YouTube2Text

Frances Haugen - Facebook’s Algorithm & Teen Mental Health | The Daily Show — Transcript

by The Daily Show · 2,502 words · 374 segments · language en · Watch on YouTube

Full transcript

  1. 0:00francis haugen welcome to the daily show
  2. 0:02thank you for inviting me happy to be
  3. 0:03here um
  4. 0:04your name
  5. 0:06isn't as big as the story that you
  6. 0:10really i mean leaked to the world
  7. 0:12you know a lot of people you say francis
  8. 0:13how can they be like who is that but
  9. 0:15then if you say to people
  10. 0:16hey remember how we all learned that
  11. 0:19facebook and what social media is doing
  12. 0:21is essentially destroying teenagers
  13. 0:22brains and it's harming all of us yeah
  14. 0:24that report happened because of you so
  15. 0:27let's start with the question of why
  16. 0:29you've worked for
  17. 0:31google you've worked for pinterest
  18. 0:32you've worked for yelp and yet you blew
  19. 0:34the whistle on facebook why did you feel
  20. 0:36like oh this this is something i can't
  21. 0:38sit on
  22. 0:39so when i joined
  23. 0:41facebook i thought i was going to work
  24. 0:42on misinformation in the united states
  25. 0:45and i was surprised i actually was
  26. 0:47charged with working on information
  27. 0:49outside only outside the united states
  28. 0:51okay and
  29. 0:52like many people and many technologists
  30. 0:54i had never really focused on facebook's
  31. 0:57impact internationally
  32. 0:58and very rapidly
  33. 1:02i realized kind of the horrifying
  34. 1:04magnitude of the danger that we are
  35. 1:05facing that facebook's algorithms give
  36. 1:08the most reach to the most extreme and
  37. 1:10divisive ideas um and that that process
  38. 1:14is destabilizing some of the most
  39. 1:16fragile places in the world like
  40. 1:17ethiopia or what happened in myanmar it
  41. 1:20you see when when you say that there are
  42. 1:22some people who will accept it
  43. 1:23immediately and then there are some
  44. 1:24people be like that's that's not true at
  45. 1:26all but i'm sure there are a lot of
  46. 1:27people who will say well but i mean
  47. 1:30aren't they always going to be extreme
  48. 1:31people in the world you know there's
  49. 1:33always somebody saying something you
  50. 1:34know facebook themselves say they go
  51. 1:36like hey we're not we're not part of the
  52. 1:37problem we're merely a platform that
  53. 1:40people put their views on we are not
  54. 1:42part of the problem and you disagree
  55. 1:43with that so let's imagine you had a
  56. 1:45relative who had particularly extreme
  57. 1:48ideas you know
  58. 1:49we often say like the crazy uncle right
  59. 1:51right some of us have those uh
  60. 1:54that person if they're if we're talking
  61. 1:56to them one-on-one or we're talking to
  62. 1:57them at a family gathering that scale
  63. 2:00allows for resolution of ideas okay
  64. 2:02what's happening on facebook right now
  65. 2:04is ideas that are the most able to
  66. 2:06elicit a reaction and the shortest path
  67. 2:09to a click is anger
  68. 2:11those get the most reach and ideas that
  69. 2:13are more moderate or that try to bring
  70. 2:15us into synthesis or to to help us find
  71. 2:17a middle path those aren't as likely to
  72. 2:20elicit a comment from you or elicit a
  73. 2:22like right and so facebook doesn't get
  74. 2:24as much distribution so facebook is
  75. 2:26essentially propelling certain ideas out
  76. 2:28there so they may say we're the platform
  77. 2:30but really what they're also doing is
  78. 2:31they're advocating for certain
  79. 2:32viewpoints because they push those out
  80. 2:34because those get people engaged and
  81. 2:35spinning in the cycle so i think it's
  82. 2:37the thing where it's not they sit out
  83. 2:38and go how can we have more extreme
  84. 2:40ideas i don't think that's what they're
  85. 2:41trying to do but they do know that there
  86. 2:43are lots of solutions that are not about
  87. 2:45censorship it's not about picking good
  88. 2:47or bad ideas or good or bad people it's
  89. 2:49about how do we change the dynamics of
  90. 2:51these systems so that you could have
  91. 2:54good speech counter counter bad speech
  92. 2:56how do we give those more of an equal
  93. 2:58say at the table you know it's
  94. 2:59interesting that you lay it out like
  95. 3:00this because i remember talking to my
