1 00:00:05,040 --> 00:00:07,480 Speaker 1: From Bloomberg News and I Heart Radio. It's the big Take. 2 00:00:07,800 --> 00:00:12,319 Speaker 1: I'm West Casova today. Why TikTok and young kids are 3 00:00:12,360 --> 00:00:29,520 Speaker 1: a bad mix? Just a note about today's episode. It's 4 00:00:29,600 --> 00:00:32,760 Speaker 1: on a difficult topic and parts of it are not 5 00:00:32,880 --> 00:00:36,199 Speaker 1: easy to listen to. If there are children nearby, you 6 00:00:36,280 --> 00:00:42,960 Speaker 1: might want to use headphones. Bloomberg Senior reporter Olivia Carville 7 00:00:43,000 --> 00:00:48,159 Speaker 1: spent months investigating the deaths of children who participated in 8 00:00:48,360 --> 00:00:52,440 Speaker 1: dangerous challenges that can be seen on TikTok, the enormously 9 00:00:52,520 --> 00:00:57,040 Speaker 1: popular video sharing social network. Her reporting shows how the 10 00:00:57,120 --> 00:01:00,880 Speaker 1: company has tried and failed to stop young kids, some 11 00:01:00,960 --> 00:01:04,640 Speaker 1: of them under ten years old, from joining the platform, 12 00:01:04,680 --> 00:01:06,880 Speaker 1: even though they're not supposed to be allowed on it. 13 00:01:07,480 --> 00:01:09,880 Speaker 1: I asked Olivia to tell us what she found in 14 00:01:09,920 --> 00:01:13,600 Speaker 1: her reporting for a Bloomberg Business Week, starting with why 15 00:01:13,720 --> 00:01:20,959 Speaker 1: TikTok has such appeal for kids? What kids want to 16 00:01:21,000 --> 00:01:24,039 Speaker 1: be when they grow up as an influencer more so 17 00:01:24,080 --> 00:01:28,720 Speaker 1: than an astronaut or a teacher or a journalist. They 18 00:01:28,720 --> 00:01:32,119 Speaker 1: want to be a YouTube or TikTok influencer. They want 19 00:01:32,120 --> 00:01:35,760 Speaker 1: to be social media famous. They are communicating with a 20 00:01:35,800 --> 00:01:39,640 Speaker 1: community of like minded people who are creating cool content, 21 00:01:39,959 --> 00:01:42,560 Speaker 1: and they're looking at people that they aspire to be. 22 00:01:43,000 --> 00:01:46,560 Speaker 1: We're talking more than two thirds of all teenagers in 23 00:01:46,600 --> 00:01:51,040 Speaker 1: America use TikTok. Millions of kids every day are on 24 00:01:51,040 --> 00:01:55,560 Speaker 1: this app more than Facebook, more than Snapchat, more than Instagram. 25 00:01:55,600 --> 00:01:58,520 Speaker 1: TikTok is the coolest place to be on the Internet 26 00:01:58,640 --> 00:02:02,120 Speaker 1: right now. One thing it's made TikTok so popular are 27 00:02:02,120 --> 00:02:06,400 Speaker 1: the viral video challenges users post for others to copy, 28 00:02:06,440 --> 00:02:10,560 Speaker 1: like a dance or a song. Overnight, millions of TikTokers 29 00:02:10,600 --> 00:02:14,120 Speaker 1: are posting videos of themselves doing their own version of it. 30 00:02:16,840 --> 00:02:19,880 Speaker 1: So the challenges on social media apps like TikTok really 31 00:02:19,919 --> 00:02:22,680 Speaker 1: started in two thousand and twenty, when the pandemic shut 32 00:02:22,720 --> 00:02:28,000 Speaker 1: down schools and kids were at home just on their phones. Initially, 33 00:02:28,040 --> 00:02:31,560 Speaker 1: these challenges were pretty harmless. They were kind of cute. 34 00:02:31,800 --> 00:02:35,520 Speaker 1: I saw pits being dressed up in different outfits and 35 00:02:35,919 --> 00:02:39,480 Speaker 1: four generations of one family dancing in the same way. 36 00:02:39,680 --> 00:02:44,040 Speaker 1: But as is usually typical when children or teenagers are involved, 37 00:02:44,080 --> 00:02:47,880 Speaker 1: they started to get edgier or more dangerous. The crate 38 00:02:48,080 --> 00:02:52,080 Speaker 1: challenge where kids would climb stepped milk crates and some 39 00:02:52,200 --> 00:02:55,480 Speaker 1: of them would fall and injure themselves. We're talking milk 40 00:02:55,560 --> 00:02:58,919 Speaker 1: crates that were like eight crates high. Or they do 41 00:02:59,160 --> 00:03:02,680 Speaker 1: challenges lie the cinnamon challenge, where they'd eat a spoonful 42 00:03:02,720 --> 00:03:06,080 Speaker 1: of cinnamon, which can actually cause minor lung damage if 43 00:03:06,120 --> 00:03:09,040 Speaker 1: they inhale it into their lungs. They were doing things 44 00:03:09,080 --> 00:03:13,480 Speaker 1: like dropping pennies into a partially plugged in phone charger 45 00:03:13,840 --> 00:03:16,800 Speaker 1: waiting for it to spark, and in some cases this 46 00:03:16,960 --> 00:03:24,800 Speaker 1: resulted in catching light and property destruction charges. We'd see 47 00:03:24,880 --> 00:03:30,720 Speaker 1: even more dangerous challenges throughout with the skull breaker challenge, 48 00:03:30,880 --> 00:03:33,680 Speaker 1: where two people would encourage a third to jump in 49 00:03:33,720 --> 00:03:36,080 Speaker 1: the ear and then they'd kick their legs out from 50 00:03:36,160 --> 00:03:39,360 Speaker 1: under them and they had land on the ground, sometimes 51 00:03:39,360 --> 00:03:42,640 Speaker 1: on their head, and this would result in concussions. We 52 00:03:42,720 --> 00:03:46,520 Speaker 1: also saw the fire accelerant challenge, where kids would actually 53 00:03:46,640 --> 00:03:52,440 Speaker 1: light themselves on fire accidentally. We've seen children who are seven, eight, 54 00:03:52,760 --> 00:03:56,320 Speaker 1: nine years old, who are too young to be on 55 00:03:56,360 --> 00:04:02,520 Speaker 1: TikTok participating in dangerous and tragically, in some cases, deadly challenges. 56 00:04:03,240 --> 00:04:05,720 Speaker 1: You heard Olivia say just now that children this young 57 00:04:05,800 --> 00:04:09,320 Speaker 1: are too young to be on TikTok. That's because TikTok, 58 00:04:09,440 --> 00:04:13,720 Speaker 1: like other social media platforms, has an age gate, kids 59 00:04:13,800 --> 00:04:16,760 Speaker 1: under the age of thirteen aren't allowed to have full 60 00:04:16,839 --> 00:04:20,520 Speaker 1: access to the apps content. What's interesting about this age 61 00:04:20,560 --> 00:04:25,080 Speaker 1: gate is actually that the number thirteen is largely arbitrary. 62 00:04:25,480 --> 00:04:30,279 Speaker 1: There's no scientific research or pediatricians or child psychologists that 63 00:04:30,400 --> 00:04:33,840 Speaker 1: say kids over the age of thirteen are mature enough 64 00:04:33,880 --> 00:04:37,320 Speaker 1: to handle this kind of content. The reason why platforms 65 00:04:37,320 --> 00:04:39,360 Speaker 1: say you have to be over the age of thirteen 66 00:04:40,000 --> 00:04:42,799 Speaker 1: to get access to them is actually because they don't 67 00:04:42,800 --> 00:04:47,760 Speaker 1: want to violate already existing child privacy laws in the US. 68 00:04:48,120 --> 00:04:51,120 Speaker 1: Children under thirteen are supposed to be directed to a 69 00:04:51,279 --> 00:04:54,640 Speaker 1: walled off version of the platform where the videos are 70 00:04:54,760 --> 00:04:59,120 Speaker 1: scrubbed of content that TikTok says isn't appropriate for kids. 71 00:05:00,080 --> 00:05:04,000 Speaker 1: Art Kids are smart, they're digitally savvy, and they don't 72 00:05:04,000 --> 00:05:06,440 Speaker 1: want to be on the child version of TikTok. They 73 00:05:06,440 --> 00:05:08,600 Speaker 1: want to be on the cool version of TikTok. They 74 00:05:08,600 --> 00:05:10,640 Speaker 1: want to be on the adult version of TikTok. So 75 00:05:10,680 --> 00:05:14,560 Speaker 1: they're lying about their age to gain access to the platform. 76 00:05:14,600 --> 00:05:17,839 Speaker 1: Millions of children today are on TikTok who are under 77 00:05:17,880 --> 00:05:20,840 Speaker 1: the age of thirteen when they shouldn't be. If you're 78 00:05:20,880 --> 00:05:23,599 Speaker 1: under the age of thirteen, and you try to sign 79 00:05:23,680 --> 00:05:26,520 Speaker 1: up for TikTok and you just say that you're older. 