1
00:00:04,400 --> 00:00:07,800
Speaker 1: Welcome to tex Stuff, a production from I Heart Radio.

2
00:00:12,200 --> 00:00:15,040
Speaker 1: Hey there, and welcome to tech Stuff. I'm your host,

3
00:00:15,280 --> 00:00:18,840
Speaker 1: Jonathan Strickland. I'm an executive producer with I Heart Radio,

4
00:00:18,880 --> 00:00:22,360
Speaker 1: and I love all things tech and I don't get

5
00:00:22,360 --> 00:00:25,720
Speaker 1: to do this very often, so indulge me. But in

6
00:00:25,840 --> 00:00:33,080
Speaker 1: Shakespeare's classic tragedy Hamlet, our protagonist and perpetually moody Dane

7
00:00:33,520 --> 00:00:37,760
Speaker 1: says at one point there is nothing either good or bad,

8
00:00:38,040 --> 00:00:42,720
Speaker 1: but thinking makes it so. What he means there is that,

9
00:00:42,880 --> 00:00:45,680
Speaker 1: at least in his eyes, all things in the universe

10
00:00:46,120 --> 00:00:49,960
Speaker 1: are just as they are. They are neither good nor bad.

11
00:00:50,440 --> 00:00:53,640
Speaker 1: It's only how we think about them that makes them

12
00:00:53,840 --> 00:00:56,560
Speaker 1: good or bad. This is similar to a philosophy and

13
00:00:56,560 --> 00:00:59,720
Speaker 1: I've talked about on this show when it comes to technology.

14
00:01:00,280 --> 00:01:03,920
Speaker 1: Technology and of itself has the capacity to help or

15
00:01:03,960 --> 00:01:08,480
Speaker 1: to harm. But the underlying tech isn't necessarily good or evil.

16
00:01:08,840 --> 00:01:12,959
Speaker 1: It's how it's put to use that makes all the difference. Now, granted,

17
00:01:13,319 --> 00:01:16,640
Speaker 1: there are some technologies that are harder to use in

18
00:01:16,680 --> 00:01:19,360
Speaker 1: a way that is a net positive, you know, like

19
00:01:19,640 --> 00:01:24,000
Speaker 1: nuclear warheads for example. Not that people haven't tried, they have,

20
00:01:24,600 --> 00:01:28,360
Speaker 1: but you get my meaning. Well, one technology that has

21
00:01:28,440 --> 00:01:32,000
Speaker 1: far reaching effects and consequences that can end up being

22
00:01:32,040 --> 00:01:35,920
Speaker 1: positive or negative. Would be the algorithms that are underlying

23
00:01:36,000 --> 00:01:40,280
Speaker 1: stuff like search and social networks. Specifically today, I'm going

24
00:01:40,360 --> 00:01:44,760
Speaker 1: to be talking about Facebook in this episode, but algorithms

25
00:01:44,840 --> 00:01:49,000
Speaker 1: underlie the various computer systems and online systems that we

26
00:01:49,040 --> 00:01:51,840
Speaker 1: rely upon, and they have the potential to be harmful

27
00:01:51,960 --> 00:01:56,360
Speaker 1: depending on their implementation. But before I start talking about

28
00:01:56,400 --> 00:02:00,320
Speaker 1: the effects of algorithms, let's back up a little bit

29
00:02:00,440 --> 00:02:06,000
Speaker 1: and define that what exactly is an algorithm. Basically, an

30
00:02:06,040 --> 00:02:10,200
Speaker 1: algorithm is a set of directions or rules, and that's

31
00:02:10,240 --> 00:02:15,560
Speaker 1: really it. There's nothing special about this definition. These rules

32
00:02:15,639 --> 00:02:20,679
Speaker 1: guide how computers calculate operations and solve problems. It's a

33
00:02:20,800 --> 00:02:25,400
Speaker 1: finite sequence of instructions for computers to follow when they're

34
00:02:25,440 --> 00:02:30,520
Speaker 1: executing a specific task. These instructions are unambiguous, which means

35
00:02:30,840 --> 00:02:34,000
Speaker 1: the computer system doesn't have to make judgment calls or

36
00:02:34,000 --> 00:02:38,519
Speaker 1: anything like that while executing the instructions. There is a

37
00:02:38,520 --> 00:02:42,480
Speaker 1: clear path to follow based on whatever the situation is

38
00:02:42,560 --> 00:02:47,079
Speaker 1: at hand. A different situation might lead to a different outcome,

39
00:02:47,240 --> 00:02:50,560
Speaker 1: but the instruction path would still remain clear. If you

40
00:02:50,600 --> 00:02:53,320
Speaker 1: were to dig down deeply enough to see what was

41
00:02:53,400 --> 00:02:57,600
Speaker 1: going on behind the computer's actions. The number of potential

42
00:02:57,639 --> 00:03:01,400
Speaker 1: paths to various outcomes could be an warmous and so

43
00:03:01,480 --> 00:03:04,360
Speaker 1: to us it would look like a really complicated system

44
00:03:04,360 --> 00:03:08,320
Speaker 1: and very difficult to grock. But if you really got granular,

45
00:03:08,800 --> 00:03:11,639
Speaker 1: you would see the logic, as it were, that connects

46
00:03:11,680 --> 00:03:16,040
Speaker 1: the beginning state to the end state of any given operation.

47
00:03:16,800 --> 00:03:21,200
Speaker 1: At Google, search algorithms include elements that determine what position

48
00:03:21,240 --> 00:03:25,320
Speaker 1: of particular query result will take on a results page.

49
00:03:25,960 --> 00:03:29,480
Speaker 1: So when we search for something using an engine like Google,

50
00:03:29,960 --> 00:03:33,160
Speaker 1: there are several things going on in that split second

51
00:03:33,240 --> 00:03:36,760
Speaker 1: before our results pop up. For one thing, Google has

52
00:03:36,800 --> 00:03:40,080
Speaker 1: an index of all the web pages it has crawled,

53
00:03:40,480 --> 00:03:43,880
Speaker 1: and so Google will reference that index to find all

54
00:03:43,920 --> 00:03:47,160
Speaker 1: the entries and that index that relate to the terms

55
00:03:47,160 --> 00:03:49,800
Speaker 1: that you put in your search query. And then those

56
00:03:49,840 --> 00:03:53,560
Speaker 1: results are ranked by a few different criteria. So way

57
00:03:53,680 --> 00:03:58,480
Speaker 1: up there is relevancy, like how relevant is each search

58
00:03:58,560 --> 00:04:01,960
Speaker 1: result to the original search query. If I were to

59
00:04:02,000 --> 00:04:05,600
Speaker 1: search for Captain America, Google is going to put the

60
00:04:05,600 --> 00:04:08,800
Speaker 1: pages about the fictional superhero way up towards the top

61
00:04:09,360 --> 00:04:12,320
Speaker 1: search results about someone who happens to hold the rank

62
00:04:12,440 --> 00:04:16,120
Speaker 1: of Captain and there's somewhere a mention of the region

63
00:04:16,160 --> 00:04:20,160
Speaker 1: of America would likely fall much lower on the search

64
00:04:20,200 --> 00:04:24,520
Speaker 1: results list, because even though those pages will have relevant

65
00:04:24,640 --> 00:04:27,880
Speaker 1: terms in them, the chances are it's not what I'm

66
00:04:27,920 --> 00:04:31,919
Speaker 1: actually looking for. A context is important, but context is

67
00:04:32,000 --> 00:04:36,120
Speaker 1: just part of it. There are thousands of pages that

68
00:04:36,200 --> 00:04:40,719
Speaker 1: mentioned Captain America. Some are from really good resources, meaning

69
00:04:40,920 --> 00:04:44,560
Speaker 1: they are resources that are dependable, they're informative, they might

70
00:04:44,560 --> 00:04:48,720
Speaker 1: be entertaining, you know whatever. Some are from garbage sites

71
00:04:48,760 --> 00:04:52,800
Speaker 1: that don't provide any real valuable experience. So then you've

72
00:04:52,839 --> 00:04:56,039
Speaker 1: got the ranking system within Google that determines how the

73
00:04:56,120 --> 00:05:01,040
Speaker 1: most relevant search engine results will appear on the results age. Now,

74
00:05:01,040 --> 00:05:04,000
Speaker 1: if there are thousands or even millions of hits, how

75
00:05:04,000 --> 00:05:06,880
Speaker 1: does Google determine which ones show up on the first

76
00:05:06,880 --> 00:05:10,520
Speaker 1: page of results. Well, there's a complicated system that determines

77
00:05:10,680 --> 00:05:13,359
Speaker 1: that ranking, and it's not really in the realm of

78
00:05:13,400 --> 00:05:16,200
Speaker 1: this episode to go into it, but suffice it to

79
00:05:16,240 --> 00:05:19,120
Speaker 1: say that Google takes a lot of things into consideration

80
00:05:19,320 --> 00:05:22,920
Speaker 1: when ordering search results. Ostensibly, this is all in an

81
00:05:22,960 --> 00:05:25,840
Speaker 1: effort to give you the most relevant and helpful results

82
00:05:26,080 --> 00:05:29,200
Speaker 1: based on your search query. For the people companies and

83
00:05:29,320 --> 00:05:32,680
Speaker 1: organizations that are making pages. It's always a goal to

84
00:05:32,880 --> 00:05:37,000
Speaker 1: rank really high on that list, preferably on that first page,

85
00:05:37,000 --> 00:05:39,719
Speaker 1: and if you can, the top half of that first page,

86
00:05:39,760 --> 00:05:42,960
Speaker 1: because most people will only look at the first few

87
00:05:43,040 --> 00:05:46,320
Speaker 1: search results and then they'll ignore everything that comes afterward.

88
00:05:46,920 --> 00:05:49,839
Speaker 1: Over the years, Google has made it really tricky to

89
00:05:49,960 --> 00:05:53,480
Speaker 1: game this engine, which means it's hard to come up

90
00:05:53,520 --> 00:05:57,440
Speaker 1: with effective strategies to show up higher in search results.

91
00:05:57,920 --> 00:06:01,000
Speaker 1: Your best bet is really to create just really good

92
00:06:01,040 --> 00:06:05,240
Speaker 1: resources around the particular topic of interest in question. And also,

93
00:06:05,400 --> 00:06:08,440
Speaker 1: you know, happened to be a really respected resource, which

94
00:06:08,520 --> 00:06:12,560
Speaker 1: is not easy to do, especially if you're just starting out. Obviously, Also,

95
00:06:12,600 --> 00:06:17,120
Speaker 1: whenever Google makes changes to this algorithm, it can drastically

96
00:06:17,120 --> 00:06:20,719
Speaker 1: affect how your page appears in search results, even with

97
00:06:20,760 --> 00:06:24,240
Speaker 1: a perfect query. And when tech stuff was still part

98
00:06:24,240 --> 00:06:27,200
Speaker 1: of how stuff works, that was something we were always

99
00:06:27,240 --> 00:06:31,040
Speaker 1: concerned about. A change in the Google algorithm would mean

100
00:06:31,080 --> 00:06:35,000
Speaker 1: that our web traffic could either skyrocket or it could plummet.

101
00:06:35,560 --> 00:06:37,760
Speaker 1: There was no change in the quality of our work.

102
00:06:38,000 --> 00:06:40,320
Speaker 1: Our work was just as good as it had been before.

103
00:06:41,000 --> 00:06:43,480
Speaker 1: It would just be a change in how our work

104
00:06:43,480 --> 00:06:47,760
Speaker 1: would show up in search results. However, that's the Google algorithm.