  96. 3:02friends about this saying have we become
  97. 3:05angrier in the world
  98. 3:07have we become less able to
  99. 3:09have misunderstandings because i because
  100. 3:11i sit with my friends and i go like i
  101. 3:12don't agree on so many things with my
  102. 3:14friends but i don't remember a time when
  103. 3:16that meant that i couldn't be friends
  104. 3:18with them yeah in the world of social
  105. 3:19media right now it feels like if i
  106. 3:20disagree with you on this i disagree
  107. 3:23with you on everything
  108. 3:24it seems like facebook goes if you
  109. 3:26disagree on this we're gonna make or
  110. 3:28we're gonna find more ways to to you
  111. 3:30know to find where you disagree or how
  112. 3:32to make you disagree yeah so some people
  113. 3:33have said this idea of like facebook's
  114. 3:35always looking for fault lines right so
  115. 3:37so a high quality piece of content is
  116. 3:40one that gets lots of engagement and it
  117. 3:42turns out an angry comment thread
  118. 3:44where you know it causes us to yell at
  119. 3:45each other that is viewed as a higher
  120. 3:47quality piece of content
  121. 3:49and so i you know during um covet i've
  122. 3:52had friends who wrote very long
  123. 3:54thoughtful pieces they went and digested
  124. 3:56lots of information that's going to
  125. 3:58cause less knee-jerk reaction than
  126. 4:00something that really offends you
  127. 4:03so masks don't work people are clicking
  128. 4:05on that here's the thing about masks we
  129. 4:07have to consider in our society boom
  130. 4:09yeah facebook's not pushing that so
  131. 4:11what's interesting you know in
  132. 4:13everything that you've done is even
  133. 4:14though you're a whistleblower even
  134. 4:16though you came up with this information
  135. 4:17you've never said i'm anti-facebook
  136. 4:19you've never said shut facebook down
  137. 4:20you've never said this thing needs to go
  138. 4:21away what you're arguing for is we
  139. 4:24change our relationship with not just
  140. 4:26facebook but social media company as a
  141. 4:27whole
  142. 4:28explain what that means so often when
  143. 4:30people talk with this idea if we need to
  144. 4:31change our relationship with social
  145. 4:32media it's more about like should we use
  146. 4:34our phones so much but i'm i'm
  147. 4:36encouraging us to have a conversation
  148. 4:37about what's our relationship with
  149. 4:39companies or what's the company's
  150. 4:40relationships with us okay facebook
  151. 4:42knows that they have a level of of um on
  152. 4:45accountability that's very different
  153. 4:47than either google or apple or other big
  154. 4:49platforms because in the case of google
  155. 4:51we can scrape and download their results
  156. 4:53and analyze them and see if there are
  157. 4:55biases in the case of apple we can take
  158. 4:57apart the phones and put up youtube
  159. 4:58videos saying apple phones do or don't
  160. 5:00work the way they claim
  161. 5:02but in case of facebook
  162. 5:04researchers and activists have been
  163. 5:06telling facebook hey we have found
  164. 5:08all these examples we think there's a
  165. 5:10pattern here we think there's too much
  166. 5:11human trafficking we think kids are
  167. 5:13suffering right and facebook keeps
  168. 5:14coming back because they know no one can
  169. 5:16call them on it
  170. 5:18and saying that's just anecdotal that's
  171. 5:19not real
  172. 5:21when you release the information
  173. 5:22realized it wasn't anecdotal it was real
  174. 5:23because facebook had done research right
  175. 5:25you see now this is interesting because
  176. 5:26this reminds me of a tobacco company yes
  177. 5:29it reminds us of the you know the fossil
  178. 5:31fuel companies
  179. 5:33they do the research they find out
  180. 5:34something really bad and then i mean
  181. 5:36obviously they go like we're not going
  182. 5:37to put this out there but but we know
  183. 5:39one of the biggest things that's really
  184. 5:41gotten people worried is how social
  185. 5:43media affects younger brains i mean it
  186. 5:46affects my brain i'm i don't consider
  187. 5:48myself younger anymore but
  188. 5:50young people are like it seems like the
  189. 5:52highest at the highest risk
  190. 5:55yeah facebook's internal research says
  191. 5:56that the highest rates of problematic
  192. 5:58use peak at age 14. so you're not