80 00:05:26,680 --> 00:05:28,719 Speaker 1: Is that it does TikTok do anything to try to 81 00:05:28,839 --> 00:05:34,479 Speaker 1: verify your age. The platform does many things to try 82 00:05:34,520 --> 00:05:38,240 Speaker 1: and weed out these underage accounts, but it's difficult. Some 83 00:05:38,320 --> 00:05:42,279 Speaker 1: of the things they do is use text based analytics 84 00:05:42,320 --> 00:05:46,240 Speaker 1: to scrape through, you know, bios and people's profile pages. 85 00:05:46,600 --> 00:05:48,919 Speaker 1: So say you download an account and you write in 86 00:05:48,960 --> 00:05:53,520 Speaker 1: your bio, I'm eleven years old. The artificial intelligence is 87 00:05:53,560 --> 00:05:56,760 Speaker 1: good enough to detect that and remove you. They also 88 00:05:56,839 --> 00:06:00,680 Speaker 1: get community users to report in suspected under age accounts. 89 00:06:00,920 --> 00:06:03,520 Speaker 1: If you're nine years old and on TikTok and posting 90 00:06:03,600 --> 00:06:07,400 Speaker 1: regularly as the account holder, other people who use TikTok 91 00:06:07,800 --> 00:06:10,800 Speaker 1: can report you to the platform and you'll get taken down. 92 00:06:11,120 --> 00:06:14,160 Speaker 1: They also train their moderators to look out for suspected 93 00:06:14,240 --> 00:06:18,640 Speaker 1: child accounts, and every three months the platform removes twenty 94 00:06:18,800 --> 00:06:23,520 Speaker 1: million suspected child accounts from TikTok. This is a huge number, 95 00:06:24,480 --> 00:06:29,760 Speaker 1: but TikTok has one billion users today worldwide. Millions of 96 00:06:29,839 --> 00:06:32,440 Speaker 1: children are on the platform and lying about their age 97 00:06:32,480 --> 00:06:35,599 Speaker 1: to access it. So all these measures that they're adopting 98 00:06:35,720 --> 00:06:39,040 Speaker 1: to try and find those suspected under age accounts largely 99 00:06:39,080 --> 00:06:47,200 Speaker 1: aren't working. Olivia's story is centered on an especially gruesome 100 00:06:47,320 --> 00:06:51,039 Speaker 1: video challenge that's been making the rounds on TikTok, the 101 00:06:51,080 --> 00:06:56,960 Speaker 1: Blackout Challenge. What it is is children choking themselves until 102 00:06:57,000 --> 00:07:01,360 Speaker 1: they lose consciousness, blacking out and then waking up and 103 00:07:01,360 --> 00:07:05,480 Speaker 1: getting that adrenaline rush or that high of regaining consciousness. 104 00:07:05,760 --> 00:07:09,320 Speaker 1: They're filming this and uploading it to social media as 105 00:07:09,360 --> 00:07:12,160 Speaker 1: if it's a fun, cool thing to do. The challenge 106 00:07:12,160 --> 00:07:16,040 Speaker 1: requests that you use a household object to try and 107 00:07:16,120 --> 00:07:21,880 Speaker 1: choke yourself, So in some cases I've seen children using belts, 108 00:07:22,520 --> 00:07:28,600 Speaker 1: using nick ties, using dog leashes, using shoelaces to choke 109 00:07:28,640 --> 00:07:32,680 Speaker 1: themselves until they pass out. If this sounds familiar, it 110 00:07:32,800 --> 00:07:36,000 Speaker 1: might be because choking games like the Blackhout Challenge didn't 111 00:07:36,080 --> 00:07:40,240 Speaker 1: start on TikTok. Well. The Blackout Challenge actually dates back to, 112 00:07:40,680 --> 00:07:43,200 Speaker 1: you know, as far as the nineteen thirties. It was 113 00:07:43,240 --> 00:07:47,560 Speaker 1: around before we even had an iPhone. But social media 114 00:07:47,640 --> 00:07:51,120 Speaker 1: has amplified its spread. Not only is it easier to 115 00:07:51,240 --> 00:07:55,640 Speaker 1: reach a huge range of kids, it's reaching children who 116 00:07:55,640 --> 00:07:58,600 Speaker 1: are younger and younger kids who are too young to 117 00:07:58,720 --> 00:08:02,239 Speaker 1: understand the risks. One of those children was a little 118 00:08:02,240 --> 00:08:07,360 Speaker 1: girl named Arianni Arroyo. Arianni Arroyo was a nine year 119 00:08:07,360 --> 00:08:11,440 Speaker 1: old girl who lived down in Milwaukee, Wisconsin. Artie was fashion. 120 00:08:11,880 --> 00:08:14,760 Speaker 1: Addie was glitter and glad, and she just had her 121 00:08:14,800 --> 00:08:19,800 Speaker 1: own little style and everything. That's Arianni's mother, Olivia, spoke 122 00:08:19,840 --> 00:08:23,160 Speaker 1: to both of her parents. My name is Christo a 123 00:08:23,320 --> 00:08:27,440 Speaker 1: Royal Roman, my name is Eddie Berto, a Royal Roman, 124 00:08:28,360 --> 00:08:34,800 Speaker 1: and my daughter is Arianni Jaylene Arroyo. She loved dancing, 125 00:08:35,360 --> 00:08:41,280 Speaker 1: and dancing was the first trend that she really started 126 00:08:41,320 --> 00:08:44,920 Speaker 1: engaging in TikTok. She would do all the dances she 127 00:08:45,200 --> 00:08:49,560 Speaker 1: record herself and um, it just grew from there. When 128 00:08:49,800 --> 00:08:52,640 Speaker 1: COVID nineteen shut down her school and she was at 129 00:08:52,640 --> 00:08:55,880 Speaker 1: home all the time, she became obsessed with TikTok. She 130 00:08:56,040 --> 00:08:58,680 Speaker 1: was using it every day. She was trying all of 131 00:08:58,720 --> 00:09:02,439 Speaker 1: these challenges with her younger brother, doing things like holding 132 00:09:02,480 --> 00:09:05,480 Speaker 1: water in your mouth until you start giggling, or one 133 00:09:05,559 --> 00:09:08,800 Speaker 1: challenge they tried was only eating red food for twenty 134 00:09:08,800 --> 00:09:11,640 Speaker 1: four hours. They loved it. They thought it was really 135 00:09:11,679 --> 00:09:17,520 Speaker 1: fun day. It was addicting. The TikTok is like something 136 00:09:17,600 --> 00:09:24,160 Speaker 1: that lures you in, makes a kid believe that they're 137 00:09:24,160 --> 00:09:28,000 Speaker 1: going to be famous. One evening in early two thousand 138 00:09:28,040 --> 00:09:32,080 Speaker 1: and twenty one, Arianni was playing with her little brother 139 00:09:32,120 --> 00:09:37,880 Speaker 1: before bedtime. It was a Wednesday, it was a normal day. 140 00:09:38,240 --> 00:09:41,319 Speaker 1: Her father was downstairs in the basement and his workshop. 141 00:09:42,040 --> 00:09:45,640 Speaker 1: From what we understand, she used a dog leash. This 142 00:09:45,760 --> 00:09:48,920 Speaker 1: was a middle dog leash. Her parents had recently purchased 143 00:09:48,920 --> 00:09:51,280 Speaker 1: it because the kids really wanted to get a puppy, 144 00:09:51,360 --> 00:09:53,840 Speaker 1: and they bought the dog leash to show their children 145 00:09:53,920 --> 00:09:56,400 Speaker 1: that they were willing to buy them a dog. And 146 00:09:56,480 --> 00:09:59,679 Speaker 1: they were playing with this dog leash, and Arianny she 147 00:09:59,760 --> 00:10:01,880 Speaker 1: read the middle dog leash around her neck and she 148 00:10:02,000 --> 00:10:05,199 Speaker 1: hooked the buckle onto the hinge of a wardrobe door, 149 00:10:05,840 --> 00:10:08,040 Speaker 1: and she slipped off the toy chests that she was 150 00:10:08,080 --> 00:10:11,199 Speaker 1: standing on, and she was left hanging about two ft 151 00:10:11,240 --> 00:10:15,280 Speaker 1: from the ground. She couldn't scream out. She couldn't cry out. 152 00:10:15,440 --> 00:10:18,400 Speaker 1: Her father was out of earshot anyway, and her little 153 00:10:18,440 --> 00:10:21,080 Speaker 1: brother had to run downstairs and tell his dad that 154 00:10:21,679 --> 00:10:25,880 Speaker 1: Sissy's tangled. Her father came upstairs to see what he 155 00:10:25,960 --> 00:10:29,520 Speaker 1: was talking about, and that's when he found Ariane lifeless, 156 00:10:30,120 --> 00:10:33,480 Speaker 1: hanging by this dog leash from the hinge of the 157 00:10:33,520 --> 00:10:38,960 Speaker 1: wardrobe door. He tried CPR, She never regained consciousness, and 158 00:10:39,040 --> 00:10:42,960 Speaker 1: she she died a week later. As you can imagine, 159 00:10:43,080 --> 00:10:49,000 Speaker 1: Arianne's parents are heartbroken. They told me that their happiness 160 00:10:49,040 --> 00:10:52,920 Speaker 1: had died with their daughter. That this feeling just weighs 161 00:10:52,960 --> 00:10:56,960 Speaker 