YT YouTube2Text

YouTube2Text → Transcripts → LIVE: Facebook Whistleblower Haugen Testifies at an EU Parliament Committee Hearing

LIVE: Facebook Whistleblower Haugen Testifies at an EU Parliament Committee Hearing — Transcript

by NTD News · 21,581 words · 3,680 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00the irene committee members they are
0:02members of the european parliament
0:07they are journalists
0:09i would like to welcome
0:13mr francis hogan here in the european
0:16parliament
0:19thank you
0:20for participating in this hearing
0:24in this hearing the inco committee and
0:27associated committees will listen to the
0:29testimony of mrs
0:32francis hogan an expert in algorithmic
0:35product management and former employee
0:38of facebook
0:40ms hogan had disclosed thousands of
0:43internal documents that she collected
0:45while working for facebook
0:49the purpose of the hearing is to enable
0:52imco and associated committees to better
0:55understand the challenges posed by
0:58social media algorithms
1:01to online user safety especially in the
1:04context of the
1:06digital services act
1:08which intends to regulate the provision
1:11of into intermediary services and in
1:15particular
1:17to achieve a safe and trusted online
1:20environment for users
1:23firstly before
1:25the hearing i would like to
1:28ask
1:30[Music]
1:32the representatives of the other
1:34committees
1:37to take the floor for opening remarks
1:41two minutes per committee
1:44and
1:45the first speaker in my list is
1:49adrian vasquez
2:27yes
2:37but we have the technical questions uh
2:40it's not
2:41only about
2:43miss hogan but other people say that
2:46there is no translation
2:56first row is blocked
2:59by
3:10sorry we we've tried to fix this problem
3:12and and i would like to ask you to
3:15to start from the very beginning again
3:17thank you
4:43let's go
5:21yeah
5:23sorry for this but now it seems we can
5:26can can start really
5:28so mr
5:29lazara
5:31the floor is yours please
6:32for example
7:14valencia
7:36thank you next i would like to ask
7:40juan fernando lopez or allowed to take
7:44the floor
7:45please one
7:47the floor is yours
7:48thank you chair good afternoon mrs
7:50hoygan i'm talking to you in my capacity
7:52as chair of the committee of liberty
7:54just as the normal person of the
7:55european parliament
7:56which has been engaged
7:59for a number of years
8:01by now
8:03making laws
8:05on fundamental rights issues related
8:07files and the protection of personal
8:09data most particularly privacy of users
8:13of platforms such as facebook
8:16have therefore
8:18been a relevant element of discussion
8:20throughout the years
8:22in the libya committee which i have the
8:24honor to share
8:26of course we'll welcome you here your
8:28testimony
8:29in this hearing on the negative impact
8:31on users
8:32of high-tech companies
8:34products and business models
8:37how eu digital rules can address these
8:41issues
8:43let me reassure you that
8:45we are
8:46very much engaged
8:48in the discussions around the digital
8:50services act and digital markets acts
8:52providing an opinion on these matters
8:56but the revelations that you have come
8:58up with
8:59are most troubling regarding the direct
9:02impacts on users
9:05but also a
9:06another potential an actual attack
9:08on their fundamental rights
9:10but this is not the first time we see
9:12ourselves
9:13facing this kind of situation
9:16three years ago we put in place a number
9:19of hearings on the abuses discovered
9:21regarding cambridge analytica
9:23proceedings and revelations
9:25in its resolution at the time this
9:28european parliament stated that data
9:30analysis and algorithms
9:33have an increasing impact i'm quoting on
9:36the information made accessible to
9:38citizens
9:39and that such techniques if misused and
9:42they are actually misused i add may
9:45endanger fundamental rights to
9:48information as well as media freedom
9:50and pluralism so your revelations
9:53are confirming that we have reached
9:56some kind of a high point a
9:58peak
9:59where the fundamental rights
10:01of users
10:03are at stake
10:04we see the impact of these alcoholisms
10:06exploiting mental health of teenagers
10:10micro targeting
10:12and impinging on people fundamental
10:15rights
10:16but not only for commercial purposes but
10:18also for political
10:21political targeting political purposes
10:23so as it appears by now
10:26that self-regulation by the companies
10:28themselves is failing utterly failing
10:31the simple question would be
10:35what is your assessment on the
10:37revelation
10:38on these
10:42proceedings by by facebook
10:44and
10:45the kind of regulation safeguards that
10:47you would like to see fit for the future
10:50particularly at the disposal of the
10:53european parliament field of action
10:55would you think that the current
10:56proposal at eu level
11:00is something of relevance
11:02according to your assessment as far as
11:05you know are they sufficient the package
11:08at site or
11:09anything else that could and should be
11:11done thank you
11:14thank you
11:15next i would like to ask a raphael
11:17glicksman to take the floor please
11:24merci merci charcoal leg
11:54is
13:29next i would like to ask uh
13:31travos tutarosh to take the floor
13:34representing either committee
13:37thank you very much chair and thanks for
13:39associating aida also to the to the
13:41hearing tonight
13:42congratulations mrs hogan for the
13:44courage to stand up for values that are
13:46so important to to us as consumers as
13:48individuals
13:49i represent the special committee of
13:51this house on artificial intelligence
13:53and we spent in fact the last year
13:55looking into the impact of ai
13:57and the underlying algorithms powering
13:59ai with the paramount will of this house
14:01to ensure that ai develops on the basis
14:03of rules that protect human rights and
14:05interests
14:06your revelations are confirming many of
14:08the concerns there i say fears
14:10that for many of us in this house had
14:12been
14:14related to the practices by big
14:15platforms particularly social media
14:17platforms dsa is already an ambitious
14:19starting point introducing an obligation
14:21for risk assessments on how platforms
14:23work
14:24i think it is a good tool to address
14:25systemic risk built into the algorithm
14:28algorithms but in his current logic it
14:30only concerns illegal content so
14:33how do you think we could regulate
14:35harmful content not yet passing the
14:37threshold of illegality
14:39and presuming you would have the rules
14:41or that platforms such as facebook would
14:43have the will to intervene by their own
14:44volition
14:46do you think you know expert knowledge
14:47that algorithms can be reverse
14:49engineered once they have grown into a
14:52given direction thank you
14:56thank you travis and i would like to
14:59thank uh all the committee's
15:01chairs for your opening remarks
15:05the hearing will be organized as follows
15:08first miss hogan will give a 15
15:12minutes presentation then
15:15there will be a q and a session
15:19there will be two rounds of questions
15:22and answers two minute question
15:25by members and two-minute answer by
15:29ms hogan
15:31and
15:32now i would like to
15:34ask ms hogan to take the floor for
15:37presentation
15:42thank you uh
15:44chair kazavi casa please
15:47forgive my pronunciation on things um
15:50kavanagh cavazani honorable member
15:53members of the european parliament
15:55thank you for the opportunity to appear
15:57before you and your interest in
15:59confronting some of the most urgent
16:01threats to the citizens of the european
16:03union
16:04my name is francis haugen
16:06i used to work at facebook and joined
16:09the company because i believe facebook
16:11has the potential to bring out the best
16:13in us
16:14but i am here today because i believe
16:17that facebook's products harm children
16:20stoke division
16:21weaken our democracy and much more
16:26the company's leadership knows ways to
16:28make facebook and instagram safer
16:30but they won't make the necessary
16:32changes because they have put their
16:33immense profits before people
16:36the consequences are severe
16:40the facebook platform today damages the
16:42health and safety of our communities and
16:44threatens the integrity of our
16:45democracies
16:47rising to meet these challenges won't be
16:49easy but democracies must do what they
16:51have always done
16:53when the actions of commerce conflict
16:55with the interests of the people and
16:57society as a whole
16:59democracies must step in and make new
17:01laws
17:03i am grateful that the european union is
17:05taking this very seriously
17:08the digital services act that is now
17:10before this parliament has the potential
17:13to be a global gold standard
17:16it can inspire other countries including
17:19my own
17:20to pursue new rules that would safeguard
17:23our democracies
17:25but the law has to be strong and its
17:27enforcement firm
17:29otherwise we will lose this once in a
17:32generation opportunity to align the
17:34future of technology and democracy
17:38my analysis and explanation of the
17:40documents in my disclosure
17:42is grounded in more than the time i
17:44spent at facebook i have worked as a
17:46product manager at large tech companies
17:48since 2006
17:50including google pinterest yelp and
17:53facebook
17:54my job has largely focused on
17:56algorithmic products like google plus
17:58search and recommendation systems like
18:01the one that powers the facebook news
18:03feed
18:04working at four major tech companies
18:06that operate different types of social
18:09networks
18:10i have been able to compare and contrast
18:13how each company
18:15approaches and deals with different
18:16challenges
18:18the choices being made by facebook's
18:20leadership are a huge problem
18:22for children for public safety
18:25and for democracy
18:26that's why i came forward
18:29and let's be clear
18:31it doesn't have to be this way
18:33we are here today because of deliberate
18:36choices facebook has made
18:38i joined facebook in 2019 because
18:40someone close to me was radicalized
18:42online
18:43i felt compelled to take an active role
18:46in
18:47role in creating a better less toxic
18:49facebook
18:50during my time at facebook first working
18:52as the lead product manager for civic
18:54misinformation and later on counter
18:57espionage
18:58i saw that facebook repeatedly
19:00encountered conflicts between its own
19:02profits
19:03and our safety
19:05facebook consistently resolved these
19:07conflicts in favor of its own profits
19:10the result has been a system that
19:12amplifies division extremism and
19:15polarization
19:17it undermines societies around the world
19:20in some cases this dangerous online talk
19:22has led to actual violence that harms
19:25and even kills people
19:27in other cases their profit optimizing
19:30machine is generating self-harm and
19:32self-hate
19:34especially for vulnerable groups like
19:36teenage girls
19:38these problems have been confirmed
19:40repeatedly by facebook's own internal
19:42research
19:43this is not simply a matter of some
19:45social media users being angry or
19:47unstable
19:49facebook became
19:51a facebook became a one trillion dollar
19:53company
19:54by by paying for its profits with our
19:56safety including the safety of our
19:58children
20:00and that is unacceptable
20:03i believe what i did was right and
20:05necessary for the common good
20:07but i know facebook has infinite
20:09resources which you could use to destroy
20:11me
20:13i came forward because i recognized a
20:15frightening truth
20:16almost no one outside of facebook knows
20:20what happens inside of facebook
20:22the company's leadership keeps vital
20:24information from the public
20:26the us government
20:27its shareholders and governments around
20:30the world
20:31the documents i have provided prove that
20:33facebook has repeatedly misled us about
20:36what its own research reveals about the
20:38safety of children
20:40its role in spreading hateful and
20:42polarizing messages
20:44and so much more
20:46the right answer to this emergency is
20:48new rules and standards
20:51the eu's digital services act has a huge
20:53potential
20:54it does not try to delete this problem
20:57with content regulations
20:59it takes a content neutral approach
21:02to address the systemic risks and harms
21:05of the overall business model and i
21:08strongly support this approach
21:10there's much to say about this
21:12legislation but i want to highlight
21:14three issues that are critically
21:15important based on my experiences
21:18first
21:18risk assessments and access to privacy
21:21aware data streams
21:22second comprehensive roles and standards
21:25for the business model
21:27third the danger of loopholes and
21:29exemptions
21:30i'll address these three in turn
21:34at the core of the problems we face
21:36right now is that no one can understand
21:38facebook's destructive choices better
21:41than facebook
21:42because only facebook gets to look under
21:44the hood
21:46facebook cannot remain the judge jury
21:50prosecutor and witness
21:52so a critical starting point for
21:54effective regulation is
21:56transparency full access to data for
21:59research not directed by facebook
22:02we need more experts studying these
22:03problems and they must have all the data
22:05they need to do it
22:07it is critical to get this right and the
22:09devil will be in the details for example
22:12if you write a broad exemption for
22:14transparency for anything classified as
22:16a trade secret
22:18the companies will say everything is a
22:20trade secret
22:22facebook would still operate in the dark
22:24continuing to make choices that go
22:26against the common good
22:28so i believe researchers should respect
22:30trade secrets once they have received
22:32proprietary data but trade secrets can
22:35and should not be an excuse for
22:36companies to refuse access to data
22:40access to data is the starting point
22:43it's what permits researchers and
22:44regulators to assess risks and harms of
22:47the whole system of profiling targeting
22:50and engagement based ranking
22:52the evidence in my disclosures to the
22:54sac make clear that engagement-based
22:57ranking systems are one of the root
22:59causes for some of the greatest systemic
23:01risks that social media poses to our
23:03societies
23:05this is a system level problem this is
23:08why new rules must not be limited to
23:10illegal content but must also include
23:13the recommendation of content that
23:15violates a platform's terms and
23:16conditions
23:18using content neutral tools the dsa
23:21should force platforms to take
23:23responsibility for the risks beyond the
23:25spread of illegal content
23:27such as manipulation of elections the
23:30explosive spread of disinformation or
23:32the harms of teenage mental health
23:35the heart of this legislation lies in
23:37the strength of articles 26 and 27.