  193. 6:01allowed to be on the platform until
  194. 6:02you're 13. so i guess it takes a little
  195. 6:04bit of time to form that habit but but
  196. 6:06the largest fraction of any of those age
  197. 6:08cohorts they sampled and they've done
  198. 6:09studies that have a hundred thousand
  199. 6:11people these are not tiny studies
  200. 6:13they find that the largest fractions say
  201. 6:16i can't control my usage and i know it's
  202. 6:18impacting my health my employment or my
  203. 6:20school
  204. 6:21is at age 14. and so there's a real
  205. 6:23thing where young people have both you
  206. 6:26know they're struggling with issues in
  207. 6:27their life yes and they don't have the
  208. 6:29self-control and they say this they say
  209. 6:31this to researchers they say i know this
  210. 6:33is making me feel bad and i can't stop i
  211. 6:36fear i'll be ostracized if i leave
  212. 6:38um and those factors mean it's not easy
  213. 6:41for kids just to walk away it's also not
  214. 6:43easy for people to understand how to um
  215. 6:45how to fix it you have lawmakers now who
  216. 6:47are trying to figure this thing out
  217. 6:49silicon valley is way ahead of them
  218. 6:51right like way ahead of them lawmakers
  219. 6:53you see these hearings they'll be like
  220. 6:54what if what if my facebook is a insta
  221. 6:56flap they don't understand what's
  222. 6:58happening so how do we begin
  223. 7:01creating a world where we're not
  224. 7:03destroying the companies but we're we're
  225. 7:05regulating them the way they should be
  226. 7:07so the main thing i'm advocating for is
  227. 7:09around transparency because one of the
  228. 7:11problems today is because facebook is
  229. 7:12the only one that holds the cards like
  230. 7:14they can see they can see whether or not
  231. 7:15they're holding you know a royal flush
  232. 7:18or not right what they claim um we need
  233. 7:20to move to a world where we have more
  234. 7:21access to data and that can be aggregate
  235. 7:23data it can be privacy sensitive data it
  236. 7:26doesn't have to you know expose people
  237. 7:27it's not it's a false choice of privacy
  238. 7:29versus transparency right and once we
  239. 7:31have the ability to have conversations
  240. 7:33we can stop talking about boogeyman
  241. 7:35social media i see and we can start
  242. 7:37having conversations on how do we solve
  243. 7:38these problems i think what we what we
  244. 7:40need to do and this is what i'm going to
  245. 7:41spend 2022 doing is we need to start
  246. 7:43organizing people like i want to plant a
  247. 7:46youth directed movement where we can
  248. 7:48begin putting pressure on facebook to
  249. 7:50release this information because i don't
  250. 7:52think facebook is just going to do it
  251. 7:53automatically out of the goodness of the
  252. 7:55heart we have to force them to
  253. 7:57and i think there's lots lots of
  254. 7:58opportunities where we can begin putting
  255. 8:00pressure on them either socially or
  256. 8:01financially as a user of social media as
  257. 8:04a person who speaks to all my friends on
  258. 8:05social media i have my enemies on social
  259. 8:07media i appreciate this conversation
  260. 8:09because i think um if we're not careful
  261. 8:11we end up in a world where we're all
  262. 8:13fighting
  263. 8:14really just because an algorithm is
  264. 8:16trying to make us fight half of those
  265. 8:17fights i've noticed even even with the
  266. 8:19show if i say a thing
  267. 8:22on the show
  268. 8:24this has been like the most interesting
  269. 8:25thing for me is depending on
  270. 8:27what i say facebook sends that to
  271. 8:29different people one time we had it on
  272. 8:31the show
  273. 8:32one yeah one time we had on the show
  274. 8:33where um
  275. 8:34literally i'd laid out both arguments i
  276. 8:36said here's here's one way to see it
  277. 8:38there's another way to see it but then
  278. 8:40on facebook only one way went to each
  279. 8:42group yeah and i was i was fascinated by
  280. 8:45that because then people like why didn't
  281. 8:46you say the other ones like but that's
  282. 8:47exactly what i said they're like oh well
  283. 8:48i didn't see that
  284. 8:50and then you realize you know so one
  285. 8:52thing that happens is um so let's say
  286. 8:54you take a video clip so let's say you
  287. 8:56have out of any given on your shows
  288. 8:58people go and put on youtube lots of