1: heavy on your chest. That it's um hard to move 162 00:10:57,000 --> 00:11:00,080 Speaker 1: on and hard to continue living when you know you 163 00:11:00,120 --> 00:11:02,160 Speaker 1: lose a child in that way. One of the things 164 00:11:02,160 --> 00:11:04,720 Speaker 1: her mother said to me is that she she knew 165 00:11:04,760 --> 00:11:07,760 Speaker 1: that she had to be protective of the outside, but 166 00:11:07,840 --> 00:11:09,880 Speaker 1: what she didn't realize is that she had to be 167 00:11:10,280 --> 00:11:13,800 Speaker 1: protective of the inside. That this danger actually came through 168 00:11:13,840 --> 00:11:19,320 Speaker 1: her child's phone, a social media app. I never thought 169 00:11:19,400 --> 00:11:23,240 Speaker 1: in my mind that that could ever result to heat 170 00:11:23,280 --> 00:11:28,720 Speaker 1: losing a daughter. As we've done everything to protect our children, 171 00:11:29,440 --> 00:11:31,640 Speaker 1: and you think they're in the safety of your home. 172 00:11:33,040 --> 00:11:38,800 Speaker 1: We're home, My kids were home, everybody was home. A 173 00:11:38,840 --> 00:11:42,559 Speaker 1: week after Arianny died, her parents asked her younger brother 174 00:11:43,200 --> 00:11:46,840 Speaker 1: what they were doing the night of the accident. He 175 00:11:46,880 --> 00:11:49,560 Speaker 1: said they were playing a game like they'd seen on TikTok. 176 00:11:49,960 --> 00:11:52,720 Speaker 1: My son told us, like we played hang the Cheen, 177 00:11:53,520 --> 00:11:56,199 Speaker 1: Like what's he in the chee. He was only four 178 00:11:56,280 --> 00:11:58,240 Speaker 1: years old, so we didn't know the name of the game. 179 00:11:58,600 --> 00:12:01,400 Speaker 1: And it wasn't until after her funeral that they were 180 00:12:01,440 --> 00:12:04,480 Speaker 1: asking her friends, you know, who were her age, who 181 00:12:04,480 --> 00:12:07,280 Speaker 1: had known what she was doing around that time, and 182 00:12:07,320 --> 00:12:10,880 Speaker 1: then her friends told me, oh, that's black Out challenging, 183 00:12:10,840 --> 00:12:13,400 Speaker 1: And then I'm like, what's black out Challenge? I didn't know. 184 00:12:13,720 --> 00:12:18,400 Speaker 1: We didn't know anything like that existed. After doing some research, 185 00:12:18,559 --> 00:12:22,960 Speaker 1: they discovered another child had died from allegedly participating in 186 00:12:22,960 --> 00:12:29,160 Speaker 1: this challenge, a ten year old girl in Italy. Our 187 00:12:29,240 --> 00:12:40,360 Speaker 1: story continues after the break, I asked Olivia what more 188 00:12:40,400 --> 00:12:43,640 Speaker 1: could TikTok have done to stop a child like Arianni 189 00:12:43,800 --> 00:12:48,400 Speaker 1: Arroyo from viewing videos like the Blackout Challenge. TikTok doesn't 190 00:12:48,480 --> 00:12:51,440 Speaker 1: want this kind of content on its platform. It causes 191 00:12:51,520 --> 00:12:54,520 Speaker 1: all kinds of issues like this exact one that we're discussing. 192 00:12:54,920 --> 00:12:58,160 Speaker 1: So TikTok has done a bunch of things to try 193 00:12:58,200 --> 00:13:01,679 Speaker 1: and reduce Blackout Challenge video is one of those things 194 00:13:01,880 --> 00:13:05,400 Speaker 1: is hiring people to sit and watch hours and hours 195 00:13:05,400 --> 00:13:08,600 Speaker 1: of TikTok videos and decide if they're okay or not. 196 00:13:09,480 --> 00:13:13,880 Speaker 1: The people who do this are called content moderators. Content 197 00:13:13,920 --> 00:13:18,720 Speaker 1: moderators for social media apps are essentially watching all of 198 00:13:18,760 --> 00:13:23,120 Speaker 1: the videos uploaded to platforms like TikTok or YouTube and 199 00:13:23,679 --> 00:13:27,480 Speaker 1: trying to find any kind of content that may violate 200 00:13:27,520 --> 00:13:31,600 Speaker 1: community guidelines that may be problematic for the platforms. They're 201 00:13:31,679 --> 00:13:36,400 Speaker 1: training their human content moderators to take down any content 202 00:13:36,480 --> 00:13:40,760 Speaker 1: that promotes or glorifies dangerous behavior, like the Blackout Challenge. 203 00:13:41,120 --> 00:13:45,360 Speaker 1: They have also band hashtags connected to the Blackout Challenge 204 00:13:45,360 --> 00:13:48,600 Speaker 1: in the app. When you type in blackout challenge on TikTok, 205 00:13:48,679 --> 00:13:52,600 Speaker 1: now you're automatically sent through to a page that says 206 00:13:52,679 --> 00:13:55,440 Speaker 1: your safety matters. You know you shouldn't be looking for 207 00:13:55,559 --> 00:14:01,640 Speaker 1: dangerous content like this, and unfortunately though children have found 208 00:14:01,640 --> 00:14:04,800 Speaker 1: ways to get around that. They have renamed the Blackout Challenge. 209 00:14:04,840 --> 00:14:07,920 Speaker 1: They have replaced the oh of blackout and put a 210 00:14:08,000 --> 00:14:11,559 Speaker 1: zero in it, or used a one instead of the L, 211 00:14:12,000 --> 00:14:16,800 Speaker 1: misspelling code words, secret phrases to keep the content spreading, 212 00:14:17,080 --> 00:14:19,920 Speaker 1: keep the challenge on the app because they like participating 213 00:14:19,960 --> 00:14:22,440 Speaker 1: in it and they want to evade or avoid or 214 00:14:22,640 --> 00:14:26,360 Speaker 1: be smarter than TikTok. It's hard for them to train 215 00:14:26,440 --> 00:14:30,920 Speaker 1: moderators on what is dangerous because the word dangerous is subjective. 216 00:14:31,320 --> 00:14:33,840 Speaker 1: Maybe what you think is dangerous is different to what 217 00:14:33,960 --> 00:14:37,000 Speaker 1: I think is dangerous. So should you remove content of 218 00:14:37,160 --> 00:14:41,360 Speaker 1: children climbing stacked milk crates? Some may say no, let 219 00:14:41,360 --> 00:14:43,000 Speaker 1: the kids do what they want to do. Don't be 220 00:14:43,080 --> 00:14:46,560 Speaker 1: overly paternal and take this content down. They're just having fun. 221 00:14:47,120 --> 00:14:50,320 Speaker 1: But what happens if a child dies? Is that now 222 00:14:50,360 --> 00:14:53,520 Speaker 1: the time to remove milk grate challenge content from TikTok. 223 00:14:54,120 --> 00:14:58,560 Speaker 1: But trying to decide from the outset how dangerous is 224 00:14:58,600 --> 00:15:01,840 Speaker 1: a challenge is one the difficulties that TikTok's trust and 225 00:15:01,840 --> 00:15:05,400 Speaker 1: safety team faces when it comes to the blackout challenge. 226 00:15:05,520 --> 00:15:10,000 Speaker 1: All along, TikTok has not wanted this content on its platform. 227 00:15:10,160 --> 00:15:14,520 Speaker 1: The difficulty comes down to removing it. How do you 228 00:15:14,680 --> 00:15:17,560 Speaker 1: remove this content when you have more than a billion 229 00:15:17,680 --> 00:15:21,200 Speaker 1: users from multiple like almost every country around the world, 230 00:15:21,600 --> 00:15:24,360 Speaker 1: uploading videos on a daily basis and doing what they 231 00:15:24,400 --> 00:15:27,320 Speaker 1: can to get around the restrictions that you're placing on 232 00:15:27,360 --> 00:15:30,240 Speaker 1: the app to try and keep kids safe. You mentioned 233 00:15:30,360 --> 00:15:33,600 Speaker 1: trust in safety, and that's another term that you hear 234 00:15:33,640 --> 00:15:36,440 Speaker 1: in connection with social media platforms quite a bit. What 235 00:15:36,720 --> 00:15:42,000 Speaker 1: is a trust in safety team? Trust and safety are 236 00:15:42,080 --> 00:15:45,680 Speaker 1: effectively the cleaners of the Internet. These guys are the fixes. 237 00:15:45,960 --> 00:15:48,720 Speaker 1: When things go wrong, trust and safety comes in to 238 00:15:48,800 --> 00:15:51,320 Speaker 1: clean up the mess. So for the past year, I've 239 00:15:51,320 --> 00:15:53,240 Speaker 1: been talking to people who work within the trust and 240 00:15:53,280 --> 00:15:56,440 Speaker 1: safety industry trying to figure out how hard is the job, 241 00:15:56,480 --> 00:15:59,360 Speaker 1: What are some of the complexities or challenges that they face. 242 00:15:59,800 --> 00:16:02,040 Speaker 1: A lot of people who enter the trust and safety 243 00:16:02,080 --> 00:16:06,800 Speaker 1: industry actually come from social justice or law enforcement backgrounds. 