105
00:06:47,800 --> 00:06:51,600
Speaker 1: With other algorithms like the ones on Facebook, those can

106
00:06:51,640 --> 00:06:54,480
Speaker 1: be easier to game if you know what you're doing.

107
00:06:54,839 --> 00:06:58,000
Speaker 1: And what's more, there are algorithms on Facebook that allow

108
00:06:58,080 --> 00:07:03,680
Speaker 1: for advertisers to engage in some discriminatory and potentially criminal behavior.

109
00:07:03,800 --> 00:07:07,280
Speaker 1: So we'll look at how the various algorithms on Facebook

110
00:07:07,560 --> 00:07:10,560
Speaker 1: can have a negative impact because of the way they

111
00:07:10,560 --> 00:07:13,800
Speaker 1: are implemented and because of the way that people are

112
00:07:13,840 --> 00:07:17,520
Speaker 1: taking advantage of them. Now, in the beginning, Facebook didn't

113
00:07:17,560 --> 00:07:21,040
Speaker 1: have an algorithm at all. Heck, it didn't even have

114
00:07:21,080 --> 00:07:24,320
Speaker 1: a news feed when it first started. Instead, you could

115
00:07:24,360 --> 00:07:27,360
Speaker 1: post stuff to your profile, but then you could visit

116
00:07:27,400 --> 00:07:30,200
Speaker 1: the profiles of friends to see what they had posted.

117
00:07:30,680 --> 00:07:34,320
Speaker 1: You could leave comments on stuff, but it still required

118
00:07:34,360 --> 00:07:38,560
Speaker 1: you to actually go over to your friends profile. There

119
00:07:38,640 --> 00:07:43,520
Speaker 1: was no aggregator for content just yet that changed in

120
00:07:43,600 --> 00:07:48,320
Speaker 1: two thousand seven when Facebook introduced the news feed. Now,

121
00:07:48,480 --> 00:07:52,320
Speaker 1: your homepage on Facebook had a record of your friends activities,

122
00:07:52,360 --> 00:07:55,720
Speaker 1: just listed in reverse chronological order. You know, the most

123
00:07:55,800 --> 00:07:58,920
Speaker 1: recent things appearing at the top of the list. You

124
00:07:58,920 --> 00:08:01,280
Speaker 1: could log in and see what your friends had been

125
00:08:01,360 --> 00:08:03,440
Speaker 1: up to since the last time you were on and

126
00:08:03,480 --> 00:08:06,800
Speaker 1: you could catch up. There's still no algorithm to speak

127
00:08:06,840 --> 00:08:10,080
Speaker 1: of yet, and as would become a hallmark of any

128
00:08:10,280 --> 00:08:14,560
Speaker 1: change made on Facebook. Ever, a lot of users objected

129
00:08:14,760 --> 00:08:17,040
Speaker 1: to the news feed, and the main objection was that

130
00:08:17,440 --> 00:08:20,640
Speaker 1: people felt it was a violation of privacy. They didn't

131
00:08:20,680 --> 00:08:23,800
Speaker 1: like the idea that the activities they were doing on

132
00:08:23,800 --> 00:08:28,400
Speaker 1: Facebook were being broadcast to all of their friends, which okay,

133
00:08:28,400 --> 00:08:30,880
Speaker 1: now this one stumps me a little bit, because the

134
00:08:30,920 --> 00:08:35,840
Speaker 1: profiles on Facebook were public unless you set privacy features

135
00:08:35,880 --> 00:08:39,520
Speaker 1: to make it otherwise. So if you change your relationship

136
00:08:39,600 --> 00:08:43,360
Speaker 1: status on your profile, for example, why would you object

137
00:08:43,360 --> 00:08:47,640
Speaker 1: to it being broadcast to your friends. If anyone visited

138
00:08:47,640 --> 00:08:50,440
Speaker 1: your profile, they would see the difference. So either you

139
00:08:50,600 --> 00:08:53,360
Speaker 1: don't change it, or you change the privacy settings on

140
00:08:53,400 --> 00:08:56,280
Speaker 1: your profile, But otherwise why would you be upset? I

141
00:08:56,320 --> 00:08:59,800
Speaker 1: think the main takeaway is just that people hate change,

142
00:09:00,160 --> 00:09:02,920
Speaker 1: and if you don't want something broadcast to everybody, maybe

143
00:09:02,960 --> 00:09:07,120
Speaker 1: you shouldn't post it. A couple of years later, Facebook

144
00:09:07,160 --> 00:09:11,240
Speaker 1: introduced the first iteration of their algorithm, which would determine

145
00:09:11,360 --> 00:09:14,320
Speaker 1: what users would see in their news feed and in

146
00:09:14,440 --> 00:09:17,560
Speaker 1: what order they would appear, so rather than getting a

147
00:09:17,640 --> 00:09:21,200
Speaker 1: full account, of all of your friends activities in reverse

148
00:09:21,280 --> 00:09:25,000
Speaker 1: chronological order. Users would see a selection of stuff that

149
00:09:25,040 --> 00:09:29,120
Speaker 1: the algorithm had determined was the best for that user.

150
00:09:29,640 --> 00:09:35,360
Speaker 1: So how did that work? Well, generally speaking, the algorithm

151
00:09:35,360 --> 00:09:39,560
Speaker 1: took three main criteria into account to select the stuff

152
00:09:39,760 --> 00:09:43,760
Speaker 1: that would show up on news feeds. One was user affinity,

153
00:09:44,080 --> 00:09:46,960
Speaker 1: which was an indicator of how close the relationship is

154
00:09:47,520 --> 00:09:51,920
Speaker 1: between one Facebook user and another, or a Facebook user

155
00:09:52,040 --> 00:09:55,960
Speaker 1: and a brand. And this was measured in one way perspectives,

156
00:09:56,559 --> 00:09:59,400
Speaker 1: and that can be a little confusing. So let's make

157
00:09:59,520 --> 00:10:03,360
Speaker 1: a an example. We'll take three fictional people to explain

158
00:10:03,480 --> 00:10:08,240
Speaker 1: how this affinity worked. So in this example, we've got Sam, Dean,

159
00:10:08,760 --> 00:10:14,080
Speaker 1: and Castiel. Okay, So Sam and Dean are super close, right,

160
00:10:14,640 --> 00:10:18,160
Speaker 1: and they often comment on each other's posts, and they'll

161
00:10:18,240 --> 00:10:21,000
Speaker 1: hit like on each other's statuses and all that kind

162
00:10:21,040 --> 00:10:24,440
Speaker 1: of stuff. So they tend to see one another's posts

163
00:10:24,520 --> 00:10:26,960
Speaker 1: a lot in their own news feeds when they log

164
00:10:27,000 --> 00:10:31,400
Speaker 1: onto Facebook. They each have a high affinity for each other.

165
00:10:32,040 --> 00:10:35,400
Speaker 1: Then we've got Castil. So Castiel is friends with both

166
00:10:35,559 --> 00:10:39,720
Speaker 1: Sam and Dean, but really he's much closer to Dean

167
00:10:39,880 --> 00:10:43,320
Speaker 1: than he is to Sam. And Castiole responds to a

168
00:10:43,320 --> 00:10:46,720
Speaker 1: lot of Dean's posts, and as a result, Castiel C's

169
00:10:46,840 --> 00:10:51,400
Speaker 1: Dean's stuff pop up in his that is, Castil's news

170
00:10:51,400 --> 00:10:55,160
Speaker 1: feed pretty regularly. Castil is not as close with Sam

171
00:10:55,200 --> 00:10:58,360
Speaker 1: and he doesn't interact with Sam as much. So sam

172
00:10:58,440 --> 00:11:01,760
Speaker 1: stuff can pop up in Castile's news feed, but it's

173
00:11:01,800 --> 00:11:04,280
Speaker 1: not as frequent or as sure a thing as when

174
00:11:04,360 --> 00:11:08,800
Speaker 1: Dean posts something. Now, on top of that, Dean likes Castiel,

175
00:11:09,080 --> 00:11:12,000
Speaker 1: but you get the feeling that this relationship is not

176
00:11:12,080 --> 00:11:16,840
Speaker 1: a perfect two way street. So Dean only interacts with

177
00:11:17,120 --> 00:11:21,240
Speaker 1: some of Castile's stuff that he posts online. So Dean

178
00:11:21,559 --> 00:11:26,040
Speaker 1: sees Castiel's posts on occasion in his news feed, but

179
00:11:26,360 --> 00:11:29,680
Speaker 1: not nearly at the same frequency at which he sees

180
00:11:30,160 --> 00:11:34,160
Speaker 1: Sam's post in his that is Dean's news feed. And yes,

181
00:11:34,920 --> 00:11:38,199
Speaker 1: I just summarized the show Supernatural in an example about

182
00:11:38,320 --> 00:11:42,480
Speaker 1: Facebook algorithms and affinity. Now, as we see a high

183
00:11:42,520 --> 00:11:46,679
Speaker 1: affinity one person has for another helps determine how much

184
00:11:47,040 --> 00:11:50,040
Speaker 1: of that other person's stuff the first person is going

185
00:11:50,120 --> 00:11:52,160
Speaker 1: to see. But it only works in that one way.

186
00:11:52,480 --> 00:11:56,200
Speaker 1: If I like your posts a lot, then I'm going

187
00:11:56,280 --> 00:11:58,600
Speaker 1: to see more of your posts in my feed. But

188
00:11:58,640 --> 00:12:01,360
Speaker 1: if you never interact with me, You're not going to

189
00:12:01,440 --> 00:12:03,800
Speaker 1: see a lot of my stuff in your feed. That's

190
00:12:03,800 --> 00:12:06,280
Speaker 1: what I mean by that one way perspective. Of course,

191
00:12:06,679 --> 00:12:10,240
Speaker 1: there are other ways to wait this, more like you

192
00:12:10,280 --> 00:12:13,439
Speaker 1: can tag someone in a post, which will definitely give

193
00:12:13,480 --> 00:12:16,839
Speaker 1: their attention much more effectively. But let's move on. So

194
00:12:16,880 --> 00:12:20,360
Speaker 1: the second component to the edge rank factor we had

195
00:12:20,600 --> 00:12:24,280
Speaker 1: user affinity as a component one. Component two is the

196
00:12:24,320 --> 00:12:28,319
Speaker 1: weight of the post W E, I, G, H T.