  289. 8:59different little chunks
  290. 9:01people can reseed those onto facebook
  291. 9:03over and over again they do right and
  292. 9:05each time that enters into facebook you
  293. 9:07know someone posted to a group someone
  294. 9:08doesn't know this it enters into the
  295. 9:10network in a different way
  296. 9:12and the algorithm begins picking up data
  297. 9:14on that and it says oh interesting like
  298. 9:16these kinds of people engage these kind
  299. 9:18of stone and so and because it's
  300. 9:20entering from so many different points
  301. 9:21it actually gets a chance to learn what
  302. 9:23kinds of people or communities it
  303. 9:24engages the most with and so that echo
  304. 9:27chamber is real like the algorithm
  305. 9:29pulls us towards homogeneity it's almost
  306. 9:32like
  307. 9:33facebook knows that you react most to
  308. 9:36your neighbor when they play loud music
  309. 9:39there's no reaction from you when your
  310. 9:40neighbor's doing other things but when
  311. 9:42your neighbor plays in doubt music
  312. 9:43that's when you react actually then
  313. 9:44facebook goes i'm just gonna it's even
  314. 9:45worse than that actually so how can it
  315. 9:47be worse let's imagine there's music
  316. 9:49that you like and there's music you
  317. 9:50don't like wow it actually knows you
  318. 9:53really don't like i don't know reggaeton
  319. 9:54or something oh wow and one thing is i
  320. 9:56think it's really sad is um one of the
  321. 9:57things that was in the documents is that
  322. 9:59people talk about um especially people
  323. 10:01from marginalized communities that they
  324. 10:04will go and and correct people who are
  325. 10:06spreading um you know racist sexist
  326. 10:10homophobic comments
  327. 10:11and guess what now facebook knows you
  328. 10:13engage with those keywords so they send
  329. 10:15you so they send one
  330. 10:16yeah i i i didn't know they were doing
  331. 10:18that they don't do it on purpose it's
  332. 10:20it's a it's a side effect it's a side
  333. 10:22effect of the way the algorithm works
  334. 10:24so the algorithm is designed to get as
  335. 10:25much engagement as possible and the
  336. 10:27engagement unfortunately you're more
  337. 10:29likely there's a there's a car accident
  338. 10:31i'm going to engage yeah yeah or or more
  339. 10:33of um if they had an option of showing
  340. 10:35you a stream of different things they
  341. 10:37know you gawk a car accident so they
  342. 10:38show you more
  343. 10:40man yeah this is hard because on the one
  344. 10:42hand i go like well i need to stop
  345. 10:43looking at car accidents but the brain
  346. 10:44is the brain it's the brain you know if
  347. 10:47if all the um ai is not intelligent
  348. 10:50right we like to say artificial
  349. 10:51intelligence yes but but people who
  350. 10:53actually study it call it machine
  351. 10:54learning because it's not intelligent
  352. 10:56it's just a hill climber it's optimizing
  353. 10:59and uh you know think of all the
  354. 11:00different pieces of content you could
  355. 11:02expose to
  356. 11:04it it's not trying to show you most
  357. 11:06extreme content it happens to be to
  358. 11:08fulfill its goal function
  359. 11:09it mindlessly pushes you towards it and
  360. 11:12so this is about facebook making choices
  361. 11:14to not fix it that facebook has lots and
  362. 11:17lots of options that don't deal with
  363. 11:18letting the machine run wild they're
  364. 11:19allowing the machine real well
  365. 11:22money money that's what it is
  366. 11:25thank you so much for being here this
  367. 11:26has really been enlightening good luck
  368. 11:27on the rest of your journey thank you so
  369. 11:29much
  370. 11:30if you want to learn more about
  371. 11:31francis's advocacy work then you can go
  372. 11:33to francis haugen.com
  373. 11:38[Music]
  374. 11:47you

About this transcript

This page contains the full transcript of Frances Haugen - Facebook’s Algorithm & Teen Mental Health | The Daily Show by The Daily Show, generated from the public captions YouTube serves with the video. The transcript has 2,502 words across 374 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to translate the transcript into 100+ languages.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, translation to 100+ languages, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.