244 00:16:07,280 --> 00:16:10,040 Speaker 1: What the ultimate goal is by joining these big tech 245 00:16:10,120 --> 00:16:13,680 Speaker 1: platforms is to try and protect users. But once they're 246 00:16:13,680 --> 00:16:16,720 Speaker 1: inside the company, and once they're actually have a seat 247 00:16:16,760 --> 00:16:20,600 Speaker 1: at the table and work there, they realize that their 248 00:16:20,680 --> 00:16:25,800 Speaker 1: job is twofold, protect the users and protect the company's reputation. 249 00:16:26,240 --> 00:16:29,240 Speaker 1: One of the most challenging things for the dozens of 250 00:16:29,280 --> 00:16:31,720 Speaker 1: trust and safety workers that I spoke with is this 251 00:16:31,840 --> 00:16:36,040 Speaker 1: feeling that you're effectively cleaning up for the company rather 252 00:16:36,080 --> 00:16:39,360 Speaker 1: than doing everything you can to protect users on the 253 00:16:39,400 --> 00:16:42,280 Speaker 1: front end, and on TikTok in particular, which is a 254 00:16:42,280 --> 00:16:46,720 Speaker 1: platform that has seen explosive growth and really high turnover. 255 00:16:47,040 --> 00:16:49,920 Speaker 1: Within the trust and safety team, that's because the job 256 00:16:50,080 --> 00:16:54,160 Speaker 1: is so damn hard, you know, trying to moderate the 257 00:16:54,240 --> 00:16:57,800 Speaker 1: platform and keep users safe. When some of the kids 258 00:16:57,920 --> 00:17:00,800 Speaker 1: are as young as six or seven years old, what 259 00:17:00,880 --> 00:17:02,800 Speaker 1: do you do when they're lying about their age? What 260 00:17:02,840 --> 00:17:04,919 Speaker 1: do you do when the appearance allow them on the 261 00:17:04,960 --> 00:17:07,639 Speaker 1: platform and allow them to lie about their age? How 262 00:17:07,680 --> 00:17:10,560 Speaker 1: do you handle that? As not only TikTok the company, 263 00:17:10,600 --> 00:17:13,200 Speaker 1: but you, the trust and safety worker who was told 264 00:17:13,240 --> 00:17:16,280 Speaker 1: to detect child accounts on the platform, how are you 265 00:17:16,280 --> 00:17:21,480 Speaker 1: going to go about doing that? This question of age 266 00:17:21,600 --> 00:17:25,560 Speaker 1: verification is now front and center as pressure increases on 267 00:17:25,680 --> 00:17:30,199 Speaker 1: social media companies to do something about underage children other platforms. 268 00:17:30,720 --> 00:17:33,159 Speaker 1: We've seen a sea change in Silicon Valley in the 269 00:17:33,240 --> 00:17:37,239 Speaker 1: last year to eighteen months where platforms are aware that 270 00:17:37,320 --> 00:17:40,720 Speaker 1: the age verification issue has kind of become the elephant 271 00:17:40,760 --> 00:17:43,680 Speaker 1: in the room. This is a massive issue not only 272 00:17:43,760 --> 00:17:47,320 Speaker 1: for TikTok, but for every social media platform is trying 273 00:17:47,320 --> 00:17:50,840 Speaker 1: to detect underage kids and keep those who use the 274 00:17:50,880 --> 00:17:55,320 Speaker 1: platform in the correct and age appropriate experience. We're seeing 275 00:17:55,560 --> 00:18:00,800 Speaker 1: regulatory shifts and legislation being written calling out TICH companies 276 00:18:00,840 --> 00:18:04,280 Speaker 1: for putting their own profits over the best interests of children, 277 00:18:04,760 --> 00:18:07,920 Speaker 1: and these new laws that are being written actually forced 278 00:18:07,920 --> 00:18:11,800 Speaker 1: TICH companies to ensure that they know the age of 279 00:18:11,840 --> 00:18:15,720 Speaker 1: their users with reasonable level of certainty, and how exactly 280 00:18:15,760 --> 00:18:18,200 Speaker 1: are they supposed to do this? Well, it turns out 281 00:18:18,240 --> 00:18:21,679 Speaker 1: there are companies that specialize in it. This third party 282 00:18:21,680 --> 00:18:24,600 Speaker 1: companies ones called Hive and they're based in San Francisco. 283 00:18:24,920 --> 00:18:27,680 Speaker 1: There's another company called Yoti that's based in the UK 284 00:18:28,240 --> 00:18:32,439 Speaker 1: and they specialize an age estimation. And we've seen different 285 00:18:32,440 --> 00:18:38,320 Speaker 1: platforms like Twitter, Be Real, Pixar, read It, Instagram partner 286 00:18:38,359 --> 00:18:41,280 Speaker 1: with these third parties to try and ensure that their 287 00:18:41,320 --> 00:18:45,440 Speaker 1: age verification efforts are as good as they possibly can be. Well, 288 00:18:45,440 --> 00:18:48,120 Speaker 1: I really wanted to know, like how effective are these 289 00:18:48,160 --> 00:18:51,960 Speaker 1: third parties? Does it actually work or you know they're 290 00:18:52,000 --> 00:18:54,639 Speaker 1: just making it up? Like how do you know that 291 00:18:54,720 --> 00:18:56,719 Speaker 1: a company that says it's going to be able to 292 00:18:56,960 --> 00:19:00,680 Speaker 1: guess a user's age can actually do it? And if so, 293 00:19:01,000 --> 00:19:04,080 Speaker 1: why isn't every platform in the world using them? The 294 00:19:04,240 --> 00:19:08,560 Speaker 1: third party company Olivia mentioned just now Hive. It's building 295 00:19:08,840 --> 00:19:15,880 Speaker 1: artificial intelligence tools to verify the age of users. How 296 00:19:15,880 --> 00:19:18,879 Speaker 1: do we build AI models that are very focused on 297 00:19:18,960 --> 00:19:24,480 Speaker 1: solving real world business problems? Problem of constant moderation, age detection, 298 00:19:25,040 --> 00:19:28,919 Speaker 1: object detection, hate speech, bulling detection. These type of aspects 299 00:19:28,960 --> 00:19:31,040 Speaker 1: that we saw that there was a lot of generally 300 00:19:31,440 --> 00:19:34,760 Speaker 1: human labor involved in and wasn't very scalable, especially when 301 00:19:34,760 --> 00:19:37,879 Speaker 1: you consider the amount of digital content being produced on 302 00:19:37,880 --> 00:19:41,960 Speaker 1: the Internet. That's Kevin who, the CEO and founder of Hive, 303 00:19:42,480 --> 00:19:44,679 Speaker 1: define what it is you're trying to detect and then 304 00:19:44,800 --> 00:19:47,600 Speaker 1: train the model on that data by actually having millions 305 00:19:47,720 --> 00:19:50,560 Speaker 1: or hundreds of millions of examples of what that means. 306 00:19:50,840 --> 00:19:53,520 Speaker 1: Imagine training a model is kind of like teaching a 307 00:19:53,560 --> 00:19:55,520 Speaker 1: small child to concepts. Right if you were to take 308 00:19:55,680 --> 00:19:59,160 Speaker 1: a different problem like recognizing you know, animals and a farm, 309 00:19:59,200 --> 00:20:01,520 Speaker 1: cats and dogs. Right, when you teach the small child, 310 00:20:01,520 --> 00:20:04,000 Speaker 1: you showed them examples in picture book. Here's what a 311 00:20:04,000 --> 00:20:05,880 Speaker 1: cat looks like, Here's what a dog looks like, Here's 312 00:20:05,880 --> 00:20:08,400 Speaker 1: what a pig looks like. And generally the child will 313 00:20:08,480 --> 00:20:10,960 Speaker 1: learn from that from those examples. In our world, how 314 00:20:11,000 --> 00:20:13,640 Speaker 1: we view this problem is you take a person, it's 315 00:20:13,680 --> 00:20:16,120 Speaker 1: on camera, it's in a video, and we do run 316 00:20:16,119 --> 00:20:18,920 Speaker 1: a model that looks at a few things. It looks 317 00:20:18,920 --> 00:20:20,960 Speaker 1: at their face, but it look also looks at the 318 00:20:21,000 --> 00:20:23,800 Speaker 1: rest of their body, and it makes the prediction of 319 00:20:23,880 --> 00:20:26,520 Speaker 1: age based on those features. So it's using the input 320 00:20:26,560 --> 00:20:28,960 Speaker 1: of the person's facial features and their body and in 321 00:20:29,000 --> 00:20:30,560 Speaker 1: general or anything else that we can kind of