23:41this brings me to my third point
23:43i have read with great interest that
23:45their efforts to exempt news media
23:47content from the rules of the dsa but if
23:50you're going to make content neutral
23:51rules then they must really be neutral
23:54nothing is singled out and nothing is
23:56exempted
23:57let me be very clear
23:59every modern disinformation campaign
24:02will exploit news media channels on
24:04digital platforms by gaming the system
24:07if the dsa makes it illegal for
24:09platforms to address these issues we
24:11risk undermining the effectiveness of
24:13the law
24:14indeed we may be worse off than today's
24:15situation
24:17if you get the dsa right for your
24:20linguistically and ethnically diverse
24:23450 million eu citizens
24:26you can create a game changer for the
24:28world
24:29you can force platforms to price in
24:31societal risk to their business
24:33operations so the decisions about what
24:35products to build
24:37and how to build them is not purely
24:39based on profit maximization
24:41you can establish systemic rules and
24:44standards that address risks while
24:46protecting free speech
24:48and you can show the world how
24:50transparency
24:51oversight and enforcement should work
24:55let's not forget
24:57we have stood at these crossroads before
24:59when tobacco companies claimed that
25:01filtered cigarettes were safer for
25:03consumers it was possible for scientists
25:06to independently invalidate
25:08that
25:09this marketing message
25:11because it was just a marketing message
25:13and confirmed that in fact they still
25:16posed a serious threat to human health
25:18they were actually in fact more toxic
25:20than unfiltered cigarettes but today we
25:23can't make this kind of independent
25:24assessment of facebook
25:26we have to just trust facebook says what
25:29facebook says is true and they have
25:31repeatedly proved they do not deserve
25:33our blind faith
25:35trust is earned it's not just given
25:38facebook's regulators can see some of
25:40these pro can see some of the problems
25:42but they are kept blind was causing them
25:45and thus can't craft specific solutions
25:48they can't even access the company's own
25:50data on product safety much less conduct
25:53an independent audit
25:54how is the public supposed to assess if
25:57facebook is resolving conflicts of
25:58interest in a way that is aligned with
26:00the public good if it has no visibility
26:03and no context into how facebook really
26:05operates
26:06this must change
26:11dear members of dear members of
26:12parliament
26:14there is a lot at stake here
26:16you have a once in a generation
26:18opportunity
26:19to create new rules for our online world
26:22a safer more enjoyable
26:24social media is possible
26:26if there are only two things that
26:28everyone takes away from these
26:29disclosures it should be
26:31first facebook chooses profit over
26:33safety every day and without bold action
26:36from lawmakers this will continue
26:40the second is that facebook has
26:41exploited its ability to hide the actual
26:45behavior of the platform
26:47to allow our safety decay to an
26:49unacceptable level
26:50if facebook is allowed to continue to
26:52operate in darkness we will only see
26:54escalating tragedies as a result
26:58i came forward at great personal risk
27:00because i believe we still have time to
27:03act
27:04but we must act now
27:05thank you
27:10[Applause]
27:21[Applause]
27:22thank you miss hogan for your
27:25presentation and now we can start with
27:27the q a session
27:29first i would like to ask andres fob
27:32from etp group
29:20thank you andreas
29:22miss hogan please
29:31sorry my apologies so that's a lot to
29:33address in two minutes but i'll do my
29:34best
29:35with regard to the risk to democracies
29:37um
29:38one of the so the inherent danger in
29:41engagement based ranking and mark
29:43zuckerberg has said this specifically he
29:46has said in 2018 even uh engagement
29:49based ranking is dangerous because
29:51people are drawn to engage with extreme
29:54content more than than more mainline
29:57content
29:58when that happens
29:59effectively we give a much much much
30:02much larger fraction of the public forum
30:04to the most extreme the most polarizing
30:07the most divisive content
30:09this also happens in ads because ads
30:12that get more engagement are cheaper
30:15running an angry hateful divisive
30:18political campaign will be five to ten
30:20times cheaper than running one that is
30:22about empathy and compassion democracies
30:25cannot exist if we can't see each other
30:27as human beings if we can't have empathy
30:30and solidarity with each other
30:33and so i believe that having a system
30:35that amplifies a divisive content like
30:38that will erode our democracies the
30:40second question is around um what can we
30:43do with regard to regulating private
30:45economy um
30:46and
30:48and
30:49in a way that is um
30:51uh
30:52i'm trying to hit both freedom of
30:54expression and we may not get there um
30:56so i'm a strong proponent of what i call
30:58one two three risk assessments so
31:01companies should assess themselves
31:03but there should also be some party that
31:05can gather feedback from the community
31:08because facebook is very homogeneous it
31:10largely speaks english it's largely
31:12isolated in california at least those
31:14who are making
31:15most decisions
31:17and unless you go and survey civil
31:20society groups you're not going to get a
31:21full picture
31:23the three is facebook should have to
31:25articulate how it will solve those harms
31:28because it has gotten into a rut where
31:30every time something is exposed all they
31:32do is say we're so sorry this is really
31:35hard we're working on it we need to
31:37change our relationship with these
31:38platforms where they have to actually
31:40articulate what the plan is for fixing
31:42them
31:43and they should have to disclose
31:46aggregate privacy conscious
31:48privacy sensitive data so that we can
31:50independently see are they actually
31:52making progress
31:53on those concerns
31:55um i think that is probably the most
31:57flexible way of regulating a private
31:59company
32:00because it can happen and be dynamic
32:02over and over again the concerns of
32:04today and the concerns of five years ago
32:06will be different concerns and as long
32:08as we have that kind of
32:10framework where the community's
32:11involvement is in place in addition to
32:14um the needs of just the company i think
32:17there's a way of covering more of those
32:18issues
32:20thank you
32:21now i would like to ask christopher
32:23demos to take the floor please
32:26thank you so much and a great thanks to
32:29you miss haugen for being here today uh
32:32i'm really happy that you took our
32:34invitation and came it's very timely for
32:36the work we do on on the dsa
32:39and i agree with you that this is a
32:41historic moment we have a chance we have
32:42a historic opportunity to do something
32:45for
32:46regulating and maybe take back control
32:48to society from the big tech
32:50companies
32:52thank you for your all your statements i
32:55would like to discuss a little bit about
32:56how we modernize the liability
32:59rules you also said it in the congress
33:01that you know
33:02in united states you have section 230
33:05and
33:07giving the tech companies immunity from
33:11civil liability and we have the same in
33:12the e-commerce they are exempted from
33:14liability the platforms
33:16but when you talk about one two three
33:19you kind of send the signal that you
33:20want the platforms to become
33:23liable accountable for their algorithms
33:26and how they work so could you maybe uh
33:28give us your suggestions on how we could
33:31make them liable or accountable at least
33:34because i think this is this is key of
33:36course they should not be liable for
33:38what i put on facebook but how they use
33:40what i put on facebook i think that
33:42could be important for them to to be so
33:44that i would be happy to to hear
33:46about that
33:48and i think that's the most important
33:50thing so i will give you more time to
33:51answer instead of using all my time
33:53thank you
33:55thank you crystal
33:56please
33:57miss hogan
33:58the question of how to increa increase
34:02the accountability of these platforms is
34:03a very important one one of the things
34:06that i think is essential is uh my
34:10understanding is today
34:11there's no
34:12there's really no practical way to bring
34:14a class action lawsuit either in the
34:16united states or in europe partially
34:18because we don't have a way of seeing
34:21differential harms
34:23when you think about mandatory data
34:25there are ways to segment the population
34:28that are privacy sensitive i worked on
34:30some of these
34:31where you
34:32produce a cohorts and you can report
34:36stats in an aggregated way on
34:38differential harms per population
34:41so what we found in the case of covid
34:43misinformation was that four percent of
34:45the population received 80 of all the
34:48coveted misinformation
34:50this is not a novel problem is a problem
34:52that repeats itself with every harm type
34:55so i think one thing is if you want to
34:57hold the platforms accountable for say
34:59some children being overexposed to
35:02self-harm content there has to be a way
35:04for us to get information on those
35:07constant that concentration of harm
35:09because like i said every harm type has
35:12these kinds of problems because of how
35:14the algorithms push people towards
35:17homogeneity
35:18i think the second question on liability
35:20is around
35:22responsibility i'm a strong proponent
35:24that uh so one of the repeated things we
35:27hear from facebook is that no one can no
35:29person can be held accountable for any
35:31decision that happens on the platform
35:33that committees of people decide
35:34everything
35:35i think requiring that a name should be
35:38associated with every change would
35:40actually radically change the culture of
35:41facebook
35:42because today people hide behind that
35:45that unified front
35:47and it means that no one stands up for
35:49what's right because no one will ever
35:50have it traced back to them
35:52i think the last question is around
35:53things like executive liability
35:56i am relatively agnostic on executive
35:59liability i think it amplifies the
36:01effect of any law that you pass so if
36:04you feel really confident in the law
36:06that you've passed then
36:07executive liability can be a good thing
36:10or at least makes them take it a lot
36:11more seriously um but at the same time
36:14if the law is not good it can cause
36:16counter effects and side effects so i am
36:19i'm not a lawmaker i i'm it's probably
36:21beyond my scope to weigh in on that
36:25thank you next i would like to ask tita
36:28franco to take the floor peter the floor
36:30is yours please thank you very much mr
36:32chair and thank you mrs haugen
36:36you worked at a unit for civic integrity
36:39as you mentioned and you also said in
36:41one of your interviews that dissolving
36:43that unit was the sign that facebook
36:46uh was not interested in fighting this
36:49information
36:50and the digital services act which i
36:54understand you you read here and you you
36:56are aware of all the details we have now
36:59different articles both on risk
37:01assessment mitigation of risks
37:04independent audit enforcement mechanisms
37:07so it would make actually mandatory that
37:10facebook would have such a unit and
37:13action will have to act against this
37:15information and harmful content which is
37:18something to my knowledge which is not