197
00:12:29,200 --> 00:12:33,040
Speaker 1: So different kinds of posts have different weights, and the

198
00:12:33,080 --> 00:12:36,240
Speaker 1: greater the weight, the more likely it will pop up

199
00:12:36,600 --> 00:12:41,360
Speaker 1: on a given friends feed. So, generally speaking, with edge rank,

200
00:12:41,559 --> 00:12:45,520
Speaker 1: stuff that had photos or videos had the most weight,

201
00:12:46,040 --> 00:12:52,000
Speaker 1: followed by posts that contained links, followed by plain text posts. Now,

202
00:12:52,040 --> 00:12:55,640
Speaker 1: the reason for this is that Facebook analysts had figured

203
00:12:55,679 --> 00:12:59,240
Speaker 1: out that posts that had stuff like photos and videos

204
00:12:59,640 --> 00:13:03,120
Speaker 1: with tend to get more engagement. I'll get back to

205
00:13:03,240 --> 00:13:06,520
Speaker 1: engagement a bit later in this episode. Just know that

206
00:13:06,600 --> 00:13:11,440
Speaker 1: it is a critical concept for Facebook. The third component

207
00:13:11,760 --> 00:13:16,200
Speaker 1: factored into edge rink is called time decay, which essentially

208
00:13:16,200 --> 00:13:20,720
Speaker 1: translates to how old is this post? And the idea

209
00:13:20,800 --> 00:13:24,400
Speaker 1: here is that the more recent posts are likely more

210
00:13:24,480 --> 00:13:28,720
Speaker 1: relevant than older posts are it doesn't do you any

211
00:13:28,760 --> 00:13:31,960
Speaker 1: good to see the announcement of an upcoming get together

212
00:13:32,600 --> 00:13:36,640
Speaker 1: if in fact that get together actually happened last week,

213
00:13:36,800 --> 00:13:41,000
Speaker 1: so you're not likely to see the initial post pop

214
00:13:41,080 --> 00:13:42,679
Speaker 1: up at the top of your news feed if you've

215
00:13:42,720 --> 00:13:47,200
Speaker 1: already missed everything. As the post ages, the value of

216
00:13:47,240 --> 00:13:51,560
Speaker 1: the post decreases. So when you log into Facebook or

217
00:13:51,760 --> 00:13:55,520
Speaker 1: open the Facebook app or whatever, this algorithm would run

218
00:13:55,679 --> 00:13:58,680
Speaker 1: with your particular situation in mind. It would scan through

219
00:13:58,760 --> 00:14:02,160
Speaker 1: all the stuff that had been posted by the friends

220
00:14:02,200 --> 00:14:05,960
Speaker 1: you have and the brands you follow since your last session,

221
00:14:06,280 --> 00:14:10,640
Speaker 1: and then it would select and rank or order those

222
00:14:10,679 --> 00:14:13,880
Speaker 1: status updates that you would see in your news feed

223
00:14:14,120 --> 00:14:18,440
Speaker 1: based off of those criteria. But this would also mean

224
00:14:18,880 --> 00:14:22,360
Speaker 1: you would miss out on stuff. No longer would you

225
00:14:22,360 --> 00:14:25,680
Speaker 1: be able to see everything that all of your friends

226
00:14:25,720 --> 00:14:30,280
Speaker 1: were doing, listed in reverse chronological order. In fact, at times,

227
00:14:30,360 --> 00:14:33,520
Speaker 1: chronology didn't have anything to do with the order in

228
00:14:33,520 --> 00:14:36,960
Speaker 1: which status updates would appear on your news feed. A

229
00:14:37,040 --> 00:14:41,160
Speaker 1: post several hours old might appear above one that was

230
00:14:41,200 --> 00:14:44,600
Speaker 1: posted just five minutes before you logged in, all because

231
00:14:44,720 --> 00:14:49,080
Speaker 1: of those other factors. For some Facebook users, including me

232
00:14:49,360 --> 00:14:53,040
Speaker 1: at the time, this was really frustrating. It made us

233
00:14:53,040 --> 00:14:55,880
Speaker 1: feel as though we were missing out on stuff our

234
00:14:55,920 --> 00:14:59,680
Speaker 1: friends were saying because we were. I found it particularly

235
00:14:59,720 --> 00:15:04,240
Speaker 1: irre tating because I wasn't one to frequently engage with updates.

236
00:15:04,240 --> 00:15:07,440
Speaker 1: I didn't like a lot of them, and I wouldn't comment.

237
00:15:07,480 --> 00:15:10,280
Speaker 1: A ton of times. I would read everything, but I

238
00:15:10,320 --> 00:15:13,240
Speaker 1: wasn't interacting with it, So the algorithm wasn't getting a

239
00:15:13,280 --> 00:15:17,880
Speaker 1: really good idea of who my good friends were versus

240
00:15:17,960 --> 00:15:21,880
Speaker 1: like people who are my acquaintances. So it meant that

241
00:15:21,960 --> 00:15:26,520
Speaker 1: some people who were my closest friends in reality were

242
00:15:26,560 --> 00:15:29,640
Speaker 1: not the ones that I was seeing, you know, being

243
00:15:29,680 --> 00:15:32,240
Speaker 1: active on Facebook. Even if they were, I just wasn't

244
00:15:32,280 --> 00:15:35,840
Speaker 1: seeing those posts. And my case was not unusual. A

245
00:15:35,840 --> 00:15:38,040
Speaker 1: lot of people went through this and and then I

246
00:15:38,120 --> 00:15:42,520
Speaker 1: missed stuff, sometimes big stuff like an engagement announcement or something.

247
00:15:43,520 --> 00:15:47,640
Speaker 1: Edge rank was just the beginning, and actually that particular

248
00:15:47,680 --> 00:15:51,240
Speaker 1: algorithm didn't stick around that long in the grand scheme

249
00:15:51,240 --> 00:15:55,760
Speaker 1: of things. By late within Facebook, no one was really

250
00:15:55,840 --> 00:15:59,160
Speaker 1: using the term edge rank anymore because they had moved

251
00:15:59,160 --> 00:16:04,080
Speaker 1: towards a more sophisticated machine learning based algorithm. This one

252
00:16:04,160 --> 00:16:06,600
Speaker 1: was far more sophisticated than edge rank. It was not

253
00:16:06,720 --> 00:16:09,760
Speaker 1: limited to just those three main factors, although they were

254
00:16:09,760 --> 00:16:14,720
Speaker 1: still components within the algorithm. Instead, according to Lar's Backstrom,

255
00:16:14,720 --> 00:16:17,920
Speaker 1: who was the engineering manager for the news feed ranking

256
00:16:17,960 --> 00:16:20,960
Speaker 1: at the time, this new one used somewhere in the

257
00:16:20,960 --> 00:16:25,800
Speaker 1: neighborhood of one hundred thousand weights to determine what users

258
00:16:25,800 --> 00:16:28,600
Speaker 1: would see and in which order. Now you would think

259
00:16:28,760 --> 00:16:30,680
Speaker 1: that with that many weights it would be hard to

260
00:16:30,720 --> 00:16:34,160
Speaker 1: gain the system, but you'd be wrong. Now, a couple

261
00:16:34,240 --> 00:16:36,320
Speaker 1: other things I want to mention before we go to break.

262
00:16:36,920 --> 00:16:39,600
Speaker 1: One of those is that, yes, there was an option

263
00:16:39,720 --> 00:16:44,080
Speaker 1: still is to try and list things in most recent

264
00:16:44,240 --> 00:16:48,040
Speaker 1: as opposed to what is most relevant or whatever Facebook

265
00:16:48,040 --> 00:16:51,480
Speaker 1: thinks is most relevant, except that you're not really seeing

266
00:16:51,600 --> 00:16:54,880
Speaker 1: everything that way. Either. You're not getting the most recent

267
00:16:55,440 --> 00:16:59,520
Speaker 1: reverse chronological list of all of your friends activities. It's

268
00:16:59,560 --> 00:17:03,120
Speaker 1: still a selection of what your friends are posting, and

269
00:17:03,120 --> 00:17:05,720
Speaker 1: it's a selection that the algorithm has determined is the

270
00:17:05,760 --> 00:17:08,959
Speaker 1: most relevant for you. It's just that they'll be listed

271
00:17:09,040 --> 00:17:12,440
Speaker 1: in reverse order instead of kind of that haphazard all

272
00:17:12,440 --> 00:17:15,200
Speaker 1: over the place method. But the fact is you're still

273
00:17:15,240 --> 00:17:18,160
Speaker 1: not seeing everything, which still frustrates a lot of people.

274
00:17:18,840 --> 00:17:22,720
Speaker 1: And when we come back we'll talk about the algorithm,

275
00:17:23,080 --> 00:17:27,840
Speaker 1: how people have used the creative approach to game the system,

276
00:17:27,880 --> 00:17:32,439
Speaker 1: and some unintended consequences that algorithm has led to. But

277
00:17:32,560 --> 00:17:42,600
Speaker 1: first let's take a quick break. So we're at the

278
00:17:42,680 --> 00:17:47,000
Speaker 1: point where an algorithm starts to serve up content catered

279
00:17:47,119 --> 00:17:51,920
Speaker 1: to each individual Facebook user, and the goal was always

280
00:17:51,920 --> 00:17:56,280
Speaker 1: the same from the very beginning keep people engaged, which

281
00:17:56,320 --> 00:18:00,800
Speaker 1: really just means keep people on Facebook for longer. And

282
00:18:00,840 --> 00:18:05,719
Speaker 1: this gets to Facebook's method of generating revenue. Facebook's primary

283
00:18:05,760 --> 00:18:10,120
Speaker 1: source of revenue is through advertising. And when I say primary,

284
00:18:10,160 --> 00:18:16,960
Speaker 1: according to Investopedia, advertising accounts for ent of Facebook's revenue,

285
00:18:17,359 --> 00:18:25,399
Speaker 1: which for Q three was twenty one point five billion dollars.

286
00:18:25,840 --> 00:18:31,280
Speaker 1: Wholly cal princely some doesn't even begin to cover it.

287
00:18:31,920 --> 00:18:35,320
Speaker 1: Facebook sales ad space to various companies to serve up

288
00:18:35,320 --> 00:18:39,080
Speaker 1: advertising to Facebook's users, so they're essentially think of it

289
00:18:39,200 --> 00:18:42,159
Speaker 1: like a billboard. Facebook is like a giant billboard, but

290
00:18:42,160 --> 00:18:46,080
Speaker 1: it's a very special billboard. It's incredibly valuable. There's another

291
00:18:46,119 --> 00:18:48,760
Speaker 1: component to this I'll touch on later, which revolves around

292
00:18:48,760 --> 00:18:53,399
Speaker 1: targeted advertising, but let's just stick with the basic value

293
00:18:53,400 --> 00:18:57,959
Speaker 1: proposition here. So Facebook has space on its apps and

294
00:18:58,119 --> 00:19:00,879
Speaker 1: its pages, and in that space ace it has the

295
00:19:01,040 --> 00:19:04,639
Speaker 1: ability to show ads that they could be on the side,

296
00:19:04,680 --> 00:19:06,880
Speaker 1: they can be fed into the news feed as well,

297
00:19:06,920 --> 00:19:10,120
Speaker 1: so they're kind of interspersed along with the status updates

298
00:19:10,119 --> 00:19:14,280
Speaker 1: from friends and stuff. Uh. The longer that people use

299
00:19:14,359 --> 00:19:18,480
Speaker 1: Facebook per any given session, the more ads they will

300
00:19:18,600 --> 00:19:21,879
Speaker 1: encounter as they interact with Facebook, which also means that

301
00:19:21,920 --> 00:19:25,520
Speaker 1: they could potentially interact with the ads. So part of

302
00:19:25,520 --> 00:19:28,840
Speaker 1: what Facebook really needs to do is to find ways

303
00:19:28,880 --> 00:19:32,800
Speaker 1: to convince people to spend as much time as possible

304
00:19:33,359 --> 00:19:39,080
Speaker 1: on Facebook. In this use case, time really is money.

305
00:19:39,320 --> 00:19:42,639
Speaker 1: Greater engagement doesn't just mean more opportunities to see ads,

306
00:19:42,920 --> 00:19:45,639
Speaker 1: it also means that Facebook can demonstrate its own value

307
00:19:45,680 --> 00:19:50,159
Speaker 1: to advertisers and thus command higher prices. To put it

308
00:19:50,240 --> 00:19:54,320
Speaker 1: another way, let's say you own a few billboards around town.