say 322 00:20:30,560 --> 00:20:33,240 Speaker 1: is it's a kind of related to the physical makeup 323 00:20:33,280 --> 00:20:35,320 Speaker 1: of that person that produces a results on age, kind 324 00:20:35,320 --> 00:20:37,119 Speaker 1: of saying, when you look at someone, you're not totally 325 00:20:37,119 --> 00:20:39,400 Speaker 1: sure exactly why you know that person seems like they're 326 00:20:39,600 --> 00:20:41,800 Speaker 1: nine years old, but your instinct is that they are 327 00:20:41,840 --> 00:20:44,600 Speaker 1: because of enough pattern recognition. That's when we trained this 328 00:20:44,640 --> 00:20:47,639 Speaker 1: model to do as well. We're actually predicting the age 329 00:20:47,960 --> 00:20:52,879 Speaker 1: without knowing identity whatsoever. One of the things that I 330 00:20:52,960 --> 00:20:56,520 Speaker 1: found really surprising during the reporting of this story is 331 00:20:56,880 --> 00:21:00,520 Speaker 1: as I was looking into Ariani A Royo's k I 332 00:21:00,560 --> 00:21:04,159 Speaker 1: actually saw a screenshot of her TikTok account. She was 333 00:21:04,200 --> 00:21:06,360 Speaker 1: only nine years old, so she wasn't meant to even 334 00:21:06,359 --> 00:21:08,880 Speaker 1: be on the platform. But when you look at her account, 335 00:21:08,920 --> 00:21:11,359 Speaker 1: you can see she's a kid. You can see in 336 00:21:11,400 --> 00:21:14,000 Speaker 1: her bio photo that it's her nine year old face. 337 00:21:14,359 --> 00:21:17,200 Speaker 1: She wasn't trying to hide herself or trying to lie 338 00:21:17,240 --> 00:21:19,880 Speaker 1: to the platform. Sure, she entered in the wrong date 339 00:21:19,920 --> 00:21:22,159 Speaker 1: of birth to access the app, but so did all 340 00:21:22,200 --> 00:21:26,159 Speaker 1: her friends. Olivia decided to put HIDES technology to the test, 341 00:21:26,840 --> 00:21:29,720 Speaker 1: so she sent them a short video clip ARIANI had 342 00:21:29,760 --> 00:21:32,400 Speaker 1: filmed of herself for social media. And the way our 343 00:21:32,520 --> 00:21:35,679 Speaker 1: models work on videos, as we look at frames and 344 00:21:35,720 --> 00:21:37,800 Speaker 1: we samples, we usually do a one frame per seconds. 345 00:21:37,840 --> 00:21:39,560 Speaker 1: In this case, I believe the video was something like 346 00:21:39,680 --> 00:21:43,520 Speaker 1: eleven seconds, So we produced eleven seconds of results, and 347 00:21:43,560 --> 00:21:46,120 Speaker 1: then we simply take the average of that to see 348 00:21:46,119 --> 00:21:48,639 Speaker 1: what our model would predict, and we've predicted that her 349 00:21:48,680 --> 00:21:52,760 Speaker 1: age was around ten years old. It took HIVES software 350 00:21:53,000 --> 00:21:56,520 Speaker 1: three seconds to determine that Ariane was ten years old, 351 00:21:57,000 --> 00:21:59,680 Speaker 1: and she died three months shy of her tenth birthday. 352 00:22:00,280 --> 00:22:04,280 Speaker 1: Kevin Guo says that level of accuracy is fairly standard 353 00:22:04,320 --> 00:22:07,800 Speaker 1: for age prediction software. There's really no reason, in our opinion, 354 00:22:07,920 --> 00:22:09,760 Speaker 1: the humans should be looking at any of this. Our 355 00:22:09,800 --> 00:22:12,920 Speaker 1: models are going to handle this at necent accuracy, really 356 00:22:12,960 --> 00:22:15,520 Speaker 1: at near Yeah. One of the stats we like to 357 00:22:15,520 --> 00:22:18,760 Speaker 1: give is our you know, we'll catch content of bad 358 00:22:18,800 --> 00:22:21,520 Speaker 1: content while making a mistake unless than one percent of it. 359 00:22:21,720 --> 00:22:23,080 Speaker 1: The way that we've thought about it with all of 360 00:22:23,160 --> 00:22:25,960 Speaker 1: our partners and platforms we work with today is we've 361 00:22:25,960 --> 00:22:28,760 Speaker 1: never yet encountered a problem where their scale was too 362 00:22:28,800 --> 00:22:30,600 Speaker 1: much for us to handle. And this includes some of 363 00:22:30,640 --> 00:22:33,320 Speaker 1: the largest platforms in the world. So I'm very confident 364 00:22:33,400 --> 00:22:36,679 Speaker 1: that if a platform we're actually aligned and trying to 365 00:22:36,680 --> 00:22:40,320 Speaker 1: solve this problem, the hardware and the costs can't keep up. Accordingly, 366 00:22:40,560 --> 00:22:43,239 Speaker 1: this is intended to be a real time product. In 367 00:22:43,240 --> 00:22:45,840 Speaker 1: this case, you could imagine a world where for users 368 00:22:45,840 --> 00:22:47,639 Speaker 1: that are underage, given the sense of the nature of that, 369 00:22:47,720 --> 00:22:49,919 Speaker 1: you would still want to escalate hume review to like 370 00:22:49,920 --> 00:22:52,119 Speaker 1: take final action, but at least the flag that this 371 00:22:52,160 --> 00:22:55,119 Speaker 1: content is something worth investigating. We believe that should be 372 00:22:55,400 --> 00:22:58,360 Speaker 1: status quo for every platform out there. This technology exists. 373 00:22:58,760 --> 00:23:05,360 Speaker 1: The technology exists, but is TikTok using it. I'm aware 374 00:23:05,480 --> 00:23:09,520 Speaker 1: that TikTok actually met with both Hive and Yote throughout 375 00:23:09,520 --> 00:23:12,520 Speaker 1: two thousand and twenty one to try and understand how 376 00:23:12,520 --> 00:23:15,040 Speaker 1: their technology works and if there could be a potential 377 00:23:15,080 --> 00:23:19,560 Speaker 1: partnership there. TikTok ultimately decided not to deploy this technology 378 00:23:19,560 --> 00:23:22,120 Speaker 1: and not to partner with those companies. So I went 379 00:23:22,160 --> 00:23:25,600 Speaker 1: to TikTok's communications team and asked them specifically, why did 380 00:23:25,600 --> 00:23:28,919 Speaker 1: the company decide not to partner with these third party 381 00:23:29,280 --> 00:23:33,640 Speaker 1: age estimation software providers. I knew that TikTok had met 382 00:23:33,720 --> 00:23:37,359 Speaker 1: with representatives from both Hive and Yote throughout two thousand 383 00:23:37,359 --> 00:23:39,679 Speaker 1: and twenty one, and that the company had decided not 384 00:23:39,760 --> 00:23:42,880 Speaker 1: to deploy this technology, but I didn't know why they 385 00:23:42,880 --> 00:23:46,080 Speaker 1: declined to comment on that particular question. I did talk 386 00:23:46,119 --> 00:23:48,280 Speaker 1: to a number of people who worked inside the trust 387 00:23:48,280 --> 00:23:51,200 Speaker 1: and safety space at TikTok to try and figure it out, 388 00:23:51,600 --> 00:23:54,040 Speaker 1: and I was told that the company was afraid of 389 00:23:54,119 --> 00:23:58,240 Speaker 1: the public relations backlash or the reputational risk of collecting 390 00:23:58,480 --> 00:24:02,720 Speaker 1: biometric data of children when there were so many suspicions 391 00:24:02,720 --> 00:24:06,879 Speaker 1: and fears around TikTok and its parent company, Bite Dance, 392 00:24:06,920 --> 00:24:11,200 Speaker 1: which is based in Beijing, spying on children or taking 393 00:24:11,280 --> 00:24:14,480 Speaker 1: data from places like the US and the UK and 394 00:24:14,600 --> 00:24:17,359 Speaker 1: giving it to the Chinese government. So the company was 395 00:24:17,400 --> 00:24:22,720 Speaker 1: afraid that there'd be an additional pr backlash around suspicions 396 00:24:22,760 --> 00:24:27,840 Speaker 1: of China spying on child users around the world. It's 397 00:24:27,840 --> 00:24:30,040 Speaker 1: important to take a moment here to spell out the 398 00:24:30,080 --> 00:24:34,040 Speaker 1: difference between the age prediction software that Hive makes and 399 00:24:34,160 --> 00:24:37,639 Speaker 1: facial recognition software, because it sounds like they do the 400 00:24:37,680 --> 00:24:41,760 Speaker 1: same thing. Facial recognition software tries to figure out the 401 00:24:41,840 --> 00:24:44,680 Speaker 1: identity of a person. That's why it raises all kinds 402 00:24:44,680 --> 00:24:49,840 Speaker 1: of privacy concerns. Age prediction