37:20required and is not a proposal in the
37:22u.s
37:24but at the same time in our proposal as
37:26it is drafted it will be still facebook
37:29itself that will choose the ways how to
37:32mitigate the risk
37:34do you think this is a good approach or
37:37do you think we should have perhaps
37:40national regulators to
37:42micromanagement
37:44manage such a massive complex system
37:48thank you
37:50thank you miss hogan please
37:54i feel fairly strongly that the role for
37:57regulators is to make sure that facebook
37:59has to
38:00follow at least a certain amount of the
38:02rules of the road so this is less so a
38:04thing of saying you must do x y z in
38:08terms of how to change your platform but
38:10is more so you know if we have like a
38:12one two three approach that they have to
38:15you know take an input from lots of
38:17community groups in a way that is not
38:19filtered by facebook itself or that they
38:22have to report on statistics that would
38:24allow us to see have they actually made
38:25progress on what they said was the
38:27solutions
38:28um i'm a strong proponent also for for
38:31criteria on what is a rigorous enough
38:34solution like it has to be a good faith
38:36plan on how will you solve these things
38:38i think it is uh it will be difficult to
38:41have a regulator that mandates specific
38:43exchanges partially because any time you
38:46draw a fence companies figure out how to
38:48run around the fence
38:50and so having something dynamic like a
38:52risk assessment process is important
38:54but the secondary thing is i i feel
38:57strongly that part of why
38:59regulators should be gathering that hav
39:02forcing platforms to disclose that data
39:04or forcing them to disclose their harms
39:07publicly is right now people never get a
39:10fair chance to assess if they want to
39:12use these products
39:14right facebook because they have misled
39:16us over and over again
39:18people's dignity is being violated
39:20because they are exposed to harms that
39:22they are told don't exist
39:24right and that's not that's not fair or
39:25just
39:26um
39:30i also think that if you have auditors
39:33that are paid for by facebook you will
39:34run into problems with um
39:38they will auditors will want to get
39:39hired again and if they are paid by
39:41facebook they will likely not do as
39:42thorough job as is necessary um so
39:45having the ability for the public to see
39:47risk assessments is actually quite
39:48important for consumers to be able to uh
39:51have informed choices
39:53thank you
39:56thank you and next uh i would like to
39:58ask alexandra gaza to take the floor
40:01please
40:02thank you chair and thank you mrs hogan
40:04for being here today for coming forward
40:06i think i can speak on behalf of almost
40:08all my colleagues saying that we're
40:10extremely grateful to you for that and
40:12you're doing that at
40:13great personal risk and we really
40:15appreciate
40:16i have two questions um you have
40:18stressed the importance of access to
40:20data for civil society
40:23do you think the current wedding in the
40:25dsa which under certain circumstances
40:28allows for access for vetted academic
40:32researchers only
40:34is enough
40:35or do you think we should also give
40:37access to ngos and investigative
40:40journalists for example under certain
40:43conditions and you might want to outline
40:44those conditions and those processes
40:47that's the first question and my second
40:49question is
40:51you have been an advocate for strong
40:53regulation
40:55that is needed to make sure facebook
40:57doesn't continue to harm society but we
40:59all know that regulation needs to be
41:02enforced to be effective
41:04what kind of enforcement do you think we
41:07need will national oversight and 27
41:10member states be able will those
41:12agencies be able to recruit enough
41:14experts or do you think a european
41:17enforcement agency with a strong center
41:19of competence for systemic risks should
41:22be established would that be a way to
41:25make sure that the dsa really has teeth
41:28and that europe could become a beacon
41:31for the world in digital regulation
41:32thank you so much
41:34thank you miss hogan please
41:37i think this question does a great job
41:39of highlighting the importance of having
41:40an ecosystem
41:42of of both analysts and ecosystem of
41:45people proposing solutions is so
41:47important
41:48so the first question is around
41:51giving data to vetted academic
41:53researchers and like whether or not that
41:54circle is broad enough
41:57just so people have context in the case
41:59of twitter which is substantially more
42:01transparent than facebook there's
42:03something called a fire hose and the
42:05fire hose contains one tenth of all the
42:08tweets that appear on on twitter
42:10there are easily ten to twenty thousand
42:12analysts around the world that analyze
42:15this public data
42:17and because they have access to it they
42:19find things like influence operations
42:21information operations being run by
42:24foreign actors like russia
42:26a huge fraction of those operations that
42:28exist on facebook were caught not on
42:31facebook they were caught on twitter
42:33and when people looked at those accounts
42:35they found identical accounts on
42:37facebook
42:38if facebook had to like i i strongly
42:40encourage you to not just give data to
42:42academics because the population of
42:45incredibly competent people who are
42:47experts in the space exist outside of
42:50academia and they exist outside of
42:51government and they will continue to do
42:53so
42:54we need to build that ecosystem of
42:57having it of of people building the
42:58muscle to understand facebook if we want
43:01to be safe
43:02and they often are very novel like they
43:04find different strategies that are are
43:06not yet um
43:08maybe have have reached academia
43:11the second question is around whether or
43:12not enforcement and actually add one
43:14thing on there i i would love to do
43:17commentary if i had data right i am not
43:19an academic i probably won't get
43:21accepted i'm not gonna be a professor
43:23probably um but you want to be able to
43:25have people like me who have the
43:26expertise from inside companies to then
43:28illustrate patterns with public data the
43:30last question is on um member states um
43:34so just to be super super blunt here the
43:37number of experts formal experts in
43:39these things so i think this is how the
43:41algorithms really work and the
43:43consequences of them there are very very
43:45few in the world because you can't get a
43:47master's degree in it you can't get a
43:48phd in it you have to go work for one of
43:50these companies and be trained up
43:52internally and i sincerely
43:55worry that if you uh delegate this
43:57functionality to 27 member states you
44:00will not be able to get critical mass in
44:02any one place
44:04um like i think it's a it'll be very
44:06very difficult to get enough experts and
44:08distribute them that broadly
44:12there are definitely people who have
44:13worked in industry who really want to
44:15give back by contributing to public
44:18accountability
44:19and so having a home for them i think is
44:21important having a place where they can
44:24cross-pollinate to academia and other
44:26outlets but having a home is going to be
44:29really important
44:32thank you next i would like to ask
44:35alexandre passo to take the flaw
44:38grazie president
44:44opportunities
46:37thank you miss hogan please
46:41um one of facebook's most common
46:43refrains in response to me has been uh
46:46it makes no business interest to design
46:49a system that intentionally spreads
46:51misinformation or spreads polarizing
46:53content
46:54and the reality is
46:56it's not a question that they were
46:57deliberate in designing these systems to
47:00do these things it is a question that it
47:03was deliberate that they did not change
47:05them once they discovered that a side
47:07effect of their choices was to spread
47:09more misinformation or more division
47:12there are many possible changes
47:15many of them content neutral
47:17to reduce the amount of misinformation
47:19division divisive content i'll give you
47:22a couple examples
47:23uh you could require clicking on a link
47:25before you reshare it
47:27that reduces misinformation
47:28substantially you could make reshare
47:31chains shorter you could say you know
47:34after you get beyond friends of friends
47:36if you want to re-share it at that point
47:38you need to copy and paste it and and
47:40you can totally share whatever you want
47:42like we're not oppressing you we're not
47:44judging your ideas we're just saying
47:46make an intentional choice if you want
47:48to spread it further that has the same
47:50impact as the entire third party
47:52fact-checking system in terms of
47:53reducing misinformation only
47:56it works for all languages in the world
47:59or things like facebook knows if they
48:00show you more content from your family
48:02and friends for free you get less
48:04misinformation less division less
48:06polarization so yes facebook knows lots
48:09of ways to change the system but it
48:11chooses not to because it will grow less
48:13you'll consume less content you'll
48:16consume less ads they'll make less money
48:19with regard to
48:21having more representative views
48:24i think as long as there is engagement
48:26based ranking it is likely that the
48:28system will continue pulling people
48:29towards polarization facebook has run
48:32the same experiment over and over again
48:33where they have a brand new account
48:36no friends no interest and they they
48:38follow just a few
48:40mainline interests so maybe they're
48:42center right or they're center left and
48:44the thing that happens every time or
48:45it's things like healthy eating and
48:47every time if you just engage with the
48:49recommendations facebook gives you
48:52it keeps getting more and more extreme
48:54you know in the case of politics it
48:56leads to content about killing your
48:57opposition or white genocide you know
49:00crazy ideas um for
49:03that healthy eating one it leads you to
49:04anorexia content
49:06so there's a real thing of we need to um
49:10be thinking about
49:11it's not bad people or bad ideas
49:14it's about a system that is biased
49:17towards spreading the most extreme ideas
49:20that's the real problem and then lastly
49:22with regard to the norwegian law i'm not
49:23familiar with this um and i don't know
49:25what modified means um so i'm
49:27unfortunately i can't comment on it but
49:30i'm happy to later
49:34thank you
49:35the next speaker is cosmos lotovsky
49:38cosmo the floor is yours
49:41thank you very much i will speak in
50:17[Music]
50:48[Music]
52:12m
52:19thank you cosmo
52:20miss hogan please
52:30thank you um
52:33so i'm going to bend more towards the
52:34bottom and go kind of upwards um so i
52:36think this question of is facebook
52:38different is a really important question
52:40facebook is substantially less
52:42transparent than other large platforms
52:45and i think that plays out in that they
52:47make less responsible decisions as a
52:49result so i'll give you a concrete
52:50example google knows their results can
52:52be downloaded and that people analyze
52:54them and as a result and they they take
52:57action they remove toxic content because
53:00they know that people are watching
53:03they also staff engineers who work on
53:05search to explain search to people
53:07twitter knows that people are analyzing
53:09the twitter twitter
53:11what tweets are getting distributed and
53:13i think they're more accountable and
53:15more responsible as a result and so