309
00:19:54,359 --> 00:19:57,080
Speaker 1: Since I've already mentioned them, so let's say you have

310
00:19:57,119 --> 00:19:59,879
Speaker 1: a billboard that is on a back road and that

311
00:20:00,359 --> 00:20:03,240
Speaker 1: gets just a little bit of traffic every day, and

312
00:20:03,320 --> 00:20:06,520
Speaker 1: you happen to own another billboard space which is on

313
00:20:06,560 --> 00:20:09,120
Speaker 1: a much more heavily used road that goes right into

314
00:20:09,160 --> 00:20:11,600
Speaker 1: the heart of town. The rates that you would be

315
00:20:11,640 --> 00:20:15,560
Speaker 1: able to charge for the back road billboard are probably

316
00:20:15,560 --> 00:20:18,200
Speaker 1: gonna be a lot lower than what you can command

317
00:20:18,240 --> 00:20:21,840
Speaker 1: with the heavily traveled thoroughfare. More folks are going to

318
00:20:21,920 --> 00:20:25,919
Speaker 1: see that second billboard, which makes it more valuable. On

319
00:20:25,960 --> 00:20:29,600
Speaker 1: a similar front, if Facebook can show numbers that indicate

320
00:20:30,040 --> 00:20:33,399
Speaker 1: that more than two billion people are spending more and

321
00:20:33,440 --> 00:20:38,000
Speaker 1: more time on Facebook, it can command higher prices for advertising.

322
00:20:38,320 --> 00:20:41,199
Speaker 1: Add companies and their clients want to be where the

323
00:20:41,240 --> 00:20:45,040
Speaker 1: eyeballs are, and Facebook fits the bill nicely. So with

324
00:20:45,200 --> 00:20:48,440
Speaker 1: that in mind, the real purpose of the algorithm ultimately

325
00:20:48,680 --> 00:20:51,240
Speaker 1: is to serve up stuff to users that will keep

326
00:20:51,320 --> 00:20:54,600
Speaker 1: them on Facebook longer and convince them to come back

327
00:20:54,640 --> 00:20:58,640
Speaker 1: to Facebook frequently. Now, some of the things that will

328
00:20:58,680 --> 00:21:01,040
Speaker 1: bring people back fall online with a few of the

329
00:21:01,119 --> 00:21:05,520
Speaker 1: ideas that Drank demonstrated. If you logged onto Facebook but

330
00:21:05,680 --> 00:21:08,240
Speaker 1: you only ever saw posts from people you only kind

331
00:21:08,280 --> 00:21:11,000
Speaker 1: of sort of know, and you always missed out on

332
00:21:11,000 --> 00:21:14,000
Speaker 1: the posts from your close friends, you probably not find

333
00:21:14,080 --> 00:21:17,280
Speaker 1: it a very satisfying experience. So at least some of

334
00:21:17,320 --> 00:21:19,479
Speaker 1: the stuff you'd be seeing would fall in line with

335
00:21:19,600 --> 00:21:22,200
Speaker 1: the stuff you'd really want to see in the first place,

336
00:21:22,720 --> 00:21:25,359
Speaker 1: But maybe there would be some posts you would really

337
00:21:25,400 --> 00:21:28,800
Speaker 1: like to see, but the algorithm, for whatever reason, doesn't

338
00:21:28,880 --> 00:21:32,760
Speaker 1: judge those as being particularly engaging. Well, the chances are

339
00:21:32,920 --> 00:21:36,240
Speaker 1: more slim that you would see those posts. If you've

340
00:21:36,280 --> 00:21:39,640
Speaker 1: ever posted something to Facebook that you thought was particularly

341
00:21:39,680 --> 00:21:44,399
Speaker 1: important or impactful, but you didn't see very much reaction

342
00:21:44,520 --> 00:21:47,360
Speaker 1: to it, there's a chance that not many people actually

343
00:21:47,440 --> 00:21:51,840
Speaker 1: saw what you posted, So that's a real possibility. Of course,

344
00:21:52,040 --> 00:21:54,240
Speaker 1: there's also a chance that they did see it, they

345
00:21:54,280 --> 00:21:56,840
Speaker 1: just didn't care. It's kind of hard to say for

346
00:21:56,880 --> 00:22:00,639
Speaker 1: any given post, but it's entirely possible for you to

347
00:22:00,680 --> 00:22:04,280
Speaker 1: post something that people would have reacted to. They just

348
00:22:04,320 --> 00:22:06,240
Speaker 1: never saw it, so they never got the chance to.

349
00:22:06,800 --> 00:22:10,120
Speaker 1: But let's say someone in your circle posts something and

350
00:22:10,240 --> 00:22:13,359
Speaker 1: it starts to catch on. Maybe it's someone in your

351
00:22:13,400 --> 00:22:16,159
Speaker 1: circle that you don't interact with much, so usually you

352
00:22:16,200 --> 00:22:19,879
Speaker 1: don't see very much from them on Facebook. But for

353
00:22:19,920 --> 00:22:23,560
Speaker 1: whatever reason, this particular post that they made is getting

354
00:22:23,560 --> 00:22:26,520
Speaker 1: a lot of interaction from other friends, which means the

355
00:22:26,520 --> 00:22:29,920
Speaker 1: algorithm might say, hey, this post is really taking off

356
00:22:30,359 --> 00:22:32,960
Speaker 1: a lot of people are interacting with it, so I'm

357
00:22:33,000 --> 00:22:36,440
Speaker 1: going to serve it up to even more people, including you,

358
00:22:36,920 --> 00:22:41,600
Speaker 1: who usually wouldn't see such things. From this particular distant friend. So,

359
00:22:41,600 --> 00:22:45,040
Speaker 1: in other words, posts that get a lot of engagement,

360
00:22:45,240 --> 00:22:49,080
Speaker 1: that being reactions and comments and shares and that kind

361
00:22:49,119 --> 00:22:53,919
Speaker 1: of stuff, they tend to rise above other posts. Engagement

362
00:22:54,040 --> 00:22:57,879
Speaker 1: doesn't have to be positive either. It's not really important

363
00:22:57,920 --> 00:23:01,560
Speaker 1: to Facebook if the stuff you see makes you happy

364
00:23:01,680 --> 00:23:04,679
Speaker 1: or unhappy, or if the stuff you see is true

365
00:23:04,880 --> 00:23:08,600
Speaker 1: or if it's false. What matters to Facebook is that

366
00:23:08,680 --> 00:23:12,199
Speaker 1: you spend as much time on the platform as they

367
00:23:12,200 --> 00:23:14,920
Speaker 1: can coax from you, and that you feel the need

368
00:23:14,960 --> 00:23:19,520
Speaker 1: to return to Facebook frequently. Maybe it turns out you

369
00:23:19,640 --> 00:23:22,720
Speaker 1: rage posts on Facebook a lot. Maybe you see stuff

370
00:23:22,760 --> 00:23:27,919
Speaker 1: that you vehemently disagree with, so you post lengthy comments,

371
00:23:28,560 --> 00:23:30,840
Speaker 1: or you share a post you don't agree with, but

372
00:23:30,960 --> 00:23:32,840
Speaker 1: you do so in an effort to put your own

373
00:23:32,880 --> 00:23:36,080
Speaker 1: spin on the thesis of the original post, or to

374
00:23:36,160 --> 00:23:40,680
Speaker 1: criticize it in some way. Your motivations or goals are

375
00:23:40,800 --> 00:23:44,840
Speaker 1: not important to Facebook. The company could not care less.

376
00:23:45,280 --> 00:23:49,320
Speaker 1: The platform just once your time. Well, that's being a

377
00:23:49,320 --> 00:23:53,320
Speaker 1: bit too reductive. Facebook also wants your posts to inspire

378
00:23:53,400 --> 00:23:57,119
Speaker 1: other people to engage using their own comments, likes, and shares,

379
00:23:57,440 --> 00:24:00,080
Speaker 1: so that they get their time too. That's really it.

380
00:24:00,359 --> 00:24:04,320
Speaker 1: They want everyone's time, not just yours. I'm sure most

381
00:24:04,359 --> 00:24:07,360
Speaker 1: of you already see where this becomes an issue. As

382
00:24:07,359 --> 00:24:10,439
Speaker 1: we've heard for a few years now, Facebook has become

383
00:24:10,480 --> 00:24:15,879
Speaker 1: a hotbed for misinformation and disinformation, something that Facebook founder

384
00:24:15,920 --> 00:24:20,600
Speaker 1: Mark Zuckerberg has tried to avoid addressing for years. There's

385
00:24:20,600 --> 00:24:25,080
Speaker 1: a saying that's frequently misattributed to Mark Twain that says

386
00:24:25,440 --> 00:24:28,679
Speaker 1: a lie can travel halfway around the world while the

387
00:24:28,680 --> 00:24:31,800
Speaker 1: truth is still putting on its shoes. I think it's

388
00:24:31,880 --> 00:24:34,560
Speaker 1: kind of fitting that Twain is often cited as the

389
00:24:34,560 --> 00:24:37,040
Speaker 1: originator of the saying, but the truth of the matter

390
00:24:37,119 --> 00:24:39,480
Speaker 1: is that it predates him. It might have been the

391
00:24:39,480 --> 00:24:42,960
Speaker 1: product of Jonathan Swift, but we don't really know. But

392
00:24:43,359 --> 00:24:47,520
Speaker 1: the saying is apt. Misinformation is not saddled with the

393
00:24:47,640 --> 00:24:53,480
Speaker 1: heavy burden of being, you know. True misinformation can tap

394
00:24:53,520 --> 00:24:59,560
Speaker 1: into biases and preconceptions. It can reinforce the beliefs we hold,

395
00:25:00,160 --> 00:25:03,199
Speaker 1: whether they're positive or negative, even if those beliefs have

396
00:25:03,440 --> 00:25:08,080
Speaker 1: no real basis in reality. And if we feel that

397
00:25:08,119 --> 00:25:12,720
Speaker 1: our beliefs are validated, we're more likely to engage and

398
00:25:12,840 --> 00:25:16,119
Speaker 1: share those posts and to feel reinforced in it to

399
00:25:16,160 --> 00:25:19,800
Speaker 1: do more, to continue sharing that kind of stuff. Now.

400
00:25:19,840 --> 00:25:24,719
Speaker 1: I've seen examples of this across different ideologies. The narrative

401
00:25:24,800 --> 00:25:28,000
Speaker 1: we frequently see here in the United States is how

402
00:25:28,200 --> 00:25:32,720
Speaker 1: right wing misinformation will spread across Facebook, and that that

403
00:25:32,880 --> 00:25:35,800
Speaker 1: certainly does happen, but I've also seen it happen for

404
00:25:35,840 --> 00:25:40,240
Speaker 1: people who have a left leaning ideology. For example, if

405
00:25:40,240 --> 00:25:44,280
Speaker 1: a post indicates that someone with a right wing philosophy

406
00:25:44,359 --> 00:25:48,600
Speaker 1: has done or said something terrible, that can spread rapidly

407
00:25:48,720 --> 00:25:52,000
Speaker 1: across Facebook. But often you can do a little fact

408
00:25:52,119 --> 00:25:56,439
Speaker 1: checking and find out that there's no actual evidence that

409
00:25:56,520 --> 00:25:59,520
Speaker 1: such a thing was ever said or done by that person.

410
00:25:59,640 --> 00:26:02,560
Speaker 1: I've seen this happen multiple times, So what I'm getting

411
00:26:02,600 --> 00:26:06,760
Speaker 1: at is that no one is immune. That being said,

412
00:26:07,119 --> 00:26:11,600
Speaker 1: certain groups, particularly those catering to more right wing audiences,

413
00:26:12,200 --> 00:26:15,879
Speaker 1: have really pounced on the opportunity more than others have.