software only tries to determine 403 00:24:49,880 --> 00:24:53,399 Speaker 1: how old and an anonymous person is. It doesn't collect 404 00:24:53,400 --> 00:24:56,359 Speaker 1: any data or try to find out who that person is. 405 00:24:56,800 --> 00:24:59,040 Speaker 1: It turns out they are related, but they're by no 406 00:24:59,160 --> 00:25:00,960 Speaker 1: means at all the same. And of course in the 407 00:25:01,040 --> 00:25:03,159 Speaker 1: database that we even trained on, there's no identities. We 408 00:25:03,200 --> 00:25:05,560 Speaker 1: don't know their names. Those pictures could have been taking 409 00:25:05,840 --> 00:25:08,040 Speaker 1: who knows when, many years ago, you know. So it's 410 00:25:08,080 --> 00:25:11,480 Speaker 1: nothing to do with real time identity tracking. Every new 411 00:25:11,600 --> 00:25:13,440 Speaker 1: image or video that comes to us. It's like we're 412 00:25:13,440 --> 00:25:16,160 Speaker 1: starting over from scratch and we're just making a guess 413 00:25:16,200 --> 00:25:19,040 Speaker 1: as to how old we think that person is. So 414 00:25:19,240 --> 00:25:24,560 Speaker 1: content moderation isn't effective. Age prediction technology isn't always being used, 415 00:25:25,000 --> 00:25:27,639 Speaker 1: and that means kids younger than thirteen are coming into 416 00:25:27,680 --> 00:25:33,000 Speaker 1: contact with all kinds of risky challenges, including the blackout challenge. Now, 417 00:25:33,160 --> 00:25:35,480 Speaker 1: most adults might see it is common sense not to 418 00:25:35,560 --> 00:25:38,240 Speaker 1: copy a video where someone is choking or hitting his 419 00:25:38,280 --> 00:25:41,439 Speaker 1: head on pavement or lighting things on fire, but the 420 00:25:41,480 --> 00:25:45,639 Speaker 1: brain of a child processes those videos differently and referred 421 00:25:45,640 --> 00:25:47,959 Speaker 1: to it as a three part recipe. UM in the 422 00:25:48,000 --> 00:25:52,440 Speaker 1: digital world. That's Dr Adam Pletter. He's a child psychologist 423 00:25:52,480 --> 00:25:56,440 Speaker 1: whose work focuses on how kids use social media. If 424 00:25:56,440 --> 00:25:59,359 Speaker 1: we talk about children or adolescents, what we're talking about 425 00:25:59,560 --> 00:26:02,720 Speaker 1: is the prefrontal cortex, the front part of your brain. 426 00:26:02,960 --> 00:26:06,720 Speaker 1: That's the control center. It controls emotions, it helps all 427 00:26:06,760 --> 00:26:11,840 Speaker 1: those executive functions that people often talk about prioritizing, sequencing, 428 00:26:12,680 --> 00:26:14,520 Speaker 1: and so that part of the brain, the front part 429 00:26:14,520 --> 00:26:19,080 Speaker 1: of your brain, develops all through childhood into adolescents. So 430 00:26:19,119 --> 00:26:22,640 Speaker 1: that's the braking system of the brain, the emotional part 431 00:26:22,640 --> 00:26:24,840 Speaker 1: of the brain. UM more of the middle part, more 432 00:26:24,840 --> 00:26:28,959 Speaker 1: primitive structures in the brain, are you know, really overactive 433 00:26:29,000 --> 00:26:31,280 Speaker 1: in childhood, which again if you think about you know, 434 00:26:31,280 --> 00:26:33,240 Speaker 1: if you go out to a playground or you know, 435 00:26:33,240 --> 00:26:35,720 Speaker 1: see a bunch of children playing, it's a lot more active. 436 00:26:36,160 --> 00:26:39,640 Speaker 1: And then the third part of the recipe is the now, 437 00:26:39,720 --> 00:26:44,439 Speaker 1: the digital world and the amazing emotional often content on 438 00:26:44,560 --> 00:26:47,840 Speaker 1: demand in often in their pockets, you know, two three 439 00:26:47,880 --> 00:26:51,280 Speaker 1: clicks away at their fingertips literally, So if you put 440 00:26:51,320 --> 00:26:55,280 Speaker 1: all those three together, a weak breaking system, an overactive 441 00:26:55,280 --> 00:26:57,600 Speaker 1: emotional part of your brain that wants, you know, that's 442 00:26:57,680 --> 00:27:00,159 Speaker 1: encouraging you to go out and seek out information, and 443 00:27:00,240 --> 00:27:04,240 Speaker 1: then the digital world that is so focused and targeted 444 00:27:04,400 --> 00:27:08,600 Speaker 1: on pulling for our emotions, it really creates, you know, 445 00:27:08,680 --> 00:27:11,200 Speaker 1: a real struggle. To say the least. We'll hear more 446 00:27:11,240 --> 00:27:14,399 Speaker 1: from Dr Pletter a bit later. The struggle he's talking 447 00:27:14,440 --> 00:27:17,719 Speaker 1: about is one that's playing out across the globe. In 448 00:27:17,800 --> 00:27:21,960 Speaker 1: August of this year, two children in the UK died 449 00:27:22,119 --> 00:27:26,720 Speaker 1: allegedly participating in the Blackout challenge. Ariani's family, along with 450 00:27:26,800 --> 00:27:31,080 Speaker 1: two others, have launched a lawsuit against TikTok, claiming wrongful death. 451 00:27:31,680 --> 00:27:34,640 Speaker 1: The case is led by the Social Media Victims Law 452 00:27:34,720 --> 00:27:38,240 Speaker 1: Center based out of Seattle. I went out to Seattle 453 00:27:38,280 --> 00:27:41,359 Speaker 1: and meet with Matthew Bergman. He founded the Social Media 454 00:27:41,440 --> 00:27:44,159 Speaker 1: Victim's Law CINA and is actually the lead attorney on 455 00:27:44,240 --> 00:27:47,399 Speaker 1: the Ariani a Royo case. It's clear that this is 456 00:27:47,400 --> 00:27:50,240 Speaker 1: an epidemic, and it's equally clear that this is not 457 00:27:50,320 --> 00:27:53,600 Speaker 1: a coincidence, and it's not an accident, but rather a 458 00:27:53,600 --> 00:27:57,080 Speaker 1: clear result from the design decisions that TikTok made when 459 00:27:57,080 --> 00:27:59,600 Speaker 1: they set up their algorithms. And that's why we're bringing 460 00:27:59,640 --> 00:28:02,400 Speaker 1: these case. Is I asked Olivia what tech talk says 461 00:28:02,440 --> 00:28:05,320 Speaker 1: about the lawsuit and about the problem of children and 462 00:28:05,400 --> 00:28:10,000 Speaker 1: its platform. TikTok says that it cannot comment on pending litigation, 463 00:28:10,200 --> 00:28:13,200 Speaker 1: which is what most companies say when they've head lawsuits 464 00:28:13,200 --> 00:28:16,560 Speaker 1: filed against them. It does say that the blackout challenge 465 00:28:17,000 --> 00:28:20,440 Speaker 1: predates the platform, that it's been around for decades before 466 00:28:20,440 --> 00:28:23,920 Speaker 1: TikTok was even created, and that's all true. It also 467 00:28:24,000 --> 00:28:27,240 Speaker 1: says that the blackout challenge has never been a trend 468 00:28:27,280 --> 00:28:29,680 Speaker 1: on the platform, that they've never found evidence of the 469 00:28:29,680 --> 00:28:33,520 Speaker 1: blackout challenge actually trending will going viral on TikTok. The 470 00:28:33,600 --> 00:28:36,880 Speaker 1: company says that it does everything it can to remove 471 00:28:36,960 --> 00:28:40,760 Speaker 1: content linked to the blackout challenge, that it's against its 472 00:28:40,800 --> 00:28:45,920 Speaker 1: community guidelines to post harmful content that glorifies or promotes 473 00:28:46,080 --> 00:28:50,560 Speaker 1: dangerous behavior like the blackout challenge. They're not seeing how 474 00:28:50,600 --> 00:28:53,440 Speaker 1: there are algorithms lead to these challenges and not seeing 475 00:28:53,480 --> 00:28:56,920 Speaker 1: how the addictive nature of their product lead to these outcomes, 476 00:28:57,320 --> 00:28:59,680 Speaker 1: and this is why we need them to be held 477 00:28:59,680 --> 00:29:02,040 Speaker 1: account of all in a court of law. The lawyers 478 00:29:02,120 --> 00:29:05,320 Speaker 1: from the Social Media Victims Law Center want to hold 479 00:29:05,400 --> 00:29:08,880 Speaker 1: these companies liable when things go wrong, and that was 480 00:29:08,920 --> 00:29:12,240 Speaker 1: one of the challenges of reporting this is how do 481 00:29:12,400 --> 00:29:17,800 Speaker 1: you prove that the child died because they saw a 482 00:29:17,880 --> 