53:17there is a real tie between transparency
53:20and the decisions and and optimizations
53:22that these platforms make
53:24the question around encroaching on
53:25freedom of speech is very very important
53:27to me
53:28i don't believe that focusing on
53:30content-based solutions is sufficient
53:33because
53:34every time you make a decision that is
53:36about content you now need to have a
53:38system that can intervene in the 5 000
53:41languages that exist on earth you have
53:44to re-tune and rebuild the ai over and
53:46over and over again
53:48and that and facebook hasn't right it's
53:51actually harder to support the third
53:53largest language in ethiopia than it is
53:55to english because the training systems
53:58the the contractors that we would use to
54:00be able to do that um
54:02don't exist those services don't exist
54:04so we need to think about how we make
54:06systematic changes around amplification
54:09how we make social media that is more
54:11about our family and friends because we
54:14like we liked social media when it was
54:15about our family and friends right i'm
54:17not asking you to give up social media
54:19i'm asking you to have human scale
54:21social media um i'll give you an example
54:25of
54:25an intervention that would radically
54:27decrease misinformation if you said if
54:30you have more than 5 000 members in a
54:32group
54:33you have to turn on self-moderation
54:35as someone who ran a group that 3 000
54:37people i know that's hard i know that
54:39less content would get out there and if
54:42we know that one of them the most
54:44dangerous kind of amplification levers
54:47is half a million million person groups
54:50those groups can still exist if their
54:52members are so committed to their groups
54:53that they're willing to staff the
54:54moderators but with great power comes
54:57great responsibility if you have a
54:59medium that can blast out a message to a
55:01million people
55:02you know having some constraints and
55:04saying hey like maybe you should
55:06moderate for yourself the content i
55:08think is an important thing
55:10um the last question is around are these
55:12things deliberate and as i mentioned
55:14before i don't believe they deliberately
55:17set out to make a polarization engine
55:20but they have known for years they've
55:22known since at least 2018 that political
55:24parties in europe were saying we are
55:27running more polarized extreme ideas at
55:31least on social media because that's the
55:32only thing that gets distributed like if
55:35we share a white paper on our
55:36agricultural policy
55:38we get crickets now we used to at least
55:40get distribution
55:42and i worry that by the time we reach
55:44the ballot box
55:46the options that we even get to choose
55:48have been constrained by facebook
55:50because
55:51if the entire time you were only able to
55:54run extreme ideas or give that the most
55:55voice have you really been given a
55:58choice to vote for
55:59for for like what would reflect your own
56:02needs because it's already been filtered
56:04by facebook's algorithm
56:06so society and and technology are now
56:09intertwined in a way that they've never
56:10been intertwined before
56:14thank you next i would like to ask
56:16martin scherzwan to take the flow
56:20martin yes thank you chair and thank you
56:22miss horn for being with us today on
56:24forbearing testimony
56:26on behalf of the left group in the
56:27european parliament i just want to tell
56:29you that we admire your courage to stand
56:31up against facebook and to reveal its
56:33business practices
56:35that driven obviously by profit interest
56:37are harmful for consumers free choice
56:41toxic for the mental health of miners
56:42and after all also dangerous for
56:44democracy
56:46as such uh your insight in the dealings
56:48of facebook's uh facebook comes at the
56:51crucial time as you know we are as
56:53european lawmakers now working on a
56:55hopefully strong regulation for big tech
56:58and working on reining in their
57:00monopolistic position related harmful
57:02business models and let me just add very
57:05briefly two more aspects to what has
57:06been said
57:08so far
57:09firstly i'd like to know if you have
57:11ever gained knowledge of decisions taken
57:13by the facebook management or maybe even
57:16by mark zuckerberg personally that were
57:18tailor-made when it comes to the
57:20european union have there been decisions
57:22taken with special focus on elections
57:24taking place in the european union or
57:27elsewhere in europe any decisions
57:29targeting in particular european
57:31consumers and if so what were these
57:34decisions and secondly
57:36i'm
57:37personally in favor of a ban
57:39of targeted advertisement i think a
57:41business model based on the exploitation
57:43of sensitive data such as political or
57:45sexual orientation or
57:48health related data is not only immoral
57:50i think it should be illegal do you
57:53think that there is a place in the
57:54future digital market for businesses
57:57that thrive through the exploitation of
57:59personal data thank you very much
58:02thank you martin miss hogan please
58:06um i am not aware of any actions that
58:09were specifically taken with the
58:11european union in mind um other than i
58:14know that germany was considered a quote
58:17at-risk country um during their
58:19elections um during the lead-up to it uh
58:22i i don't know this for sure but i'm
58:24pretty sure
58:25the reason why germany was given that
58:27privilege um were those extra resources
58:29was because they feared things like the
58:31dsa right the other countries on the
58:34atlas that were tier one and the at-risk
58:36countries are places where
58:38you know there's people actively killing
58:40each other um and so it shows that
58:42facebook invests resources for safety
58:45not based on
58:47equity right every election in europe is
58:50important they should all get
58:51supervision with regard to influence
58:53operations or things like that um
58:56facebook should have to staff enough
58:58security personnel that that's possible
59:00because right now those functions are
59:03all hidden and how much they're willing
59:05to invest in those that's hidden
59:08with regard to targeted ads
59:11i'm a strong proponent that people
59:12should be allowed to make choices with
59:14regard to how they are targeted and i
59:16encourage um prohibiting dark dark
59:19patterns that force people into opting
59:21into those things
59:22um platforms should have to be
59:24transparent about how they use that data
59:26and i'm a big proponent of it there's a
59:28lot
59:29they should have to also publish
59:31policies like
59:32do they give flat ad rates for all
59:35political ads because you know we
59:36shouldn't be subsidizing hate in
59:38political ads
59:40um
59:41with regard to will there be businesses
59:43that can be successful without
59:44exploiting data
59:46i strongly believe that facebook will
59:49not be dislodged from what i call
59:51personal social media so there's
59:53broadcast social media which is things
59:55like tick tock people create on tick
59:57tock because they want to reach the
59:58largest audience possible the market
1:00:00works for things like tick tock and you
1:00:03see lots of tick tock clones youtube has
1:00:05one snapchat has one instagram has one
1:00:08um
1:00:09but you don't see clones of facebook and
1:00:11the reason for that is in personal
1:00:13social media people create because they
1:00:15want to reach very specific people and
1:00:17unless you move the entire front network
1:00:19somewhat simultaneously that's not going
1:00:21to happen
1:00:22and we can talk about interoperability
1:00:24but it's outside the scope of this
1:00:25question i don't think interoperability
1:00:27will work and so in a world where um
1:00:30it is almost impossible to escape that
1:00:33center of mass you know three years ago
1:00:35even there were still some countries
1:00:37that had alternative personal media
1:00:39now there are there really aren't any
1:00:41countries except for china because they
1:00:43banned facebook um and so i
1:00:46i
1:00:46i think there is a space for companies
1:00:48that are more just with information
1:00:51but we're not gonna we've we've lost the
1:00:53chance for that world facebook is
1:00:55personal social media and will be for
1:00:57the foreseeable future
1:01:00thank you next i would like to ask maria
1:01:03marisa matthias to take the floor please
1:02:04foreign
1:03:25thank you miss hogan please
1:03:35um
1:03:35[Music]
1:03:37before i dive into my answer i'll do a
1:03:38quick plug for whistleblower protections
1:03:42we will only need whistleblowers more in
1:03:44the future right unless we have
1:03:46substantially more transparency into
1:03:48technology as a whole and remember the
1:03:51dsa is going to cover platforms online
1:03:53it's not going to cover
1:03:54self-driving cars it's not going to
1:03:56cover other systems that might be
1:03:57harmful
1:03:59it is extremely important to make sure
1:04:00that there are protections and avenues
1:04:03for whistleblowers of all levels to feel
1:04:05safe coming forward
1:04:07because technology is only going to
1:04:08accelerate and become more dangerous
1:04:11the second question is around automatic
1:04:13control systems
1:04:14um i am extremely skeptical on ai being
1:04:18the solution
1:04:19um this is because it is very difficult
1:04:22to like ai is not actually intelligent
1:04:25it is a i've heard the metaphor that ai
1:04:28is like having a million first graders
1:04:31solve a problem if a million first
1:04:33graders together can solve it ai will be
1:04:35great for you but if it requires context
1:04:38and understanding of a larger sphere ai
1:04:41is very bad at that and we see this over
1:04:43and over again in facebook's own
1:04:45documents things like counterterrorism
1:04:47speech being consistently labeled
1:04:50terrorism speech
1:04:51and that's because the words are similar
1:04:53they don't understand the nuance in the
1:04:54context
1:04:55i'm a strong proponent that facebook
1:04:57should have to disclose all of its
1:05:00safety systems in terms of you know are
1:05:02you looking for terrorism content are
1:05:04you looking for self-harm are you
1:05:05looking for hate speech what are those
1:05:07systems
1:05:08and they should have to disclose which
1:05:09languages they exist in and they should
1:05:11have to disclose the performance of
1:05:13those systems per language because the
1:05:15reality is even when they say they they
1:05:17implement it in something other than
1:05:18english often they don't do a good job
1:05:20of having actually be performant
1:05:23and so requiring things like disclosing
1:05:25a thousand samples at each score
1:05:27percentile would give us an ability to
1:05:30come and say wow your terrorism detector
1:05:32is stripping all the counterterrorism
1:05:34out
1:05:35and you see this over and over again in
1:05:37the case of myanmar you know i was asked
1:05:40by a journalist who had read through the
1:05:41documents he said how is it possible
1:05:44that there is no misinformation
1:05:46classifier in myanmar like a system for
1:05:48labeling what is misinformation it's
1:05:50because facebook hasn't really invested
1:05:52in third-party fact checking there like
1:05:54it's difficult for them to have a
1:05:55misinformation classifier
1:05:57so this question of you know can
1:05:59automatic systems be the solution i
1:06:01think you have to make these platforms
1:06:03smaller more human scale
1:06:05slower
1:06:07because that will automatically reduce a
1:06:09lot of the harmful content and if you
1:06:11focus on these automatic systems they
