cliff lampe 02 — Transcript

Full transcript

1. 0:00hi this is cliff lampe again from the
2. 0:02school of information
3. 0:04another thing people often ask me when
4. 0:06it comes to what's the relationship
5. 0:07between social media and misinformation
6. 0:09is why don't the social media platforms
7. 0:11just stop this why don't facebook or
8. 0:13twitter just say
9. 0:14enough we're going to police our
10. 0:15platforms and have no more information
11. 0:17there's a variety of reasons for this i
12. 0:19want to get into a couple of them
13. 0:20because some of them are technical
14. 0:21and some of them are legal from the
15. 0:23legal side it's important to know about
16. 0:25a piece of legislation called the
17. 0:26communications decency act
18. 0:28of 1996. and especially section 230 of
19. 0:32that act
20. 0:33most people these days just shorthand
21. 0:35that as cda section 230.
22. 0:37now what that does is it indemnifies
23. 0:39platforms
24. 0:40from being responsible for the content
25. 0:42that they carry
26. 0:44back in 1996 there were not very many
27. 0:46social media platforms of course
28. 0:48it was to protect the big telecom
29. 0:50companies from being responsible
30. 0:51for anything people sent over the wires
31. 0:54that they were using to deliver
32. 0:56internet at that time because platforms
33. 0:59like twitter and facebook were
34. 1:00determined to be platforms and not
35. 1:02services
36. 1:02they became under the jurisdiction of
37. 1:05section 230
38. 1:06and that means that there's no legal
39. 1:08recourse to hold them responsible
40. 1:10for the information that they share on
41. 1:12their platforms so that's the legal side
42. 1:14of this there's a
43. 1:15really strong well-tested law in the u.s
44. 1:17legal system section 230 of the cda
45. 1:20that protects platforms from being
46. 1:21responsible for the content
47. 1:23meaning that there's no legal way to
48. 1:25hold them responsible
49. 1:26the other effect of this legal
50. 1:28protection that they have is that the
51. 1:30platforms don't really want to be
52. 1:31arbiters of what's true or not
53. 1:33we've heard mark zuckerberg of facebook
54. 1:35say this pretty often where he doesn't
55. 1:36think it's appropriate for facebook to
56. 1:38decide
57. 1:38what is good information or what is bad
58. 1:40information all of the platforms have
59. 1:42been struggling recently for what is the
60. 1:44line
61. 1:45where they define what is going to be
62. 1:47misinformation versus
63. 1:49wrong simply or what's uh what's
64. 1:52misinformation versus being somebody's
65. 1:53perspective or opinion
66. 1:55these are very fine details and facebook
67. 1:58nor twitter
68. 1:59really don't want to be responsible for
69. 2:01determining what's true for everybody
70. 2:03uh partially because people get really
71. 2:04mad at them when people when they do
72. 2:06just try to decide to do that
73. 2:08so most of the companies now are
74. 2:09deciding facebook for instance is
75. 2:11deciding that
76. 2:12something is misinformation if the
77. 2:14information be presented is false
78. 2:16and could potentially hurt somebody
79. 2:18twitter has a slightly different
80. 2:19guideline from that
81. 2:20instagram which is owned by facebook has
82. 2:22a slightly different guideline from both
83. 2:24so all the platforms are making these
84. 2:25decisions now the other big challenge
85. 2:27that platforms have
86. 2:28is a technical hurdle it's hard to
87. 2:30imagine the scale of posting that occurs
88. 2:32on platforms like youtube and facebook
89. 2:35on youtube for instance more video is
90. 2:37uploaded every minute than a person
91. 2:39could watch in a year
92. 2:40right it's impossible for humans to
93. 2:43police the amount of content
94. 2:44that is appearing on these platforms
95. 2:46facebook has hired 40
96. 2:48000 human content moderators across the
97. 2:51world
98. 2:52and they are barely able to hold to keep
99. 2:54up with the reports that they get from
100. 2:56other folks
101. 2:57so facebook and other social media
102. 2:58platforms are depending on two major
103. 3:00mechanisms to determine how to
104. 3:02fight against kind of the misinformation
105. 3:04plague that they're seeing
106. 3:06the first is they depend on people
107. 3:07reporting so when you see bad
108. 3:09information and you report it
109. 3:10that goes into a hopper which then gets
110. 3:12reported to their human content
111. 3:14moderators and eventually is reviewed
112. 3:17system is abused a lot as people report
113. 3:19information as being false that they
114. 3:20simply disagree with
115. 3:21and so there's a lot of sorting out and
116. 3:24determination that people have to
117. 3:25determine what's false or not so it's
118. 3:27really hard at scale for that system to
119. 3:29keep up with the bad information
120. 3:32so the other mechanism that the
121. 3:33platforms are really trying to depend on
122. 3:34is artificial intelligence or machine
123. 3:36learning algorithms which are basically
124. 3:38automated ways to go out and try to
125. 3:40detect bad information
126. 3:42so you can go out and you can see
127. 3:44certain keywords or there's a
128. 3:46different types of approaches that
129. 3:47involve network approaches who's who's
130. 3:49retweeting whom or who's
131. 3:51connected to which site or who's also
132. 3:53posted here there's a lot of different
133. 3:55ways to do that
134. 3:56all of them right now are still pretty
135. 3:58bad so all of the machine learning
136. 4:00mechanisms the automated robot ways of
137. 4:02determining bad information that the
138. 4:04platforms have
139. 4:05still aren't that great and so the
140. 4:07platforms are really struggling with how
141. 4:08to
142. 4:09make that work for them so those are the
143. 4:12two major reasons why platforms can't
144. 4:14solve this problem for themselves
145. 4:16right one is a legal one they simply
146. 4:18don't have the incentive to do so
147. 4:20because they don't want to be arbiters
148. 4:21of the truth and there's no legal
149. 4:23recourse to
150. 4:24force them to do that the second is a
151. 4:26technological hurdle
152. 4:27they work at these absolutely massive
153. 4:29scales and
154. 4:30at those scales human moderation can't
155. 4:32keep up and machine learning moderation
156. 4:34or robotic moderation still
157. 4:36isn't sophisticated enough to really
158. 4:38solve the problem