zuckerberg argues

22
00:01:30,801 --> 00:01:35,583
the danger of limiting free speech as it
blocks out potentially positive voices for

23
00:01:35,583 --> 00:01:36,553
social change.

24
00:01:36,553 --> 00:01:40,745
Instead, he focuses on
eliminating fake accounts known

25
00:01:40,745 --> 00:01:44,161
as bots that spread misinformation and
hate.

26
00:01:44,161 --> 00:01:47,980
Facebook's AI systems can
now detect clusters of fake

27
00:01:47,980 --> 00:01:51,224
accounts that aren't behaving like humans.

28
00:01:51,224 --> 00:01:54,125
Government regulation can
force companies to do

29
00:01:54,125 --> 00:01:56,753
more to manage their platform's content.

30
00:01:56,753 --> 00:02:01,470
Germany has a law against hate speech,
and this has resulted in Facebook

31
00:02:01,470 --> 00:02:05,814
doing extra work to filter hateful
content from users in Germany.

32
00:02:05,814 --> 00:02:10,892
HOPE: Content heavy apps also struggle
with the spread of misinformation.

33
00:02:10,892 --> 00:02:16,334
Twitter has begun evaluating tweets
based on their propensity for harm.

34
00:02:16,334 --> 00:02:20,790
Tweets that contain potentially
misleading content such as this one

35
00:02:20,790 --> 00:02:25,564
mentioning COVID-19 are labeled
with a link to trusted information.

36
00:02:25,564 --> 00:02:29,435
Tweets suspected to be inaccurate,
not credible, or

37
00:02:29,435 --> 00:02:33,484
have more propensity for
harm are marked with a warning.

38
00:02:33,484 --> 00:02:37,542
Tweets confirmed to be false are removed.

39
00:02:37,542 --> 00:02:38,676
On a final note,

40
00:02:38,676 --> 00:02:44,170
online harassment is rampant across
social media apps and forums.

41
00:02:44,170 --> 00:02:49,103
We don't have time to get into it,
but there's cyberstalking,

42
00:02:49,103 --> 00:02:54,494
fake impersonations, catfishing,
doxxing, trolling, and so on.

43
00:02:54,494 --> 00:02:58,513
Block Party is an app to
help with online harassment.

44
00:02:58,513 --> 00:03:03,894
It mutes accounts on social media that
are likely to send unwanted content.

45
00:03:03,894 --> 00:03:09,693
The lockout folder contains the filtered
content and is accessible anytime.

46
00:03:09,693 --> 00:03:15,334
There's also the Helper View which
enables a user's friends to review their

47
00:03:15,334 --> 00:03:21,333
lockout folder content and take action
on their behalf, such as blocking users.

48
00:03:21,333 --> 00:03:24,146
Check out the teacher's notes for
further reading.

49
00:03:24,146 --> 00:03:29,110
While many apps have anti-harassment
rules, they are inconsistently enforced.

50
00:03:29,110 --> 00:03:32,986
We need better policies, procedures,
and tools to mitigate and

51
00:03:32,986 --> 00:03:34,897
prevent harassment and abuse.

52
00:03:34,897 --> 00:03:38,910
Even more, we need people like you
to advocate for change within.