1:06:13will not work for the the most
1:06:16ethnically diverse places in the world
1:06:17with linguistically diverse places in
1:06:19the world which are often the most
1:06:21fragile i guarantee you there's a lot of
1:06:23languages even in europe that have no
1:06:25safety systems or they have a very
1:06:27minimal number of safety systems for
1:06:28this reason
1:06:30the last question on individual
1:06:31responsibility i'm a strong proponent of
1:06:34every change should have to have a name
1:06:36associated with it because facebook has
1:06:38hidden behind
1:06:40its collaborative culture to remove
1:06:42responsibility from anyone they say
1:06:44committees make the choices not
1:06:45individuals
1:06:46and even the action of saying you just
1:06:48have to have a name on each change
1:06:50because someone needs to be able to
1:06:52stand up and say i believe in this thing
1:06:54because today like that culture of a
1:06:56lack of accountability means people also
1:06:58don't push back
1:06:59they just go with the flow
1:07:03thank you and next i would like to ask
1:07:06oxel was to take the floor hawks the
1:07:09floor is yours please
1:07:11thank you but i would also like to
1:07:14speak in german
1:07:32[Music]
1:07:48[Music]
1:08:27[Music]
1:09:02please
1:09:12with regard to the change
1:09:14of name for facebook from facebook to
1:09:16meta
1:09:18i am extremely concerned about meta and
1:09:21the metaverse
1:09:22partially because i think it illustrates
1:09:24a meta problem of facebook which is they
1:09:26really like to move on
1:09:28they really like to prioritizing growth
1:09:30and expansion over being actually
1:09:32finishing what they build like making
1:09:34sure it's safe and responsible so the
1:09:36fact that they can afford 10 000 more
1:09:39engineers to build video games
1:09:41when they allegedly can't afford to have
1:09:4410 000 engineers working on our safety i
1:09:46find that unacceptable and i find it
1:09:48unconscionable um with regard to what
1:09:51the direct harms will be of of metaverse
1:09:53not just these you know you know
1:09:55zero-sum things um i am very concerned
1:09:58about having lots and lots more sensors
1:10:01in our homes and our offices
1:10:03because like to do a lot of this
1:10:05metaverse stuff you have to actually
1:10:07expose yourself a lot more and facebook
1:10:10has demonstrated that they lie
1:10:12anytime it's useful to them
1:10:14and so the idea that we're supposed to
1:10:16fill our homes and offices with lots
1:10:17more sensors from a corporation that is
1:10:19not transparent and like we have no
1:10:22transparency into i think is a bad idea
1:10:25um along with that in the case of
1:10:26workplaces at least like people often
1:10:29don't get to opt in on what technology
1:10:31is used by their employer so if someone
1:10:33says hey i've read up on this i don't
1:10:35feel safe having facebook have even more
1:10:37data on me
1:10:38if your employer says oh well we're a
1:10:40metaverse company like you have to
1:10:42accept this is how we do voice you know
1:10:44how we do meetings you don't there's no
1:10:47consent there and i think that also is
1:10:49dangerous
1:10:50um the only thing that i am excited
1:10:52about metaverse is i think it is much
1:10:54more human skill
1:10:56so most of the interactions are at the
1:10:57level of one to maybe 10.
1:11:00um i think vr becomes pretty
1:11:01overwhelming beyond that point
1:11:03um and and that's a great illustration
1:11:05of systems that are just intrinsically
1:11:08safer like your uncle can have really
1:11:11out their views and he might ruin
1:11:13thanksgiving but he's not getting a
1:11:15megaphone to a million people
1:11:16right so by default when you design
1:11:18human scale systems you get for free
1:11:21less spread of extreme ideas
1:11:24um with regard to
1:11:26uh blocking more um
1:11:28[Music]
1:11:30blocking too much free expression um i i
1:11:33that's part of why i'm saying we need to
1:11:34have
1:11:35samples like for any safety system they
1:11:37have we should be allowed to see a
1:11:39thousand samples at various different
1:11:40scores because i think we would be
1:11:42really shocked at how much good content
1:11:45gets thrown in with the bad content um
1:11:47especially on things like
1:11:48counterterrorism versus terrorism
1:11:50so i'm much more a proponent of how do
1:11:53we design systems to be safe
1:11:56versus how do we touch the content a lot
1:11:58because like i said before i don't
1:11:59believe um content-based solutions scale
1:12:02to the most fragile places in the world
1:12:04then the last question around
1:12:06balance balanced on teen advertising i
1:12:09think it's a really hard thing because
1:12:11right now facebook claims that they
1:12:14don't target ads like that on teenagers
1:12:16like you can you can say i want to reach
1:12:18you know women who are between 13 and 16
1:12:20or whatever but the reality is ai learns
1:12:24biases and ai learns things about you
1:12:26even if you're not aware that it's in
1:12:28those systems like they're going to
1:12:29learn keywords in those ads and it's
1:12:31going to act like targeting on
1:12:33vulnerable teenagers and so um i think
1:12:36there are ways that you can do that in
1:12:38terms of um
1:12:40having it be a little bit more like tv
1:12:42you know where ads are blasted out to a
1:12:44much more diverse and less confined
1:12:47audience but as long as ai can learn
1:12:49those relationships it's going to micro
1:12:50target
1:12:53thank you the next speaker is tom wonder
1:12:58thank you the floor is yours please
1:12:59thank you chair and thank you miss hogan
1:13:01first of all also from my myself thank
1:13:03you for being here
1:13:04it's not only very courageous but in a
1:13:06way it's also necessary because facebook
1:13:08to us functions like a factory with with
1:13:12closed factory with big concrete walls
1:13:14and you're right the digital services
1:13:16act will oblige them to put at least
1:13:18some windows in those big concrete walls
1:13:21they will be obliged to provide insight
1:13:22into the processes of the algorithms
1:13:24that determine what we see and why we
1:13:26see it and that's an important first
1:13:28step but i think
1:13:29it it should not stop there it's
1:13:31unacceptable like you said that
1:13:33instagram for example would encourage
1:13:34eating disorders among teenage girls or
1:13:37that right-wing voters are pushed
1:13:38towards extreme right-wing groups via
1:13:41the algorithm without them actually even
1:13:43realizing it now my question's on
1:13:45algorithms the first
1:13:46one which i want to ask you is how we as
1:13:48lawmakers with limited knowledge and
1:13:50resources can get a grip on the
1:13:53algorithms of specialized tech giants
1:13:55how do we ensure that we always stay one
1:13:57step ahead especially because those
1:13:59algorithms will continue to evolve also
1:14:02second question
1:14:03concerns the input for algorithms
1:14:06because as i understand it facebook
1:14:08collects the data of its users on a
1:14:10large scale and then uses these same
1:14:11data to manipulate these users
1:14:13themselves
1:14:14cynical if you ask me it raises
1:14:16questions about privacy and about what a
1:14:18data company may use to build their
1:14:20algorithm so that's why i would like to
1:14:22know from you
1:14:23whether
1:14:24we should set limits on the data that
1:14:26facebook might collect on the data that
1:14:29facebook might collect and use as inputs
1:14:30for algorithms would a restriction in
1:14:33this way of our solution and if so what
1:14:36data should we keep out of the algorithm
1:14:38i look forward to your answers thank you
1:14:42thank you miss hogan please
1:14:46and with regard to this question of how
1:14:48can lawmakers stay one step ahead
1:14:50especially when these systems change so
1:14:51much like that's part of why having
1:14:53dynamic systems like the one two three
1:14:57safety audits is so important um because
1:14:59the reality is any fence you put up now
1:15:02like anything where we say you can't use
1:15:04these three things in reality we need to
1:15:06be able to continually evolve because
1:15:08these systems evolve so fast um i think
1:15:10that's a also a reason why facebook
1:15:13should have to disclose what data goes
1:15:15into things like their ad targeting
1:15:17algorithms or into how they rank the
1:15:18news feed right being able to disclose
1:15:20like these are the factors that were
1:15:22used to prioritize content is important
1:15:24i'm a strong proponent that if facebook
1:15:26has a fire hose like we discussed at
1:15:29least for for some fraction of the
1:15:30content um
1:15:32they should have to disclose what those
1:15:34scoring parameters were because if you
1:15:37allowed people like independent
1:15:38researchers and academics to have access
1:15:40to it they would be able to summarize
1:15:42for you what are the factors that are
1:15:43disproportionately amplifying the worst
1:15:45content
1:15:46and the reality is right now a very
1:15:48small number of people inside of
1:15:50facebook are asking that question
1:15:51because these things are very
1:15:53complicated right when facebook says
1:15:55that they're accurate um to give you a
1:15:58sense of how limited our resource for
1:16:00people who can even analyze these
1:16:01problems like i'm not i'm not trying to
1:16:03brag here i think there's probably on
1:16:05the order of 200 or maybe 300 people in
1:16:08the world who have the depth of
1:16:10awareness that i do around how these
1:16:12systems work and i think that alone is a
1:16:14national security problem like we we
1:16:16need to figure out ways where we can get
1:16:18more eyes on these systems because right
1:16:21now
1:16:21many of those people who fall in that
1:16:23category are alumni of facebook or the
1:16:26alumni of a place like google um and we
1:16:29have to figure out ways where they can
1:16:30continue to be in the conversation
1:16:32because part of why a lot of those
1:16:33people left was because they didn't get
1:16:35to work on keeping people safe enough
1:16:38with regard to um
1:16:40like inputs for these systems
1:16:43and i think as we have visibility into
1:16:46how they currently use the data we will
1:16:48realize data that we feel uncomfortable
1:16:50having in there um and that's part of
1:16:53why we need to involve not just
1:16:54facebook's own perception of its terms
1:16:57we need to allow community members say
1:16:59hey i feel really uncomfortable about
1:17:01whatever parameter is being used there
1:17:03because i guarantee you civil society
1:17:05groups will have very different opinions
1:17:06than facebook will
1:17:10thank you next i would like to ask
1:17:12natalie lazor to take the floor
1:19:34and
1:19:35with regard to does facebook does social
1:19:38media be just restarted from the
1:19:39beginning the reality is that facebook
1:19:41has many tools at its disposal which it
1:19:44has chosen not to use that are content
1:19:46agnostic it doesn't require us to pick
1:19:49good people or good ideas
1:19:51like these are uh
1:19:53these are systems that make this
1:19:54platform slower that give people more
1:19:57time to think before they re-share and
1:19:59that has the side effect of less
1:20:01misinformation that that those ideas
1:20:03that are when you need your share them
1:20:06uh they often but you wouldn't if you
1:20:07thought about a little bit that it slows
1:20:09those down um some examples are things
1:20:12like i talked about before of like
1:20:13having to click a link before you
1:20:14reshare it or cutting reshare chains
1:20:17um but there are other things like um
1:20:19[Music]