414
00:26:16,359 --> 00:26:19,560
Speaker 1: And again I'm not saying left leaning organizations and people

415
00:26:19,720 --> 00:26:23,520
Speaker 1: haven't done the same thing. They just haven't been as

416
00:26:23,640 --> 00:26:27,479
Speaker 1: fast to jump on it, and it's less common they

417
00:26:27,480 --> 00:26:30,640
Speaker 1: don't do it as much. Right now, I think it's

418
00:26:30,680 --> 00:26:34,720
Speaker 1: pretty fair to say that no ethical organization or person

419
00:26:35,200 --> 00:26:40,520
Speaker 1: would want to perpetuate misinformation or create disinformation. This is

420
00:26:40,560 --> 00:26:43,879
Speaker 1: a form of manipulation to get people to support something

421
00:26:44,240 --> 00:26:47,560
Speaker 1: without having to go to the trouble of actually really

422
00:26:48,000 --> 00:26:51,080
Speaker 1: giving evidence that that point of view is the best one.

423
00:26:51,560 --> 00:26:55,200
Speaker 1: If you're using lies to get people to support you, that's,

424
00:26:55,640 --> 00:27:00,359
Speaker 1: you know, not great. But on Facebook that sort of

425
00:27:00,400 --> 00:27:03,879
Speaker 1: thing can really take off, and as a post gets

426
00:27:03,960 --> 00:27:07,240
Speaker 1: traction with people sharing it or commenting on it, or

427
00:27:07,280 --> 00:27:12,200
Speaker 1: reacting to it, whether positively or negatively, Facebook will start

428
00:27:12,400 --> 00:27:16,959
Speaker 1: serving that content to progressively larger audiences. And you can

429
00:27:17,000 --> 00:27:19,840
Speaker 1: think of this as like ripples in a pond. That

430
00:27:19,960 --> 00:27:24,240
Speaker 1: initial circle of a ripple that's really small, right, But

431
00:27:24,320 --> 00:27:27,480
Speaker 1: let's say that that circle represents a certain group of

432
00:27:27,560 --> 00:27:30,760
Speaker 1: users and they react very strongly to that content. Well,

433
00:27:30,800 --> 00:27:34,000
Speaker 1: Facebook will then expand that to the next circle outward

434
00:27:34,480 --> 00:27:36,800
Speaker 1: to affect a larger group of users, and if the

435
00:27:36,840 --> 00:27:40,040
Speaker 1: trend continues, it just keeps getting bigger and bigger. Now,

436
00:27:40,040 --> 00:27:42,400
Speaker 1: while this has been an issue with Facebook for ages,

437
00:27:42,520 --> 00:27:45,960
Speaker 1: we really saw it come under scrutiny in when it

438
00:27:46,000 --> 00:27:50,720
Speaker 1: became clear that Russian operatives were leveraging Facebook to manipulate

439
00:27:50,760 --> 00:27:55,840
Speaker 1: Americans and spread disinformation, particularly when they came to politics.

440
00:27:56,680 --> 00:28:01,679
Speaker 1: Complicating matters are fake Facebook accounts also known as bot accounts.

441
00:28:02,160 --> 00:28:06,280
Speaker 1: These accounts don't represent actual users, but rather are part

442
00:28:06,400 --> 00:28:10,879
Speaker 1: of larger systems designed to amplify the signal of certain messages.

443
00:28:11,560 --> 00:28:15,280
Speaker 1: Every content creator out there would love for every post

444
00:28:15,400 --> 00:28:18,159
Speaker 1: they make to go viral, that is, to become the

445
00:28:18,200 --> 00:28:20,919
Speaker 1: sort of thing that tons of people see and share

446
00:28:20,920 --> 00:28:25,119
Speaker 1: with others. Making viral content is pretty darn hard to

447
00:28:25,160 --> 00:28:29,240
Speaker 1: do on purpose. Often the stuff happens organically, then it

448
00:28:29,280 --> 00:28:33,399
Speaker 1: becomes difficult or even impossible to recapture that same magic

449
00:28:33,440 --> 00:28:36,440
Speaker 1: on subsequent efforts. I'm sure you can think of lots

450
00:28:36,440 --> 00:28:40,240
Speaker 1: of famous viral videos from people that never seemed to

451
00:28:40,280 --> 00:28:43,000
Speaker 1: pop up again afterward, or if they did, it didn't

452
00:28:43,000 --> 00:28:47,120
Speaker 1: have nearly the same effect. There countless examples of that.

453
00:28:48,040 --> 00:28:52,040
Speaker 1: But by creating lots of fake accounts, you can amplify

454
00:28:52,160 --> 00:28:54,880
Speaker 1: a message. You can get that effect of a viral

455
00:28:55,000 --> 00:28:59,200
Speaker 1: breakout because you're manipulating the system. You're using bots that

456
00:28:59,240 --> 00:29:01,920
Speaker 1: are either all blasting out the same post at the

457
00:29:01,960 --> 00:29:05,240
Speaker 1: same time, which is actually really easy to detect. You

458
00:29:05,240 --> 00:29:08,479
Speaker 1: can often see this on Twitter, where you'll find the

459
00:29:08,520 --> 00:29:12,760
Speaker 1: exact same message posted by a bunch of seemingly disconnected

460
00:29:12,800 --> 00:29:15,600
Speaker 1: Twitter accounts, which indicates they're all bots that are coming

461
00:29:15,600 --> 00:29:18,720
Speaker 1: out of the same source, or more likely, you would

462
00:29:18,760 --> 00:29:22,920
Speaker 1: have the bots interacting with a post and sharing that

463
00:29:23,000 --> 00:29:27,400
Speaker 1: post and liking the post, thus amplifying the post's importance

464
00:29:27,840 --> 00:29:30,800
Speaker 1: on Facebook in general and making it more visible as

465
00:29:30,840 --> 00:29:34,040
Speaker 1: a result. Now, the visibility is one thing, but that

466
00:29:34,120 --> 00:29:38,520
Speaker 1: interaction also creates a perception among the real humans who

467
00:29:38,560 --> 00:29:42,719
Speaker 1: are using Facebook that that original post must have merit.

468
00:29:43,320 --> 00:29:45,800
Speaker 1: If you look at a post and you see that

469
00:29:45,880 --> 00:29:49,880
Speaker 1: it has thousands of reactions connected to it, you're more

470
00:29:49,920 --> 00:29:52,680
Speaker 1: likely to pay attention and perhaps even take that post

471
00:29:52,800 --> 00:29:56,760
Speaker 1: to heart. You know that that has a certain cash

472
00:29:56,800 --> 00:30:02,320
Speaker 1: a that that social acceptance of a piece. And by

473
00:30:02,400 --> 00:30:06,240
Speaker 1: lots of fake accounts, I really do mean lots. Last

474
00:30:06,320 --> 00:30:09,200
Speaker 1: year in twenty nineteen, Facebook announced it had shut down

475
00:30:09,200 --> 00:30:12,840
Speaker 1: more than three billion fake accounts over the course of

476
00:30:12,920 --> 00:30:18,360
Speaker 1: half a year, three billion in six months. Even after that,

477
00:30:18,720 --> 00:30:23,200
Speaker 1: Facebook estimated that somewhere around five of all accounts still

478
00:30:23,240 --> 00:30:27,120
Speaker 1: on Facebook we're actually bots. Considering Facebook has more than

479
00:30:27,160 --> 00:30:32,720
Speaker 1: two billion users, five percent represents a lot of fake accounts. Now,

480
00:30:32,800 --> 00:30:35,560
Speaker 1: this doesn't mean that every user is going to see

481
00:30:35,600 --> 00:30:39,480
Speaker 1: the posts that get that kind of engagement. The algorithm

482
00:30:39,520 --> 00:30:42,720
Speaker 1: is taking a lot more into account, for example, the

483
00:30:42,760 --> 00:30:46,160
Speaker 1: types of pages you like on the platform. That matters.

484
00:30:46,440 --> 00:30:48,720
Speaker 1: If you like a lot of pages that have a

485
00:30:48,800 --> 00:30:52,240
Speaker 1: more left leaning slant, you know, stuff that falls into

486
00:30:52,320 --> 00:30:56,560
Speaker 1: like progressive politics, that category, you're less likely to see

487
00:30:56,920 --> 00:31:00,479
Speaker 1: really hard right leaning posts as a rule, and in

488
00:31:00,520 --> 00:31:04,320
Speaker 1: this way, Facebook can become a true echo chamber. The

489
00:31:04,360 --> 00:31:07,400
Speaker 1: platform again wants to keep you there as long as possible,

490
00:31:07,880 --> 00:31:12,080
Speaker 1: as many times as possible, so the algorithm is prioritizing

491
00:31:12,200 --> 00:31:15,440
Speaker 1: content that it calculates has the best chance of keeping

492
00:31:15,440 --> 00:31:19,920
Speaker 1: you on Facebook. If they drive you away, they've done

493
00:31:20,240 --> 00:31:23,720
Speaker 1: the wrong thing, at least as far as Facebook's revenue

494
00:31:23,760 --> 00:31:28,560
Speaker 1: generating plan goes. Facebook's crackdown on fake accounts was part

495
00:31:28,680 --> 00:31:32,320
Speaker 1: of a larger effort to address the problems with misinformation

496
00:31:32,600 --> 00:31:35,760
Speaker 1: that we're growing concerning enough for the US Congress to

497
00:31:35,800 --> 00:31:39,400
Speaker 1: call Zuckerberg to Washington, d C. To answer questions about

498
00:31:39,640 --> 00:31:43,480
Speaker 1: fake news and the like. Zuckerberg had long maintained that

499
00:31:43,920 --> 00:31:46,160
Speaker 1: he didn't want Facebook to get into the business of

500
00:31:46,320 --> 00:31:49,960
Speaker 1: arbitrating which bits of information were legit and which ones

501
00:31:50,040 --> 00:31:53,360
Speaker 1: represented misinformation, as kind of this idea of it's the

502
00:31:53,480 --> 00:31:59,480
Speaker 1: free market of ideas all ideas are are given equal platforms,

503
00:31:59,840 --> 00:32:02,560
Speaker 1: And you can sort of understand his reasoning from a

504
00:32:02,640 --> 00:32:06,959
Speaker 1: business perspective, because all that engagement really meant money for

505
00:32:07,040 --> 00:32:10,640
Speaker 1: his company. But with Facebook under scrutiny and criticism, there

506
00:32:10,720 --> 00:32:13,880
Speaker 1: was a very real danger of the platform being seen

507
00:32:14,080 --> 00:32:19,160
Speaker 1: as a toxic place for advertisers. Companies that aren't too

508
00:32:19,240 --> 00:32:21,880
Speaker 1: keen on getting looped in with content that falls in

509
00:32:21,920 --> 00:32:25,280
Speaker 1: with radical ideology, we're getting nervous, and in fact, in

510
00:32:26,120 --> 00:32:30,120
Speaker 1: we've seen advertisers boycott Facebook for various reasons, some of

511
00:32:30,120 --> 00:32:34,280
Speaker 1: which relating back to Facebook's failure to address things like

512
00:32:35,080 --> 00:32:39,479
Speaker 1: groups that are radicalizing individuals, that sort of stuff. So

513
00:32:39,520 --> 00:32:43,200
Speaker 1: it does have an impact. Shutting down body accounts was

514
00:32:43,320 --> 00:32:46,880
Speaker 1: one way to diminish the spread of misinformation on Facebook

515
00:32:46,920 --> 00:32:50,840
Speaker 1: without actually taking the additional step of Facebook becoming a

516
00:32:51,040 --> 00:32:55,520
Speaker 1: judge of what is and isn't legit information. But that

517
00:32:55,600 --> 00:32:58,600
Speaker 1: wasn't quite enough to satisfy critics who are pointing out

518
00:32:58,640 --> 00:33:01,880
Speaker 1: that people were using face to spread false narratives and

519
00:33:01,920 --> 00:33:05,760
Speaker 1: to mislead people and to manipulate large populations, while Facebook

520
00:33:05,800 --> 00:33:08,360
Speaker 1: meanwhile just raking in the profits from all of that.