00:29:22,520 Speaker 1: video on TikTok. To prove which social media platform sent 483 00:29:22,640 --> 00:29:26,640 Speaker 1: that content to that child is incredibly difficult. You need 484 00:29:26,680 --> 00:29:29,360 Speaker 1: to do a forensic analysis of their cell phone. You 485 00:29:29,440 --> 00:29:32,080 Speaker 1: need to find when that specific piece of content was 486 00:29:32,120 --> 00:29:34,440 Speaker 1: sent through to the child, where did they watch it, 487 00:29:34,480 --> 00:29:36,320 Speaker 1: when did they watch it, and how do you know 488 00:29:36,800 --> 00:29:39,240 Speaker 1: that when they died. What they were trying to do 489 00:29:39,400 --> 00:29:43,560 Speaker 1: was emulate that challenge. Even when the evidence may seem 490 00:29:43,640 --> 00:29:46,360 Speaker 1: cut and dried, the law may not be on the 491 00:29:46,440 --> 00:29:49,120 Speaker 1: side of parents. There was a case of a ten 492 00:29:49,160 --> 00:29:53,520 Speaker 1: year old girl in Pennsylvania who died allegedly participating in 493 00:29:53,560 --> 00:29:58,200 Speaker 1: the Blackout challenge. She accidentally hanged herself with the purse 494 00:29:58,240 --> 00:30:02,520 Speaker 1: strap in her mother's ward robe. In that case, her 495 00:30:02,600 --> 00:30:05,640 Speaker 1: lawyers say that police did a forensic analysis of her 496 00:30:05,640 --> 00:30:09,520 Speaker 1: cell phone and that there is evidence that TikTok's algorithm 497 00:30:09,720 --> 00:30:12,960 Speaker 1: sent her a video showing her how to hang herself 498 00:30:13,440 --> 00:30:16,920 Speaker 1: with the purse strap of a handbag. TikTok filed a 499 00:30:17,000 --> 00:30:20,200 Speaker 1: motion to dismiss this case, arguing that it couldn't be 500 00:30:20,240 --> 00:30:24,440 Speaker 1: held liable for her death because under the Communications Decency Act, 501 00:30:24,760 --> 00:30:28,440 Speaker 1: platforms cannot be held liable for the content that is 502 00:30:28,480 --> 00:30:32,320 Speaker 1: posted on their sites by third parties. It is an immunity. 503 00:30:32,480 --> 00:30:35,320 Speaker 1: It is a shield for technology platforms from being held 504 00:30:35,360 --> 00:30:38,920 Speaker 1: responsible when things like this happen. A judge in that 505 00:30:39,040 --> 00:30:43,560 Speaker 1: case actually dismissed the lawsuit, saying even if TikTok had 506 00:30:43,680 --> 00:30:47,040 Speaker 1: sent the blackout challenge content, it could not be sued 507 00:30:47,280 --> 00:30:51,280 Speaker 1: because Section two thirty of the Communications Decency Act shielded 508 00:30:51,320 --> 00:30:55,400 Speaker 1: it from being held liable. This problem exists across social 509 00:30:55,400 --> 00:30:59,240 Speaker 1: media platforms. We've seen challenges like the blackout Challenge on 510 00:30:59,640 --> 00:31:03,120 Speaker 1: you for example. I've also seen reports that it's been 511 00:31:03,200 --> 00:31:08,800 Speaker 1: on Facebook, Instagram, Snapchat. What makes TikTok different, though, is 512 00:31:08,840 --> 00:31:12,920 Speaker 1: that this is a highly performative app, where in order 513 00:31:13,000 --> 00:31:16,640 Speaker 1: to gain followers and build your presence on the app, 514 00:31:16,680 --> 00:31:21,960 Speaker 1: you're encouraged to do extreme things. The algorithm really rewards 515 00:31:22,000 --> 00:31:27,480 Speaker 1: extreme behavior. Thanks to Olivia Carville for sharing her reporting. 516 00:31:28,320 --> 00:31:32,120 Speaker 1: You can read Olivia Carville's Business Week story about TikTok 517 00:31:32,480 --> 00:31:36,760 Speaker 1: on Bloomberg dot com. When we come back, what parents 518 00:31:36,840 --> 00:31:53,240 Speaker 1: can do to protect their kids? Obviously this is pretty 519 00:31:53,360 --> 00:31:56,840 Speaker 1: difficult to listen to. And if you're a parent of 520 00:31:56,920 --> 00:31:59,960 Speaker 1: young kids or even a teenager, you might be one 521 00:32:00,040 --> 00:32:02,880 Speaker 1: during how you're supposed to protect them from seeing or 522 00:32:02,920 --> 00:32:07,280 Speaker 1: possibly copying challenges like these. Doctor Platter, who you heard 523 00:32:07,320 --> 00:32:10,840 Speaker 1: a little earlier, specializes in helping parents talk to kids 524 00:32:10,920 --> 00:32:14,960 Speaker 1: about social media, so I asked him what advice he 525 00:32:15,000 --> 00:32:18,920 Speaker 1: would give, so finding ways of connecting around the concerns 526 00:32:19,000 --> 00:32:22,960 Speaker 1: without overdoing it, helping the children think things through, having 527 00:32:22,960 --> 00:32:25,200 Speaker 1: a plan of how they're going to handle certain things. 528 00:32:25,720 --> 00:32:29,160 Speaker 1: So much of the concerns on TikTok and and other 529 00:32:29,240 --> 00:32:32,800 Speaker 1: places are predictable, so I try to come at it 530 00:32:32,920 --> 00:32:37,000 Speaker 1: less reactive. You want to start off with understanding that 531 00:32:37,320 --> 00:32:41,680 Speaker 1: these concerns are there, and having some plan, whether it's 532 00:32:41,760 --> 00:32:45,360 Speaker 1: written down media plan, where here are some rules. Having 533 00:32:45,600 --> 00:32:48,240 Speaker 1: basic parental controls. I usually lean on the ones that 534 00:32:48,280 --> 00:32:51,480 Speaker 1: are built into the phones. Um not that any of 535 00:32:51,480 --> 00:32:54,040 Speaker 1: the parental control systems are perfect, and you can't just 536 00:32:54,160 --> 00:32:56,920 Speaker 1: sort of set it and forget it, but it serves 537 00:32:57,240 --> 00:32:59,840 Speaker 1: once it's set up, it does serve a good role. 538 00:33:00,240 --> 00:33:03,760 Speaker 1: The prince controls that is in encouraging I'll say, if 539 00:33:03,800 --> 00:33:07,320 Speaker 1: not forcing the dialogue, family sharing. That's something that you 540 00:33:07,360 --> 00:33:09,479 Speaker 1: want to do when you're first setting up the phone, 541 00:33:09,640 --> 00:33:12,720 Speaker 1: but you can do it at any point where each phone, 542 00:33:12,760 --> 00:33:16,480 Speaker 1: the young child's phone, even a young young kid has 543 00:33:16,560 --> 00:33:18,440 Speaker 1: their own Apple I D And this way it's all 544 00:33:18,520 --> 00:33:21,760 Speaker 1: under the same umbrella. If you go under settings, you 545 00:33:21,840 --> 00:33:23,760 Speaker 1: just click on your name and you basically it says 546 00:33:23,800 --> 00:33:26,360 Speaker 1: like set up your family sharing and so that's all 547 00:33:26,400 --> 00:33:28,800 Speaker 1: through the iPhone set up. And then once you have 548 00:33:28,880 --> 00:33:32,560 Speaker 1: that set up, you set up screen time. Uh, if 549 00:33:32,560 --> 00:33:36,360 Speaker 1: you want WiFi controls. Most of the larger internet connections, 550 00:33:36,440 --> 00:33:38,640 Speaker 1: they have their own apps now, whether it's ex Affinity, 551 00:33:38,800 --> 00:33:41,880 Speaker 1: Verizon or whatever the company is that you use for 552 00:33:41,960 --> 00:33:46,200 Speaker 1: your at home WiFi. And I also recommend that setting 553 00:33:46,240 --> 00:33:50,160 Speaker 1: that up separately because then you could turn other devices 554 00:33:50,240 --> 00:33:54,440 Speaker 1: chromebooks off, Xboxes during the school day, during study hours, 555 00:33:54,520 --> 00:33:57,360 Speaker 1: you can easily just toggle things on and off, which 556 00:33:57,440 --> 00:34:00,560 Speaker 1: again forces the dialogue. If your child wants to play Xbox, 557 00:34:00,600 --> 00:34:02,720 Speaker 1: they can come say hey, can I play Xbox? And 558 00:34:02,760 --> 00:34:04,280 Speaker 1: you say, sure, what are you playing? Who are you 559 00:34:04,280 --> 00:34:06,320 Speaker 1: playing with? How long do you think you're gonna play? 560 00:34:06,360 --> 00:34:08,319 Speaker 1: There a little bit of that dialogue back and forth, 561 00:34:08,360 --> 00:34:10,400 Speaker 1: and then you take out your phone. It's all remote 562 00:34:10,680 --> 00:34:12,799 Speaker 1: and you turn it on and then when they're done, 