1:20:21and
1:20:23demoting content that is not from
1:20:25someone you know
1:20:26right even demoting a little because the
1:20:28worst things that spread misinformation
1:20:30are how facebook has pushed people into
1:20:32groups
1:20:34um that are large that act like viral
1:20:36variant factories you know a group that
1:20:38has a million members and produces a
1:20:40thousand pieces of content a day
1:20:42that is the pro the biggest problem
1:20:45because right now the algorithm picks
1:20:47three of those pieces of content and
1:20:49spreads them out to a million people but
1:20:51unfortunately the algorithm
1:20:53differentially picks they can it's
1:20:55biased towards polarizing divisive
1:20:57content
1:20:58so a lot of these things if facebook was
1:21:00required to do a risk assessment
1:21:03they had to listen to the community and
1:21:05they had to articulate how they were
1:21:07going to get better and what data would
1:21:08allow you to hold them accountable they
1:21:10actually were getting better
1:21:12they would likely fall back on some of
1:21:13these things because for a lot of these
1:21:16things they don't lose that much profit
1:21:18they lose slivers like less than one
1:21:20percent because it's true that people
1:21:22consume a little less content when it's
1:21:23safer
1:21:24but the platform is also a lot nicer
1:21:26they might actually be more profitable
1:21:28in five years or ten years because more
1:21:29people would still be on the platform a
1:21:31lot of people are quitting because it's
1:21:33toxic
1:21:34um but once you have that transparency
1:21:37once you have to articulate how we're
1:21:38going to deal with these problems it
1:21:40becomes people inside the company get
1:21:41the space to actually do the right
1:21:44things because right now there is lots
1:21:46and lots of weight pulled towards what's
1:21:48good for the shareholders and there's
1:21:50not a countervailing force that gives
1:21:52space for people who want to do the
1:21:54right thing to do the right thing
1:21:56with regard to having to publish what
1:21:59content is taken down i think that's a
1:22:00great idea because you would learn and
1:22:03you have to do it by type like was
1:22:04taking down and why because immediately
1:22:07you would learn that lots of stuff is
1:22:08not being taken down
1:22:10in a variety of languages you'd be like
1:22:12oh interesting you're not actually doing
1:22:14integrity actions in
1:22:16xyz language
1:22:17um so anything that involves linguistic
1:22:19transparency i think is vitally
1:22:20important because right now facebook is
1:22:22differentially harming
1:22:24on
1:22:26people who don't speak english and
1:22:27that's unacceptable
1:22:29the last question is like how do they
1:22:31distinguish between various forms of
1:22:33content they use ai systems that allow
1:22:35them to classify to label those those
1:22:37things
1:22:38and uh the quality is widely divergent
1:22:41between languages so again if the way
1:22:44that they are able to keep these
1:22:45platforms safe is with these ais they
1:22:48should have to publish enough data that
1:22:49we feel confident that they are
1:22:50investing they are being equitable in
1:22:52safety because right now they're
1:22:55investing in safety in the united states
1:22:56because they're afraid of being
1:22:57regulated but they're not investing in
1:22:59safety in places that they aren't afraid
1:23:03thank you next uh i would like to ask
1:23:05evan maitel to take the floor please
1:23:08thank you very much uh chair miss haugen
1:23:11thank you for your testimony here today
1:23:13and for being with us
1:23:15um i think
1:23:18if we take
1:23:20many messages from you i think one
1:23:23message you probably can take away with
1:23:25you is that we've heard you we hear you
1:23:27very well and um i think um it's so
1:23:30brave but also extremely helpful what
1:23:33you've been doing
1:23:34over the past couple of weeks together
1:23:37with many of my colleagues here i also
1:23:38work on the digital single market and
1:23:41the digital
1:23:43markets act the services act and markets
1:23:46act
1:23:46because i believe it's important to end
1:23:48this sort of digital anarchy that you
1:23:51also partially described uh today but
1:23:54today i'll speak on uh uh in my role as
1:23:57sitting in the special committee on ai
1:24:00because you yourself are an ai expert uh
1:24:03as well um and i think you've you know
1:24:06largely addressed that topic uh in in
1:24:09the remits of the dsa and the dma
1:24:12um but
1:24:14i think you also have heard that many
1:24:17have used your very uh helpful as i said
1:24:20and also brave whistleblowing to ask for
1:24:22complete fans of ai in various uh um
1:24:26places when it comes to advertising when
1:24:29it comes to recommender system or even
1:24:31systems meant to remove illegal um or
1:24:34dangerous content um
1:24:36but i believe and also what we've heard
1:24:38today you have worn against the misuse
1:24:40of ai particularly in the context of
1:24:43facebook and but i don't have the
1:24:45impression you're asking for complete
1:24:48bans
1:24:48and but
1:24:50rather as you've a number of times
1:24:52emphasize
1:24:53you're asking for more transparency to
1:24:55reduce ai usage to where it makes sense
1:24:59for more sensible government
1:25:02regulation and
1:25:03laws but also for the need for consumers
1:25:06to be able to make choices which i think
1:25:09is very important and i conquer on all
1:25:11of the statements um that you made
1:25:13because i simply don't believe that all
1:25:15the work you have been also doing on ai
1:25:17um you would you know uh
1:25:20think it's worthy of banning in one way
1:25:22or another
1:25:24entirely um but i also don't think that
1:25:26the the many young and energetic people
1:25:28that work in startups in in europe um
1:25:31with ai want to act natickly and
1:25:34necessarily
1:25:35and i think these are dii developers
1:25:38that could work on certain issues to
1:25:39find solutions and this is why i wanted
1:25:42to hear from you how do we actually
1:25:45improve the building of the usage of ai
1:25:48for the benefit of mankind how do we
1:25:51make sure that we get to these systems
1:25:53to be safe as you said
1:25:56and you've elaborated on a number of
1:25:58points but perhaps you could now go a
1:26:00little bit while uh into the ai itself
1:26:03and not just into the remits of the dsa
1:26:06or the dna
1:26:08if you ever
1:26:10miss hold on please
1:26:13um i want to be super clear i i don't
1:26:15advocate for getting rid of ai i'm more
1:26:18about that especially in a place like
1:26:20facebook where it is effectively our
1:26:23communication platform it will be the
1:26:25personal social media platform for the
1:26:27foreseeable future that we can only have
1:26:30ai if we have more transparency on ai
1:26:33because right now like for example after
1:26:35my congressional testimony and mark came
1:26:38out and said you're worried about
1:26:40extreme and polarizing content
1:26:42fine we'll take all political content
1:26:44out of the news feed
1:26:46if we don't know how they trained those
1:26:47systems like what what were the examples
1:26:49that were considered political what
1:26:51topics were considered political and we
1:26:53don't know how those are actually being
1:26:55used like seeing examples of how things
1:26:58were classified
1:27:00like that that can shape our democracies
1:27:02because they're choosing what ideas can
1:27:04they can't be distributed anymore um so
1:27:07i want to be clear i have a game for
1:27:08transparency not for getting rid of ai
1:27:11with regard to how we can make these
1:27:14systems more generally safer
1:27:17i think thinking about how how can we
1:27:20generalize
1:27:21requiring people to have some visibility
1:27:23into these systems
1:27:25um i understand the importance of having
1:27:28competitive secrets but at the same time
1:27:31there are lots of circumstances where
1:27:33people's lives are pretty dramatically
1:27:34affected by decisions that are made
1:27:37made by ais and if at a minimum
1:27:40academics and regulators aren't allowed
1:27:42data access
1:27:43um that that i think is unjust because
1:27:46the thing that we have seen over and
1:27:47over again is that these things are
1:27:49rarely equitable
1:27:51right that you know african americans in
1:27:53the united states are routinely highly
1:27:55disadvantaged by these systems we see
1:27:57that today at facebook where on things
1:27:59like reals
1:28:00like the ai was never taught racism but
1:28:03it learned racism by seeing how people
1:28:05responded to different kinds of content
1:28:07and it noticed oh african american
1:28:09content doesn't get as much engagement
1:28:11we don't need to show it
1:28:13right so these are the kinds of things
1:28:14where we need to have transparency or
1:28:16we'll have
1:28:17unjust systems
1:28:21thank you
1:28:22i would like to acknowledge that the uk
1:28:25mp mr damian collins chair of the uk
1:28:29parliament's joint committee for the
1:28:31online safety bill is also taking part
1:28:35in our hearing welcome
1:28:38mr damon collins
1:28:42and before starting with the second
1:28:45round of uh uh questions uh
1:28:47i proposed to to make a very short break
1:28:51it was asked for five minutes but uh for
1:28:54only three minutes
1:28:55so
1:28:56thank you
1:28:58break
1:36:09so dear colleagues to our guests it
1:36:12seems
1:36:13we can continue
1:36:16and we can start with our second round
1:36:18of
1:36:19questions and answers first i would like
1:36:22to ask
1:36:23pablo arias to take the floor
1:36:36today
1:36:41[Music]
1:37:31foreign
1:38:14online
1:39:33thank you pablo
1:39:36please
1:39:39i'll hit on something from the end of
1:39:40that question first so this question of
1:39:43you know algorithms are trade secrets
1:39:45and we have to protect them um any law
1:39:48that allows us to not disclose data in
1:39:51order to protect trade secrets is is
1:39:53going to fail right like facebook is
1:39:56going to do exactly what you described
1:39:57and they are going to wrap themselves in
1:39:59trade secrets and say you know we can't
1:40:01show any data on our algorithm because
1:40:02that will violate our trade secrets um
1:40:05so i think having some some
1:40:08discretion on that is an important thing
1:40:11but with regard to age verification um i
1:40:14would like to introduce an idea to you
1:40:16too that may not be obvious which is
1:40:17there is a difference between a stated
1:40:19age and an estimated age
1:40:22so when we shared documents with
1:40:24congress one of the staffers came back
1:40:26and said
1:40:28does this chart mean 10 to 20 of 10 year
1:40:31olds are on facebook um and it was a
1:40:34cohort chart it looked at each birth
1:40:36year
1:40:37and it using and i looked at i was very
1:40:40confused initially i was like how could
1:40:42they know this and then it clicked for
1:40:43me facebook had gone
1:40:46and looked at each of those users
1:40:47friends they looked at their interests
1:40:50various things and they were able you're
1:40:52able to with great precision guess what
1:40:54someone's ages
1:40:55based on because most most people don't
1:40:57lie about their age enough kids wait to
1:40:59join the platform that you can get a
1:41:01good estimate of what the actual age of
1:41:03people is
1:41:04facebook should have to publish
1:41:06what fraction of people under were under
1:41:0913
1:41:10what at what age do they come on the
1:41:12platform for each cohort so for every
1:41:15year when kids turn 13 14 facebook can
1:41:17now publish a backwards state and say
1:41:19what fraction of those people of 14 year