521
00:33:08,920 --> 00:33:11,000
Speaker 1: In fact, a recent article in The New York Times

522
00:33:11,000 --> 00:33:15,280
Speaker 1: titled on Facebook, misinformation is more popular now than in

523
00:33:16,600 --> 00:33:21,400
Speaker 1: Reporter Davey Alba shared research from the German Martial Fund

524
00:33:21,600 --> 00:33:26,000
Speaker 1: Digital which found that people were sharing, liking, and commenting

525
00:33:26,120 --> 00:33:30,240
Speaker 1: on three times as many articles from outlets that regularly

526
00:33:30,320 --> 00:33:35,120
Speaker 1: published misinformation in than they were back in ten. So

527
00:33:35,160 --> 00:33:38,120
Speaker 1: we're actually seeing more of it today. That's not a

528
00:33:38,120 --> 00:33:41,680
Speaker 1: great trend there, and it's another indication that Facebook is

529
00:33:41,760 --> 00:33:45,960
Speaker 1: reluctant to intervene in this and as Karen Cornbla, the

530
00:33:46,040 --> 00:33:49,880
Speaker 1: director of the research firm, would say, counteracting this trend

531
00:33:50,080 --> 00:33:55,360
Speaker 1: quote just runs against their economic incentives. In the quote,

532
00:33:56,360 --> 00:34:01,760
Speaker 1: when we come back, it gets worse. But first let's

533
00:34:01,800 --> 00:34:11,760
Speaker 1: take another quick break. So I mentioned before the break

534
00:34:12,040 --> 00:34:14,880
Speaker 1: that it gets worse, and early in this episode I

535
00:34:14,920 --> 00:34:18,959
Speaker 1: alluded to the concept of targeted advertising, and this gets

536
00:34:18,960 --> 00:34:23,920
Speaker 1: into another aspect of algorithms on Facebook that can be harmful. Okay,

537
00:34:24,000 --> 00:34:28,720
Speaker 1: So the basic idea behind targeted advertising is really simple,

538
00:34:29,080 --> 00:34:31,120
Speaker 1: and the goal is to get the right ads in

539
00:34:31,200 --> 00:34:33,920
Speaker 1: front of the right people. The goal is always to

540
00:34:34,040 --> 00:34:37,320
Speaker 1: increase the odds that someone is going to act upon

541
00:34:37,360 --> 00:34:41,359
Speaker 1: that ad. Otherwise advertising is just throwing money away, right,

542
00:34:41,960 --> 00:34:45,560
Speaker 1: So going back to my billboard example from earlier, there's

543
00:34:45,600 --> 00:34:48,520
Speaker 1: only so much targeting you can do with billboards. Now,

544
00:34:48,560 --> 00:34:50,160
Speaker 1: you might choose to put up an add on a

545
00:34:50,200 --> 00:34:53,120
Speaker 1: billboard that's in a part of town that most closely

546
00:34:53,239 --> 00:34:57,760
Speaker 1: matches the demographic of your average customer, Meaning the people

547
00:34:58,000 --> 00:35:00,960
Speaker 1: that you cater to happen to live in a certain

548
00:35:01,000 --> 00:35:03,000
Speaker 1: part of town, so it makes more sense to put

549
00:35:03,000 --> 00:35:05,560
Speaker 1: your billboard in that part of town. But that's a

550
00:35:05,640 --> 00:35:11,120
Speaker 1: pretty primitive approach to targeted advertising. Facebook provides a laser focused,

551
00:35:11,440 --> 00:35:17,640
Speaker 1: individual precision approach. Every user activity on Facebook gives Facebook

552
00:35:17,760 --> 00:35:21,440
Speaker 1: more information about the the user in question and what

553
00:35:21,520 --> 00:35:25,760
Speaker 1: they like, which is obvious, but we need to start there.

554
00:35:25,840 --> 00:35:29,320
Speaker 1: So the pages that you visit on Facebook, the posts

555
00:35:29,320 --> 00:35:32,600
Speaker 1: you interact with on Facebook, the general information in your

556
00:35:32,640 --> 00:35:37,160
Speaker 1: profile like your birth date, your location, your relationship status,

557
00:35:37,640 --> 00:35:41,040
Speaker 1: all of these are valuable pieces of information. You can

558
00:35:41,080 --> 00:35:44,920
Speaker 1: actually go into Facebook's settings and the ad preferences page

559
00:35:45,120 --> 00:35:48,760
Speaker 1: and you can see what Facebook has deduced about you.

560
00:35:48,760 --> 00:35:51,600
Speaker 1: You know, the company will list out what it thinks

561
00:35:51,640 --> 00:35:54,719
Speaker 1: your interests are, so you can see, oh, this is

562
00:35:54,760 --> 00:35:57,200
Speaker 1: why I'm getting all of those ads for such and such.

563
00:35:57,239 --> 00:36:00,920
Speaker 1: It's because Facebook thinks I'm really interested in that. For example,

564
00:36:01,480 --> 00:36:04,839
Speaker 1: when I was first getting into exercising, which I really

565
00:36:04,880 --> 00:36:08,120
Speaker 1: need to get back to. Uh, Facebook would serve me

566
00:36:08,239 --> 00:36:11,719
Speaker 1: up all these ads for things like muscle enhancement, protein

567
00:36:11,840 --> 00:36:15,879
Speaker 1: shake things. They're all these insanely buff dudes popping up

568
00:36:15,920 --> 00:36:18,959
Speaker 1: on my Facebook profile news feed, and I was like, well,

569
00:36:19,000 --> 00:36:21,680
Speaker 1: you know, good on you guys. You've done some great work.

570
00:36:21,719 --> 00:36:25,320
Speaker 1: But that's not really it's not really my jam. But anyway,

571
00:36:25,440 --> 00:36:28,160
Speaker 1: that that is one way that Facebook starts to build

572
00:36:28,160 --> 00:36:30,600
Speaker 1: out information on you. But that's just the tip of

573
00:36:30,640 --> 00:36:35,839
Speaker 1: the iceberg. Facebook also has marketing partners, a lot of

574
00:36:35,960 --> 00:36:40,360
Speaker 1: marketing partners, and these partners are also collecting information about

575
00:36:40,400 --> 00:36:43,480
Speaker 1: the people who are visiting their pages and their habits,

576
00:36:43,520 --> 00:36:47,319
Speaker 1: including stuff like what items people might have purchased or

577
00:36:47,400 --> 00:36:50,520
Speaker 1: perhaps just searched for or looked at. Like think of

578
00:36:50,560 --> 00:36:54,040
Speaker 1: something that you've searched for, maybe on a site like Amazon,

579
00:36:54,680 --> 00:36:56,840
Speaker 1: and you're looking at stuff you haven't pulled the trigger

580
00:36:56,880 --> 00:37:00,560
Speaker 1: on buying anything yet, but you're just kind of comparison shopping. Well,

581
00:37:00,719 --> 00:37:04,160
Speaker 1: these sort of partners share that data back with Facebook,

582
00:37:04,760 --> 00:37:08,960
Speaker 1: and then Facebook can leverage that information and target specific

583
00:37:09,000 --> 00:37:12,600
Speaker 1: ads to you based on what you've been doing off

584
00:37:12,640 --> 00:37:15,200
Speaker 1: of Facebook in other parts of the web. So if

585
00:37:15,200 --> 00:37:17,439
Speaker 1: you ever done a search for something like an air

586
00:37:17,560 --> 00:37:20,440
Speaker 1: fryer and then you notice that whenever you go to

587
00:37:20,480 --> 00:37:24,200
Speaker 1: Facebook you're seeing ads for air fryers and other kitchen

588
00:37:24,239 --> 00:37:28,200
Speaker 1: gadgets just in your news feed, that's what's going on now.

589
00:37:28,200 --> 00:37:30,280
Speaker 1: I saw this happen a lot when I was shopping

590
00:37:30,280 --> 00:37:33,160
Speaker 1: around for a guitar. Every time I would log onto Facebook,

591
00:37:33,440 --> 00:37:36,000
Speaker 1: I would see a ton of ads for various music shops,

592
00:37:36,640 --> 00:37:39,759
Speaker 1: some of which were clearly scams. There's a lot of

593
00:37:39,760 --> 00:37:42,480
Speaker 1: that on Facebook. Whether you'll see an ad for you know,

594
00:37:42,560 --> 00:37:44,239
Speaker 1: you'll see an ad for something that looks like an

595
00:37:44,280 --> 00:37:49,640
Speaker 1: incredible product, something there's phenomenal, amazing mask or costume piece,

596
00:37:49,800 --> 00:37:54,600
Speaker 1: or you know, a special kit that seems to be

597
00:37:54,640 --> 00:37:58,120
Speaker 1: catered right to your interests, and more than that, it's

598
00:37:58,160 --> 00:38:00,520
Speaker 1: an unbelievably low price, like this is gonna be for

599
00:38:00,520 --> 00:38:02,840
Speaker 1: forty bucks. You're like, wow, that would normally be hundreds

600
00:38:02,840 --> 00:38:06,160
Speaker 1: of dollars. Well, this falls into the general category of

601
00:38:06,440 --> 00:38:09,000
Speaker 1: if it sounds too good to be true, it probably is.

602
00:38:09,480 --> 00:38:12,200
Speaker 1: A lot of these company pages are really just fronts

603
00:38:12,200 --> 00:38:17,400
Speaker 1: for stores that sell cheap knockoffs, often with huge shipping

604
00:38:17,440 --> 00:38:19,480
Speaker 1: costs because a lot of them are located in places

605
00:38:19,480 --> 00:38:22,600
Speaker 1: like China. So you order the thing that looks incredible,

606
00:38:22,760 --> 00:38:24,480
Speaker 1: you're like, wow, this would normally set me back a

607
00:38:24,560 --> 00:38:27,759
Speaker 1: grant and they're asking for forty bucks. And then if

608
00:38:27,800 --> 00:38:30,759
Speaker 1: you do get a product it doesn't resemble what you

609
00:38:30,800 --> 00:38:33,120
Speaker 1: were promised. You end up getting a super cheap knockoff

610
00:38:33,239 --> 00:38:36,480
Speaker 1: of whatever it was you were ordering it. But the

611
00:38:36,520 --> 00:38:40,760
Speaker 1: expense of returning that product would be so much higher

612
00:38:40,840 --> 00:38:44,439
Speaker 1: because of shipping costs, that it would not make any

613
00:38:44,480 --> 00:38:46,560
Speaker 1: sense to return it. You would still be out more

614
00:38:46,680 --> 00:38:49,600
Speaker 1: money by returning the product, so you just kind of

615
00:38:49,640 --> 00:38:51,719
Speaker 1: suck it up. But that's kind of a tangent. Let's

616
00:38:51,800 --> 00:38:55,759
Speaker 1: let's get back to targeting now. After some criticism, Facebook

617
00:38:55,800 --> 00:38:58,600
Speaker 1: has made it possible for people to tweak their settings

618
00:38:58,640 --> 00:39:02,839
Speaker 1: on the platform to bit targeted advertising, So you can

619
00:39:02,920 --> 00:39:05,080
Speaker 1: change things so that the ads you get are more

620
00:39:05,200 --> 00:39:08,960
Speaker 1: general and they're no longer hinging upon your identity or

621
00:39:09,000 --> 00:39:13,080
Speaker 1: your behaviors. Not that Facebook isn't still collecting of these information.