563 00:34:12,840 --> 00:34:15,840 Speaker 1: you turn it off. Do the things you're describing here, 564 00:34:16,680 --> 00:34:21,200 Speaker 1: uh work practically. UM, have you seen that this is 565 00:34:21,239 --> 00:34:24,920 Speaker 1: an effective way for parents to keep their kids safe 566 00:34:25,440 --> 00:34:28,400 Speaker 1: while not turning their kids away from them? So I 567 00:34:28,440 --> 00:34:30,440 Speaker 1: want to say yes because this is what I do. 568 00:34:30,600 --> 00:34:35,840 Speaker 1: You know, typically parents, UM, providing supervision and providing a 569 00:34:35,880 --> 00:34:40,880 Speaker 1: regulatory system, one that is uh that progresses and evolves 570 00:34:40,880 --> 00:34:45,640 Speaker 1: over time works. So it's not just that parental reaction 571 00:34:45,719 --> 00:34:48,520 Speaker 1: to rain things in protect your kid by simply turning 572 00:34:48,520 --> 00:34:50,879 Speaker 1: it off saying no, you're not allowed to have it, 573 00:34:50,920 --> 00:34:53,920 Speaker 1: but doing it in conjunction with talking to the kid 574 00:34:53,920 --> 00:34:56,799 Speaker 1: about why, so that they are kind of part of 575 00:34:56,800 --> 00:34:59,920 Speaker 1: a conversation, you know, the counter of what you just suggest. 576 00:35:00,000 --> 00:35:03,280 Speaker 1: It is not possible, especially coming out of the pandemic. 577 00:35:03,320 --> 00:35:04,960 Speaker 1: But I would have said this if we spoke about 578 00:35:04,960 --> 00:35:09,400 Speaker 1: this five years ago, but especially now, screens are everywhere. 579 00:35:09,440 --> 00:35:12,759 Speaker 1: It's for education, it's for work. You know, every aspect 580 00:35:12,960 --> 00:35:16,279 Speaker 1: of that young child's life is connected in some way 581 00:35:16,320 --> 00:35:21,560 Speaker 1: to some digital format. Some of the kids who are 582 00:35:21,719 --> 00:35:24,359 Speaker 1: engaging in the Blackout Challenge and other things like that 583 00:35:24,760 --> 00:35:27,719 Speaker 1: are much younger. They're ten years old. So they're not 584 00:35:27,840 --> 00:35:31,840 Speaker 1: even teenagers yet. How do you talk to those kids? 585 00:35:31,840 --> 00:35:36,239 Speaker 1: How do you protect those kids who don't even have 586 00:35:36,360 --> 00:35:39,279 Speaker 1: the sophistication of a young teen when it comes to 587 00:35:39,400 --> 00:35:42,600 Speaker 1: understanding what's dangerous and what's not. So I'd be talking 588 00:35:42,680 --> 00:35:45,160 Speaker 1: to the parents more than the kids when you get 589 00:35:45,200 --> 00:35:50,480 Speaker 1: below ten, and then the parents um having to up 590 00:35:50,520 --> 00:35:53,920 Speaker 1: that dialogue with their younger kids. But that is the 591 00:35:53,960 --> 00:35:57,160 Speaker 1: path for the parent to be talking and knowing that 592 00:35:57,239 --> 00:35:59,719 Speaker 1: all of these dangerous things are out there, not to 593 00:35:59,800 --> 00:36:02,880 Speaker 1: just be freaked out, but to use that as motivation, 594 00:36:03,040 --> 00:36:08,239 Speaker 1: protective anxiety, parental anxiety, to talk and to make sure 595 00:36:08,320 --> 00:36:11,239 Speaker 1: that some of these safeguards are in place where you know, 596 00:36:11,320 --> 00:36:13,600 Speaker 1: if you have a nine year old, you know, and 597 00:36:13,600 --> 00:36:16,839 Speaker 1: they're on TikTok, there has to be an awareness from 598 00:36:16,840 --> 00:36:19,719 Speaker 1: my point of view, that that nine year old might 599 00:36:19,800 --> 00:36:23,560 Speaker 1: come in contact with not only strangers reaching out, but 600 00:36:23,719 --> 00:36:27,920 Speaker 1: also some of these challenges that everyone else is doing, 601 00:36:28,000 --> 00:36:30,840 Speaker 1: so it can pull them in in very, very scary, 602 00:36:30,960 --> 00:36:35,920 Speaker 1: dangerous ways. Social media is changing so quickly. There's always 603 00:36:35,920 --> 00:36:38,279 Speaker 1: new sights, and the ones that were all familiar with 604 00:36:38,320 --> 00:36:41,600 Speaker 1: suddenly are kind of less attractive to kids, and they're 605 00:36:41,640 --> 00:36:45,800 Speaker 1: going to something else. That's always been the case with everything. Um, 606 00:36:46,080 --> 00:36:48,759 Speaker 1: how should parents keep up? Because they're often the last 607 00:36:48,800 --> 00:36:50,120 Speaker 1: to know, you know, by the time all the kids 608 00:36:50,160 --> 00:36:52,200 Speaker 1: are on TikTok, they're just kind of first hearing about it. 609 00:36:52,239 --> 00:36:54,480 Speaker 1: And that's a little bit by design, because kids don't 610 00:36:54,480 --> 00:36:56,680 Speaker 1: want to be on the same platforms as their parents 611 00:36:56,719 --> 00:36:58,920 Speaker 1: a lot of the time. How do you even keep up? 612 00:36:58,920 --> 00:37:01,920 Speaker 1: As someone who makes it a practice to keep up? 613 00:37:02,239 --> 00:37:04,960 Speaker 1: So part of my secret is that I work and 614 00:37:05,080 --> 00:37:07,480 Speaker 1: talk to kids and teenagers all day long, so I'm 615 00:37:07,520 --> 00:37:10,360 Speaker 1: hearing that firsthand. UM, I hear from my own kids. 616 00:37:10,360 --> 00:37:13,080 Speaker 1: I've got two teenagers at home, so I try my best. 617 00:37:13,719 --> 00:37:16,799 Speaker 1: There's some great websites out there that do some of 618 00:37:16,800 --> 00:37:19,719 Speaker 1: the work for you. Common sense media comes to mind 619 00:37:19,960 --> 00:37:22,320 Speaker 1: is probably my go to, UM if I have a 620 00:37:22,400 --> 00:37:24,600 Speaker 1: question or just to kind of see what's on their 621 00:37:24,640 --> 00:37:28,200 Speaker 1: front page. And then you know, for ironically, if a 622 00:37:28,280 --> 00:37:31,000 Speaker 1: parent is on you know, one of the social media apps, 623 00:37:31,000 --> 00:37:35,000 Speaker 1: whether it's Instagram or Facebook, there's tons and tons of 624 00:37:35,160 --> 00:37:38,880 Speaker 1: Facebook groups you know, focused on parental controls or often 625 00:37:38,960 --> 00:37:42,600 Speaker 1: run by a certain parental control software. But you know, 626 00:37:42,640 --> 00:37:45,160 Speaker 1: it's not just for that software. UM, And there's a 627 00:37:45,200 --> 00:37:48,120 Speaker 1: lot of forums out there where people are talking, and 628 00:37:48,200 --> 00:37:50,720 Speaker 1: sometimes you can get pretty anxious reading some of the stuff. 629 00:37:50,719 --> 00:37:52,239 Speaker 1: But as long as you go in with the right 630 00:37:52,320 --> 00:37:55,359 Speaker 1: mindset that you're looking to educate yourself and not just 631 00:37:55,440 --> 00:38:00,320 Speaker 1: be freaked out, it should serve that purpose. Dr Adam Platter, 632 00:38:00,600 --> 00:38:06,240 Speaker 1: thanks for speaking with me. Thank you, thanks for listening 633 00:38:06,280 --> 00:38:09,040 Speaker 1: to us here at The Big Take, the daily podcast 634 00:38:09,120 --> 00:38:12,920 Speaker 1: from Bloomberg and I Heart Radio. For more shows from 635 00:38:12,920 --> 00:38:17,200 Speaker 1: my Heart Radio, visit the i Heart Radio app, Couple Podcast, 636 00:38:17,320 --> 00:38:21,160 Speaker 1: or wherever you listen. Read today's story and subscribe to 637 00:38:21,200 --> 00:38:25,279 Speaker 1: our daily newsletter at Bloomberg dot com slash Big Take, 638 00:38:25,840 --> 00:38:28,640 Speaker 1: and we'd love to hear from you. Email us with 639 00:38:28,800 --> 00:38:32,759 Speaker 1: questions or comments to Big Take at Bloomberg dot net. 640 00:38:35,080 --> 00:38:38,040 Speaker 1: The supervising producer of The Big Take is Vicky Burgalina. 641 00:38:38,560 --> 00:38:42,840 Speaker 1: Our senior producer is Katherine Fink. Our producer is Rebecca Chasson. 642 00:38:43,440 --> 00:38:47,279 Speaker 1: Our associate producer is sam Goa Bauer. Hilda Garcia is 643 00:38:47,320 --> 00:38:53,000 Speaker 1: our engineer. Original music by Leo Sidrin. I'm West Kasova. 644 00:38:53,239 --> 00:39:01,120 Speaker 1: We'll be back tomorrow with another Big Take on four