1:41:21olds on the platform what fraction of
1:41:23them were on the platform at 10 11 12
1:41:25because that will allow us to know did
1:41:28they do a good enough job of keeping
1:41:30kids off the platform today we don't
1:41:32have any yardstick we have to just trust
1:41:34them and so at a minimum i believe they
1:41:36should have to disclose that estimated
1:41:38age data
1:41:39the second question is around how can we
1:41:42verify kids ages without having to give
1:41:44up anonymity so anything that requires
1:41:47us to provide identification i think is
1:41:49a solution that will fail
1:41:51because the reality is most places in
1:41:53the world do not have highly databased
1:41:56identification
1:41:58and people as long as they can have a
1:42:00vpn will go into those other go into
1:42:03those other countries and register
1:42:04accounts because facebook's not going to
1:42:06give up on growth from you know a large
1:42:09swath of the world
1:42:10and so if you have any system that is
1:42:12around you have to verify identity i
1:42:15think people will just go around it
1:42:17um
1:42:19because facebook is also not going to
1:42:21you know fence in those users and say
1:42:23you come from africa therefore you can't
1:42:24talk to people in europe right i i don't
1:42:26think will happen i do think though
1:42:28there are solutions like um
1:42:30you can there are now service providers
1:42:32that you could do this even on a device
1:42:34it doesn't even have to go to facebook
1:42:36where uh taking a picture of someone
1:42:38they can estimate your age for kids
1:42:40within about half a year
1:42:42right and so facebook has lots of tools
1:42:44it could be using
1:42:46they should have to publish data that
1:42:47confirms that you're actually doing
1:42:49enough
1:42:50um and they should have to also disclose
1:42:52what are they doing because then
1:42:54academics independent researchers can
1:42:55say here are all the other tools that
1:42:57you might want to use
1:43:00thank you next uh i would like to invite
1:43:02developer to take the floor picture
1:43:05yeah thank you sweetie
1:43:16[Music]
1:43:22foreign
1:43:43[Music]
1:45:07[Music]
1:45:33miss hogan please
1:45:45um to be really clear what i said
1:45:47earlier was that i did not expect
1:45:50smaller platforms to emerge
1:45:53because there are lock locked in network
1:45:56effects
1:45:57and that is why the few countries that
1:45:59did have their own systems
1:46:01so for personal social media creators
1:46:03creates they want to reach specific
1:46:05people
1:46:06and so the only way you can have
1:46:07competition is if you could bring that
1:46:09whole network with you wherever you want
1:46:15interoperability i think opens up far
1:46:18bigger problems than we currently have
1:46:20today and would require a much deeper
1:46:23level of regulation i'll give you an
1:46:24example let's all imagine we're email
1:46:27servers so email has inter
1:46:29interoperability
1:46:30but if i want to send an email to
1:46:32someone on your server i don't have to
1:46:35broadcast it out to every email server
1:46:37in the world it just goes to one server
1:46:39and we have a contract that is easy to
1:46:41enforce
1:46:42once an email leaves my server and goes
1:46:44to your server i can never call it back
1:46:46i've committed
1:46:48we have certain expectations with social
1:46:49media we're like let's imagine my
1:46:52friends scatter across 50 social
1:46:54networks so there's facebook there's new
1:46:57book there's new book two new book three
1:47:00if we have interoperability because uh
1:47:03we have an expectation that our social
1:47:05media loads really quickly right if
1:47:07things don't happen almost
1:47:08instantaneously people people leave
1:47:11um
1:47:12to have interoperability when i post a
1:47:15picture of my baby on
1:47:17new book
1:47:18it will get broadcast out to the 50
1:47:20other platforms that my friends are
1:47:22scattered across and that sounds fine
1:47:24you're like oh good now we can all have
1:47:26our own rankings our own filterings but
1:47:28it becomes a problem if i ever want to
1:47:29take that content down
1:47:31so let's say tears go by and i no longer
1:47:33want pictures of my children on social
1:47:35media
1:47:36i have to trust that when my new book
1:47:40contacts those 50 platforms and says you
1:47:41need to delete that content that it's
1:47:43actually deleted
1:47:44we would have to do a very invasive
1:47:47level of regulation of all the facebook
1:47:49clones to ensure that they actually
1:47:51complied with how they were operating
1:47:53how do we know that that server wasn't
1:47:56built by russia right they built a
1:47:58social network they funded it because
1:48:00they were never going to delete the
1:48:01content
1:48:02right there's a lot of these dangers
1:48:04where it may not be obvious why
1:48:05interoperability is is is probably much
1:48:08more of a mixed bag and much harder to
1:48:09do in a way that is actually functional
1:48:11that's part why i advocate for the one
1:48:13two three approach on risk assessments
1:48:15and getting transparent access to data
1:48:19with regard to excessively personal
1:48:21targeted information i think there's
1:48:22interesting conversations to be had on
1:48:24how to simplify those models to try to
1:48:27avoid as many rabbit holes pushing
1:48:29people towards them
1:48:31and with regard to acquisitions i
1:48:33encourage you to think more broadly than
1:48:35just the obvious other social platforms
1:48:38with regard to banning acquisitions by
1:48:40facebook one of the most important
1:48:42startups in terms of holding facebook
1:48:44accountable was crowdtangle
1:48:47so crowdtangle had a way of of
1:48:49monitoring what happened on facebook and
1:48:51facebook bought it
1:48:53and when
1:48:54researchers started to notice
1:48:56how toxic facebook was by using
1:48:58crowdtangle facebook shuttered that team
1:49:02like it still exists but they've stopped
1:49:03developing it and they likely will close
1:49:06it in the future so when you're thinking
1:49:08about what acquisitions to not allow to
1:49:10protect competitiveness or protect
1:49:12accountability
1:49:13you need to also look at should facebook
1:49:15be allowed to buy startups that hold it
1:49:17accountable
1:49:18because if they buy it they can kill it
1:49:21and then the last question on data
1:49:23access and how it can be used i strongly
1:49:25i'll repeat again i strongly encourage
1:49:27you to include the independent
1:49:28researcher community you can have an
1:49:30accreditation process that's totally
1:49:32fine but i encourage you to have some
1:49:34avenue for participation because we need
1:49:36to build the ecosystem of accountability
1:49:39because remember you know let's say
1:49:41someone like me
1:49:43isn't interested in going to a
1:49:44university but we're really interested
1:49:45in running a youtube channel
1:49:47that could actually inform consumers and
1:49:49give them choices by translating this
1:49:51complex information into something they
1:49:53could consume so i encourage you to be
1:49:55inclusive in who gets data
1:50:00thank you
1:50:02next i would like to ask alicia
1:50:06and among her to take the floor
1:50:09you very much chair and thank you mrs
1:50:12hogan um for coming here and sharing
1:50:15your views with us
1:50:16your revelations have confirmed my
1:50:18belief that we've let online platforms
1:50:20make up their own rules and regulations
1:50:22for too long and with your call for more
1:50:25transparency and democratic oversights
1:50:27you are really preaching to the choir in
1:50:29this house
1:50:30and i would also like to address the
1:50:32youngest users of social media platforms
1:50:34because we know that one in three
1:50:36internet users is below the age of 18.
1:50:39uh we also know that the internet was
1:50:41never designed with children in mind and
1:50:43while platforms have opened up many
1:50:45opportunities for children to learn to
1:50:47communicate to develop
1:50:48their design more often than not still
1:50:51goes after the data of these youngest
1:50:53users so children are now leaving
1:50:55digital footprints behind many times the
1:50:57size of their own shoe size
1:50:59and besides this use you have also worn
1:51:02for harmful content that is served up to
1:51:04these miners
1:51:07and i have listened to your answer on
1:51:09the age verification very clearly and
1:51:11thank you for this i was very interested
1:51:13in this but i'm also very interested to
1:51:15hear how you think that we should really
1:51:17secure the safety and rights of our
1:51:19children online and whether you have
1:51:21concrete proposals to improve platforms
1:51:23services for children
1:51:26and notably in the data processing
1:51:29and because you have had the chance to
1:51:31look at the research that facebook has
1:51:33been done
1:51:34has done do you think that we can really
1:51:36cater to children and what they really
1:51:39understand that they are giving up
1:51:41uh
1:51:42in
1:51:43trading for uh for free games for
1:51:45example do they know what they are doing
1:51:48and can we
1:51:49somehow make this clear to children so
1:51:52thank you
1:51:54thank you alicia
1:51:56miss hogan please
1:51:58i think it's very important for us to
1:52:01acknowledge
1:52:02the
1:52:03uh
1:52:04the network of influences that
1:52:06a child faces when they use things like
1:52:09instagram so because it is the default
1:52:12platform for people under 18
1:52:15people
1:52:16feel a fear like facebook's own research
1:52:19says you know kids say
1:52:20i know it makes me unhappy i know i i i
1:52:24don't want to use it as much as i do i
1:52:26can't stop myself and the most important
1:52:29one and i fear being ostracized if i
1:52:31leave right that if i leave in isolation
1:52:34if i make a choice for myself i will be
1:52:36cut off from my friends and family
1:52:38people are willing to do some pretty
1:52:40self-harming things if it means they can
1:52:43keep their friends right and we see this
1:52:45with teenagers every day peer pressure
1:52:47is a real thing
1:52:48um i think it's important when we think
1:52:50about how to secure the rights of young
1:52:52people that we keep that in mind that we
1:52:54really need systems level solutions not
1:52:58you know blaming parents because kids
1:53:00use social media
1:53:02i think one of the most important ways
1:53:03that we can help protect the rights of
1:53:05young users is by giving parents better
1:53:07resources
1:53:08so one of the things from facebook's
1:53:10research is that kids
1:53:12because their parents did not come of
1:53:14age during a time when instagram was
1:53:16ubiquitous
1:53:18that you find that parents often give
1:53:20really harmful advice to their kids or
1:53:23advice that makes kids feel alone
1:53:25right so parents will uh
1:53:27just say like why don't you just not use
1:53:29it or they might get angry at their
1:53:31child because their child can't can't
1:53:33stop using it and so i think there's a
1:53:35lot of value in trying to get better
1:53:37information to the hands of parents so
1:53:38that
1:53:39if we can't get the platform safer at
1:53:41least parents can try to help their kids
1:53:43more
1:53:44um
1:53:45because right now we're expecting
1:53:46parents to know about neurobiology and
1:53:48that's unaccept that's unreasonable
1:53:52with regard to can children understand
1:53:54what they're trading away um i think
1:53:56almost certainly kids like we don't we
1:53:59as adults we don't know what we've
1:54:00traded away right we don't know what
1:54:02goes in the algorithms we don't know how
1:54:04we're targeted so the idea that children