622
00:39:13,160 --> 00:39:16,120
Speaker 1: They are, it's just you're not seeing it in the

623
00:39:16,160 --> 00:39:20,040
Speaker 1: effect of the advertising that appears on Facebook. But it's

624
00:39:20,080 --> 00:39:22,520
Speaker 1: not always easy to spot these settings, and a lot

625
00:39:22,520 --> 00:39:24,719
Speaker 1: of people just never bother to even check into it.

626
00:39:25,120 --> 00:39:28,279
Speaker 1: And that suits Facebook just fine, because remember when I

627
00:39:28,280 --> 00:39:32,399
Speaker 1: said Facebook can charge more for ads because of their

628
00:39:32,520 --> 00:39:35,319
Speaker 1: high engagement, the fact that people spend so much time

629
00:39:35,320 --> 00:39:39,080
Speaker 1: on Facebook. That gets boosted even more when Facebook can

630
00:39:39,080 --> 00:39:43,080
Speaker 1: tell an advertiser that they're able to put their ads

631
00:39:43,239 --> 00:39:45,440
Speaker 1: in front of the people who are most likely to

632
00:39:45,480 --> 00:39:48,560
Speaker 1: respond to those ads. So not only are people spending

633
00:39:48,600 --> 00:39:52,640
Speaker 1: more time on the platform, Facebook can tell the advertisers

634
00:39:52,760 --> 00:39:55,200
Speaker 1: you want to you want to target this specific person

635
00:39:55,600 --> 00:39:58,759
Speaker 1: because they're the one most likely to respond to your ad. Now,

636
00:39:58,800 --> 00:40:02,399
Speaker 1: let me be clear, this doesn't have to be a

637
00:40:02,480 --> 00:40:05,600
Speaker 1: bad thing necessarily. I mean, you might end up seeing

638
00:40:05,640 --> 00:40:08,760
Speaker 1: ads that are really relevant to you and they could

639
00:40:08,800 --> 00:40:11,600
Speaker 1: be really helpful. Maybe you would see something that you

640
00:40:11,640 --> 00:40:14,719
Speaker 1: wouldn't have come across otherwise. But it can also come

641
00:40:14,760 --> 00:40:18,480
Speaker 1: across as really invasive or creeping. But now let's go

642
00:40:18,640 --> 00:40:21,959
Speaker 1: to something that is just a bad thing. Period See

643
00:40:22,040 --> 00:40:25,920
Speaker 1: targeting ads to get your most likely customer. Sounds fairly

644
00:40:26,120 --> 00:40:28,760
Speaker 1: on the up and up, except the practice can also

645
00:40:28,800 --> 00:40:33,640
Speaker 1: be discriminatory. For example, let's say there's a big company

646
00:40:33,719 --> 00:40:38,000
Speaker 1: and they own upscale apartment complexes in different cities, and

647
00:40:38,080 --> 00:40:41,000
Speaker 1: this company wants to advertise on Facebook because it's opening

648
00:40:41,080 --> 00:40:44,360
Speaker 1: up a brand new apartment building. And in the process,

649
00:40:44,640 --> 00:40:48,640
Speaker 1: the company specifically opts to show the ad only to

650
00:40:48,800 --> 00:40:54,440
Speaker 1: white people, because that's something Facebook used to allow advertisers

651
00:40:54,480 --> 00:41:00,480
Speaker 1: to do. Facebook was allowing for racial discrimination in advertising. Now,

652
00:41:00,719 --> 00:41:04,000
Speaker 1: if you took a bird's eye view of these practices,

653
00:41:04,360 --> 00:41:08,319
Speaker 1: you would see some really ugly trends here. Ads for

654
00:41:08,400 --> 00:41:12,880
Speaker 1: companies that were hiring fast food workers were disproportionately directed

655
00:41:12,960 --> 00:41:17,480
Speaker 1: towards black users. Ads for companies hiring cleaners or nurses

656
00:41:17,840 --> 00:41:24,160
Speaker 1: disproportionately targeted women users. The practice was reinforcing disparities in

657
00:41:24,239 --> 00:41:29,799
Speaker 1: the job market. It was perpetuating and and and heightening

658
00:41:30,000 --> 00:41:34,759
Speaker 1: these disparities. It was an effect making things worse. And

659
00:41:34,840 --> 00:41:38,920
Speaker 1: this practice first came to light way back in sixteen,

660
00:41:39,400 --> 00:41:43,520
Speaker 1: but it wasn't until twenty nineteen that Facebook was dealing

661
00:41:43,520 --> 00:41:46,720
Speaker 1: with a lawsuit and the company promised that it would

662
00:41:46,840 --> 00:41:51,840
Speaker 1: end the practice. The announcement also wasn't exactly transparent. In

663
00:41:51,920 --> 00:41:55,120
Speaker 1: a post, the Facebook for Business division said that it

664
00:41:55,160 --> 00:41:58,239
Speaker 1: was removing the option as part of its quote efforts

665
00:41:58,280 --> 00:42:02,520
Speaker 1: to simplify and stream line are targeting options end quote,

666
00:42:02,800 --> 00:42:07,440
Speaker 1: and that it would be quote removing multicultural affinity segments

667
00:42:07,719 --> 00:42:11,520
Speaker 1: and encouraging advertisers to use other targeting options such as

668
00:42:11,600 --> 00:42:17,880
Speaker 1: language or culture end quote. Multicultural affinity segments was the

669
00:42:17,960 --> 00:42:23,080
Speaker 1: euphemism the company used to essentially talk about racial discrimination now.

670
00:42:23,080 --> 00:42:27,600
Speaker 1: According to a Facebook insider named Kion Lvy, team members

671
00:42:27,680 --> 00:42:31,520
Speaker 1: within Facebook had been pushing for the removal of these

672
00:42:32,280 --> 00:42:36,800
Speaker 1: factors for advertisers for more than three years. In addition,

673
00:42:37,160 --> 00:42:40,399
Speaker 1: it's not the only discrimination that people have faced when

674
00:42:40,440 --> 00:42:44,080
Speaker 1: it comes to trying to do advertising on Facebook. One

675
00:42:44,080 --> 00:42:47,040
Speaker 1: thing I didn't touch on earlier in this episode is

676
00:42:47,080 --> 00:42:50,480
Speaker 1: how brands have been affected by the various tweaks to

677
00:42:50,600 --> 00:42:54,080
Speaker 1: the news feed algorithm and the news feed itself. Brands

678
00:42:54,200 --> 00:42:57,160
Speaker 1: can make pages, and users can visit the pages, and

679
00:42:57,160 --> 00:42:59,840
Speaker 1: they can like those pages. And in the old days,

680
00:43:00,280 --> 00:43:04,040
Speaker 1: if a business posted to their page, then that update

681
00:43:04,120 --> 00:43:07,120
Speaker 1: would go out to a decent number of the followers

682
00:43:07,160 --> 00:43:10,720
Speaker 1: who had liked that page. But Facebook tweaked the news

683
00:43:10,760 --> 00:43:13,120
Speaker 1: feed a couple of times, and each time it would

684
00:43:13,120 --> 00:43:17,360
Speaker 1: affect the organic reach of a business's post. So, for example,

685
00:43:18,280 --> 00:43:22,279
Speaker 1: there is a text of Facebook page. I think. I

686
00:43:22,280 --> 00:43:24,719
Speaker 1: mean I quit Facebook a couple of weeks ago, so

687
00:43:25,080 --> 00:43:28,640
Speaker 1: I mean I assume it's still there, but it wasn't

688
00:43:28,719 --> 00:43:31,879
Speaker 1: updated frequently. And I'll be up front with you why

689
00:43:31,960 --> 00:43:35,360
Speaker 1: that is because there wasn't much point to updating the

690
00:43:35,360 --> 00:43:39,360
Speaker 1: Facebook page due to the way the news feedworks. Posting

691
00:43:39,680 --> 00:43:42,320
Speaker 1: to the text off page would mean only a small

692
00:43:42,400 --> 00:43:46,359
Speaker 1: fraction of the people who were actually following that page

693
00:43:46,560 --> 00:43:50,000
Speaker 1: would even see the update. So if a thousand people

694
00:43:50,200 --> 00:43:53,520
Speaker 1: had liked the page, maybe only ten of them get

695
00:43:53,520 --> 00:43:57,520
Speaker 1: the update. That's not a good return on investment. So

696
00:43:58,880 --> 00:44:02,360
Speaker 1: ostensibly this was on an effort to deliver a better

697
00:44:02,400 --> 00:44:06,279
Speaker 1: experience to users. You know, Zuckerberg always positioned this as

698
00:44:06,280 --> 00:44:10,040
Speaker 1: saying the ideas they wanted the users to see more

699
00:44:10,120 --> 00:44:14,160
Speaker 1: posts from the actual people they know, their friends and family,

700
00:44:14,600 --> 00:44:19,359
Speaker 1: and fewer posts from businesses and media outlets, even though

701
00:44:19,640 --> 00:44:24,840
Speaker 1: we should remember these were outlets and businesses that users

702
00:44:24,840 --> 00:44:30,160
Speaker 1: had gone to and liked on Facebook, so presumably the

703
00:44:30,960 --> 00:44:34,799
Speaker 1: posts from those entities were things that the users were

704
00:44:34,840 --> 00:44:38,359
Speaker 1: interested in the first place. But what was really going

705
00:44:38,400 --> 00:44:42,360
Speaker 1: on here was that Facebook was creating a market for itself.

706
00:44:42,920 --> 00:44:46,279
Speaker 1: Why give away free access to users when you can

707
00:44:46,320 --> 00:44:50,880
Speaker 1: create an incentive for businesses to pay for promoted posts,

708
00:44:51,320 --> 00:44:53,960
Speaker 1: and that's what was really going on here. Facebook had

709
00:44:54,000 --> 00:44:57,840
Speaker 1: created a system where brands felt the necessity for creating

710
00:44:57,840 --> 00:45:00,000
Speaker 1: a Facebook page, and you had to have a face

711
00:45:00,000 --> 00:45:04,040
Speaker 1: fcebook page. There are two billion people on Facebook, so

712
00:45:04,120 --> 00:45:06,000
Speaker 1: you want to be where the people are, you gotta

713
00:45:06,040 --> 00:45:09,279
Speaker 1: have a Facebook page. But to reach a significant percentage

714
00:45:09,280 --> 00:45:11,360
Speaker 1: of the people who have actually gone to that page

715
00:45:11,360 --> 00:45:14,120
Speaker 1: and liked it, you would also have to pay for

716
00:45:14,200 --> 00:45:17,759
Speaker 1: that privilege. Now, I'm not going to argue against Facebook's

717
00:45:17,800 --> 00:45:21,279
Speaker 1: strategy here. It kind of stinks if you happen to

718
00:45:21,360 --> 00:45:24,759
Speaker 1: run a brand. But from Facebook's perspective, I can understand

719
00:45:24,800 --> 00:45:29,200
Speaker 1: the reasoning. So while it kind of is tough if

720
00:45:29,239 --> 00:45:33,239
Speaker 1: you're someone who's trying to reach your fans, uh, from

721
00:45:33,239 --> 00:45:35,960
Speaker 1: Facebook's perspective, I get the I get the idea, I

722
00:45:36,000 --> 00:45:40,399
Speaker 1: get the business case. However, that discrimination that we would

723
00:45:40,440 --> 00:45:43,319
Speaker 1: see directed toward users would creep in again, this time

724
00:45:43,320 --> 00:45:48,840
Speaker 1: towards advertisers. There's a band called Unsung Lily, which is

725
00:45:48,880 --> 00:45:52,840
Speaker 1: an independent pop band. It's headed by a same sex couple,

726
00:45:53,320 --> 00:45:57,440
Speaker 1: Frankie and Sarah Golding Young. These two are attempting to

727
00:45:57,480 --> 00:46:02,279
Speaker 1: promote an album on Facebook this year, in because, like

728
00:46:02,320 --> 00:46:05,320
Speaker 1: I think pretty much everyone in the world, the pandemic

729
00:46:05,440 --> 00:46:08,800
Speaker 1: changed all of their plans for the year. They wouldn't

730
00:46:08,800 --> 00:46:11,040
Speaker 1: be able to tour, they wouldn't be able to market

731
00:46:11,080 --> 00:46:14,160
Speaker 1: their album the way they would typically do so, and

732
00:46:14,200 --> 00:46:17,640
Speaker 1: so they decided to run and add on Facebook. But

733
00:46:17,760 --> 00:46:21,480
Speaker 1: they also knew that if they relied solely on posting

734
00:46:21,760 --> 00:46:24,279
Speaker 1: the the ad to the band's page, like if they

735
00:46:24,320 --> 00:46:28,520
Speaker 1: just made a post on their fan page, the way

736
00:46:28,600 --> 00:46:31,040
Speaker 1: Facebook serves up that kind of stuff means that they

737
00:46:31,040 --> 00:46:34,440
Speaker 1: wouldn't reach nearly enough of their fans for it to

738
00:46:34,480 --> 00:46:37,879
Speaker 1: make an impact. So they decided to do it as

739
00:46:37,920 --> 00:46:41,160
Speaker 1: an paid ad instead. So they shot an ad, they

740
00:46:41,200 --> 00:46:44,399
Speaker 1: put it together, they edited it, they submitted it. They

741
00:46:44,400 --> 00:46:47,920
Speaker 1: received a rejection, and under the reason why the ad

742
00:46:47,920 --> 00:46:51,680
Speaker 1: had been rejected, it said that the ad contained inappropriate

743
00:46:51,760 --> 00:46:57,280
Speaker 1: sexually explicit content, except it didn't. What it did feature

744
00:46:57,800 --> 00:47:01,400
Speaker 1: was an image of Sarah and Frank Key leaning into

745
00:47:01,440 --> 00:47:04,560
Speaker 1: each other with their foreheads touching, kind of just an

746
00:47:04,600 --> 00:47:09,640
Speaker 1: image of affection. There was nothing remotely sexual or explicit

747
00:47:09,880 --> 00:47:15,200
Speaker 1: about this photograph. Now, the rejection notice didn't specify that

748
00:47:15,320 --> 00:47:18,879
Speaker 1: the image was the reason for the rejection, so they

749
00:47:18,920 --> 00:47:21,040
Speaker 1: decided they would run a test to see if, in

750
00:47:21,120 --> 00:47:24,960
Speaker 1: fact that was the problem. First, they submitted an AD

751
00:47:25,000 --> 00:47:28,320
Speaker 1: that had all the same content as the original piece,

752
00:47:28,680 --> 00:47:32,680
Speaker 1: only with this one image swapped out for a quote

753
00:47:32,719 --> 00:47:38,240
Speaker 1: unquote non romantic picture of the couple boom Facebook excepts

754
00:47:38,239 --> 00:47:43,160
Speaker 1: the ad. To follow that up, they also submitted another ad.

755
00:47:43,200 --> 00:47:46,520
Speaker 1: They still used the same video, the same copy, but

756
00:47:46,600 --> 00:47:50,200
Speaker 1: they got a heterosexual couple making the same pose that

757
00:47:50,280 --> 00:47:52,680
Speaker 1: they had made with the two four heads together, so

758
00:47:52,680 --> 00:47:56,160
Speaker 1: it's a man and a woman foreheads touching. They just

759
00:47:56,239 --> 00:47:59,960
Speaker 1: used that to replace their own image. Boom Facebook except

760
00:48:00,200 --> 00:48:03,239
Speaker 1: that ad too. Okay, so I guess I don't have

761
00:48:03,280 --> 00:48:06,960
Speaker 1: to explain that this is a really bad thing. The

762
00:48:07,000 --> 00:48:11,160
Speaker 1: fact that Facebook ads deemed a heterosexual display of affection

763
00:48:11,680 --> 00:48:15,600
Speaker 1: as perfectly acceptable and the same sex display of the

764
00:48:15,680 --> 00:48:21,280
Speaker 1: exact same act of affection as offensive is beyond infuriating.

765
00:48:22,000 --> 00:48:25,920
Speaker 1: The band contacted the American Civil Liberties Union, which facilitated

766
00:48:25,960 --> 00:48:28,719
Speaker 1: some communication with Facebook. Facebook said, oh, no, no, no no,

767
00:48:28,880 --> 00:48:31,920
Speaker 1: this whole thing was a mistake. And besides, the actual

768
00:48:31,960 --> 00:48:35,319
Speaker 1: objectionable part wasn't that photograph. It was because of some

769
00:48:35,400 --> 00:48:39,520
Speaker 1: of the dancing those in the video. But again, Facebook

770
00:48:39,600 --> 00:48:43,600
Speaker 1: accepted two versions of the video that contained the dancing,

771
00:48:43,680 --> 00:48:47,520
Speaker 1: It just didn't have that photograph of Sarah and Frankie,

772
00:48:47,920 --> 00:48:52,440
Speaker 1: which seems pretty fishy to me. Meanwhile, the company continues

773
00:48:52,480 --> 00:48:56,160
Speaker 1: to face criticism that it fails to curb truly objectionable

774
00:48:56,239 --> 00:49:00,279
Speaker 1: stuff like hate speech. So hate speech spreads like easy

775
00:49:00,320 --> 00:49:04,000
Speaker 1: on Facebook, but an ad that had a photograph of

776
00:49:04,239 --> 00:49:09,000
Speaker 1: two women with their foreheads together was offensive. There are

777
00:49:09,000 --> 00:49:12,440
Speaker 1: studies that are also showing that black users are disproportionately

778
00:49:12,520 --> 00:49:16,839
Speaker 1: more likely to have their accounts suspended by automated moderating

779
00:49:16,960 --> 00:49:21,320
Speaker 1: systems than white users are, which is another bad algorithm example.

780
00:49:22,000 --> 00:49:25,719
Speaker 1: And there are real concerns that the targeting criteria Facebook

781
00:49:25,760 --> 00:49:31,080
Speaker 1: continues to use exacerbate socioeconomic disparities in the real world.

782
00:49:31,960 --> 00:49:35,759
Speaker 1: None of this is good news. Facebook has been under

783
00:49:36,120 --> 00:49:41,319
Speaker 1: intense scrutiny recently from various governments around the world that

784
00:49:41,400 --> 00:49:45,480
Speaker 1: have all floated variations of the possibility that Facebook could

785
00:49:45,640 --> 00:49:48,960
Speaker 1: face some really big consequences down the road, like getting

786
00:49:48,960 --> 00:49:51,719
Speaker 1: broken up. There are a lot of different entities that

787
00:49:51,840 --> 00:49:56,160
Speaker 1: argue Facebook is a monopoly or is monopolistic, and it's

788
00:49:56,160 --> 00:50:00,360
Speaker 1: pretty clear that Facebook and its incarnation as it exists

789
00:50:00,440 --> 00:50:04,120
Speaker 1: right now really has some big problems to iron out,

790
00:50:04,640 --> 00:50:08,200
Speaker 1: things that are affecting people in a truly negative way.

791
00:50:08,320 --> 00:50:13,080
Speaker 1: These are all parts of the reason why I left Facebook.

792
00:50:13,200 --> 00:50:16,399
Speaker 1: I did so for my own personal reasons, and I'm

793
00:50:16,400 --> 00:50:20,040
Speaker 1: not advocating that anyone listening to this just go out

794
00:50:20,040 --> 00:50:22,400
Speaker 1: and shut down your account right now. I'm not saying

795
00:50:22,440 --> 00:50:26,520
Speaker 1: that for me that was the right decision. I needed

796
00:50:26,560 --> 00:50:30,600
Speaker 1: to do it for my own mental health, honestly, So

797
00:50:31,560 --> 00:50:34,160
Speaker 1: I felt like this was an important thing to talk about,

798
00:50:34,239 --> 00:50:37,080
Speaker 1: because again, it's those algorithms that, at the end of

799
00:50:37,080 --> 00:50:42,080
Speaker 1: the day, we're really just designed to optimize the amount

800
00:50:42,080 --> 00:50:44,320
Speaker 1: of time people would spend on Facebook and the number

801
00:50:44,480 --> 00:50:47,680
Speaker 1: and quality, or at least the variety of ads that

802
00:50:47,760 --> 00:50:51,400
Speaker 1: they would see while they spent their time there. That

803
00:50:51,520 --> 00:50:54,719
Speaker 1: was really the only thing that mattered to Facebook. It

804
00:50:54,800 --> 00:50:58,640
Speaker 1: was otherwise kind of distanced from any of the content stuff.

805
00:50:58,760 --> 00:51:02,799
Speaker 1: Content was not at all the concern. The concern was

806
00:51:03,400 --> 00:51:05,480
Speaker 1: how long can we keep them there, how many ads

807
00:51:05,480 --> 00:51:07,399
Speaker 1: can we show them, how much money can we get

808
00:51:07,440 --> 00:51:10,080
Speaker 1: for those ads? That was really all that mattered. And

809
00:51:10,160 --> 00:51:14,719
Speaker 1: now we're starting to see when you disassociate the content

810
00:51:15,080 --> 00:51:19,280
Speaker 1: from the approach how that can have a profoundly negative

811
00:51:19,280 --> 00:51:22,600
Speaker 1: impact on the world, and we're trying to get to

812
00:51:22,600 --> 00:51:24,680
Speaker 1: the point where we figure out what to do next.

813
00:51:25,520 --> 00:51:29,240
Speaker 1: I don't have the answers um other than I walked away.

814
00:51:30,960 --> 00:51:35,480
Speaker 1: So hopefully people at Facebook will start to come up

815
00:51:35,520 --> 00:51:39,120
Speaker 1: with ways to approach the business from a more ethical

816
00:51:39,200 --> 00:51:43,359
Speaker 1: standpoint that will have less of a negative impact on

817
00:51:43,520 --> 00:51:45,759
Speaker 1: the people who actually use the service, whether it's the

818
00:51:45,800 --> 00:51:50,279
Speaker 1: advertisers or the end users. I certainly hope. So if

819
00:51:50,320 --> 00:51:52,640
Speaker 1: you found this interesting, let me know if you have

820
00:51:52,840 --> 00:51:56,240
Speaker 1: suggestions for future topics on tech stuff. Whether it's a company,

821
00:51:56,520 --> 00:51:59,720
Speaker 1: a trend in technology, maybe it's a person in tech.

822
00:52:00,400 --> 00:52:03,239
Speaker 1: Maybe it's a specific technology you've always wanted to know

823
00:52:03,320 --> 00:52:05,960
Speaker 1: how it worked. Let me know. The best way to

824
00:52:06,000 --> 00:52:09,080
Speaker 1: do that is clearly to go over to Twitter. Not

825
00:52:09,239 --> 00:52:12,600
Speaker 1: that that's a perfect service, but that's a different podcast

826
00:52:12,640 --> 00:52:15,840
Speaker 1: for a different time. But use the handle text stuff

827
00:52:16,160 --> 00:52:20,160
Speaker 1: h s W and I'll talk to you again really soon.

828
00:52:24,840 --> 00:52:27,839
Speaker 1: Text Stuff is an I Heart Radio production. For more

829
00:52:27,920 --> 00:52:31,320
Speaker 1: podcasts from my Heart Radio, visit the i Heart Radio app,

830
00:52:31,440 --> 00:52:34,600
Speaker 1: Apple Podcasts, or wherever you listen to your favorite shows.