1:54:06can be give informed consent i i like i
1:54:09don't think we give informed consent and
1:54:11they have less capability
1:54:13um
1:54:15i think there's a real need like in
1:54:17terms of research to be funded um
1:54:19one of the things that i'm i'm most
1:54:21intrigued by is there's a misinformation
1:54:24researcher out of singapore and she's an
1:54:26anthropologist and she had some of the
1:54:28most interesting observations on how
1:54:30misinformation was being conducted that
1:54:33we would have never been able to
1:54:34discover on our own and the way she
1:54:35found it was by just interviewing kids
1:54:38i think funding more research where we
1:54:40ask kids about their lived experience of
1:54:42social media is really important because
1:54:44right now facebook is going to be
1:54:46discouraged from doing research like
1:54:48that because of my disclosure right
1:54:51though i don't take responsibility for
1:54:53it because their cowardice is not my
1:54:55fault
1:54:56um and so yeah investing more in
1:54:59listening to children i think is really
1:55:00valuable if we want to protect their
1:55:02rights
1:55:04thank you uh next i would like to ask
1:55:07deirdre cloon to take the floor
1:55:10thank you very much and thank you miss
1:55:11hagan for being with us today um and i
1:55:13hope this is the start of mo we'll have
1:55:15other opportunities as well so thank you
1:55:17and and i would i'd like one of the
1:55:20points i wanted to raise actually was
1:55:22minors and i think the comments you've
1:55:24made um really
1:55:26well will highlight the fears that so
1:55:28many parents have
1:55:30of how their children can be and their
1:55:31young people can be manipulated um
1:55:34online and i wanted to ask you if you
1:55:36were
1:55:37mr miss facebook making decisions what
1:55:39would you do how could you change it
1:55:41just give us an insight into how
1:55:42facebook approach um or they think about
1:55:45their approach approach to minors uh
1:55:48that'd be most interesting and actually
1:55:50your comments about the dsa i mean um i
1:55:52think are very encouraging for us i know
1:55:54and just looking at the words we have
1:55:56describing us transparency oversight
1:55:59accountability
1:56:00risk assessment is there uh access for
1:56:02researchers even though i i take your
1:56:04point that maybe there's so many more
1:56:06out there so much more expertise out
1:56:07there that we could be availing of asp
1:56:09to to um to data i think that's all that
1:56:12so that's very very encouraging i fear
1:56:15that you know
1:56:16we they'll always be one step ahead of
1:56:18us and so how can we future proof our
1:56:21legislation and
1:56:22bearing in mind it's not just about
1:56:24facebook it's about other other
1:56:25platforms as well
1:56:27i think that that would be important for
1:56:29us because we have a lot of key
1:56:31parameters are there it's just really if
1:56:33we can we never we i know we never will
1:56:36i would always be behind but really to
1:56:38future-proof i think would be very
1:56:40important for us and just a small
1:56:42another point i could ask if facebook
1:56:44changed its algorithms in 2018 to
1:56:46emphasize um
1:56:48meaningful social interactions and to
1:56:50give more weight to uh re-sharing and to
1:56:53friends
1:56:54did that work or what happened to that
1:56:56just
1:56:57did did it work i think i don't think it
1:56:59probably didn't but just um why was that
1:57:01decision taken 2018 and did it work so
1:57:04thank you
1:57:06thank you miss hogan please
1:57:09with regard to how can we future proof
1:57:11the dsa
1:57:13i believe there's a lot of power in
1:57:14having a one two three approach so that
1:57:17means they can disclose risks the
1:57:19community identifies risks because
1:57:21listening to different groups is going
1:57:22to give us more things that we need to
1:57:24flag and facebook should propose
1:57:26solutions but
1:57:28we should hold them accountable those
1:57:29are rigorous solutions and they should
1:57:31disclose data that would allow us to
1:57:33track progress on those solutions if you
1:57:35don't have all three parts i don't think
1:57:37it'll be effective but if you do have
1:57:39all three parts developing that cycle
1:57:42where data is continuously disclosed
1:57:44that's something we need that i think i
1:57:46encourage you to modify the dsa if they
1:57:48pull data for you once
1:57:50it's easy for them to pull it every week
1:57:53because they have already written the
1:57:54queries to pull that information so if
1:57:56they tell you it's too onerous to do it
1:57:58twice a year guess what giving you a
1:58:00stream where they just rerun the same
1:58:01queries every week it's it's very little
1:58:04additional work for them so don't
1:58:06i encourage you to be questionable in
1:58:07them
1:58:08that process of having
1:58:10building an ecosystem of contributors
1:58:13having data that regularly comes out so
1:58:15we can monitor when step changes happen
1:58:17because i i saw over and over again at
1:58:19facebook that they would have
1:58:20regressions that a feature would launch
1:58:23that no one could have anticipated was
1:58:24going to greatly amplify misinformation
1:58:27but i guarantee you if 10 000 people
1:58:28were watching those streams
1:58:31they would notice that something had
1:58:32changed and so it is important to build
1:58:34the ecosystem
1:58:35so the next question is around msi so
1:58:37msi was done specifically for business
1:58:39interests so they claimed they wanted to
1:58:41increase meaningful social interactions
1:58:44but those interactions could have been
1:58:45hate speech it could have been bullying
1:58:47today they only fixed that part for
1:58:49english in november 2020 right if you
1:58:52did bullying or hate speech in the
1:58:53comments they still count it as a
1:58:55meaningful social interaction and for
1:58:57the vast majority of languages in the
1:58:58world today and i guarantee you a
1:59:00majority of european languages you can
1:59:02do hate speech in the comments today and
1:59:03still look considered a meaningful
1:59:05social interaction the reason why
1:59:06facebook did that change was they had
1:59:08tried many many things to get people to
1:59:11produce more content because people were
1:59:13starting to produce less and less
1:59:14because it was so toxic and what they
1:59:16found was the only thing that worked
1:59:18that had the least side effects was
1:59:20giving you more little hits of dopamine
1:59:23that if you got more likes more comments
1:59:25more research you produced more content
1:59:28and they ran experiments on creators to
1:59:31figure this out
1:59:33when they went and did surveys after the
1:59:35fact so six months down the line they
1:59:37found that people said their feeds were
1:59:39less meaningful after they implemented
1:59:41meaningful social interactions
1:59:43because like we have talked about
1:59:44repeatedly during today's session
1:59:47when you prioritize for content that can
1:59:49provoke a reaction
1:59:51you
1:59:52end up biasing towards extreme
1:59:53polarizing divisive content
1:59:56we have seen political parties picked up
1:59:58on this publishers picked up on this
2:00:00like facebook zone research says that
2:00:01when you use engagement based ranking
2:00:03like this you end up seeing that
2:00:07the more toxic your comment thread is
2:00:09like the more negative the common thread
2:00:11is the more likely they click out to
2:00:12your site
2:00:14and i guarantee you publishers notice
2:00:15this
2:00:16right uh buzzfeed which is a large large
2:00:18outline in the united states they wrote
2:00:20to facebook and said the content that
2:00:22we're the least proud of is the content
2:00:24that gets the most distribution on
2:00:25facebook you clearly have a problem
2:00:28but because they needed it to get people
2:00:30to keep creating content that's why they
2:00:32kept captain in place
2:00:36thank you
2:00:37next i would like to ask kim von
2:00:39sparanthak to take the floor king
2:00:42the floor thank you very much and also
2:00:44thank you to madame hogan for being here
2:00:46today
2:00:48i think you've explained it very clearly
2:00:50for years facebook has profited from the
2:00:52large-scale spread of this information
2:00:54and hate and what we see online is
2:00:56driven by clicks interaction and profit
2:00:58and this means recommender algorithms
2:01:00are programmed to show people content
2:01:02that sparks anger hate and even sends
2:01:04people down rabbit holes of
2:01:05radicalization
2:01:07facebook knows the impact of their
2:01:08algorithm but refuses to do anything
2:01:10about it and this only reaffirms what we
2:01:12already know it's time for strong rules
2:01:16my group the greens and i fully agree
2:01:17with you that to hold companies
2:01:19accountable we have to know what is
2:01:21actually going on
2:01:23we need to understand in detail how
2:01:24these algorithmic systems work how they
2:01:27are optimized and how they are
2:01:29influenced
2:01:30only then we can take the power over
2:01:31what happens online out of the hands of
2:01:33private companies and place it firmly
2:01:35back in the hands of the people
2:01:37but tech companies and some right-wing
2:01:39politicians framed such transparency as
2:01:41an unreasonable burden for companies and
2:01:45you've already also touched upon this
2:01:46trade secrets issue
2:01:48so my first question is could you
2:01:50explain to us how feasible it is to have
2:01:52transparency of recommender algorithms
2:01:54and reporting on content moderation for
2:01:57companies
2:01:58and yeah knowing what happens is a
2:02:00crucial first step but holding platforms
2:02:02accountable and giving users more
2:02:03control is the next step transparency
2:02:06alone will not be enough if people
2:02:07cannot act upon this information and if
2:02:09nobody holds platforms accountable so my
2:02:12second question to you is how do you
2:02:14think we can empower people and give the
2:02:16public more control
2:02:18over what they see online and which role
2:02:20could independent oversight researchers
2:02:23and civil society play as a part of the
2:02:25solution thank you
2:02:29thank you ms hogan please
2:02:31with regard to how feasible is it for
2:02:33them to provide data
2:02:35i think it's important for us to
2:02:36differentiate between the first time
2:02:38they do it and the ongoing times it is
2:02:40likely that for the foreseeable future
2:02:43many of the things that are flagged via
2:02:45the risk assessment will continue to be
2:02:46problems right and so yes the first time
2:02:50they may need to write that the way we
2:02:52get data out of these systems is via a
2:02:54query
2:02:55right we query the system
2:02:57they will have to write many queries
2:02:58likely the first time though many of
2:03:00them will be things they already have
2:03:02they will already run this analysis
2:03:04somewhere and they have at least
2:03:05facebook has a search engine that lets
2:03:06them find those queries
2:03:08um
2:03:11each additional time it is likely that
2:03:13they won't have to write as many queries
2:03:16because many of the problems are likely
2:03:17going to stay the same time to time we
2:03:19might need to ask slightly different
2:03:20questions of what information we would
2:03:21need to know to get progress because
2:03:24likely once that data is published once
2:03:26those streams are released every week
2:03:28ongoing basis that people will be going
2:03:30to say oh interesting actually i wanted
2:03:31it a little bit different and the next
2:03:33time around you can tweak it
2:03:35um
2:03:38we are we're not asking for something
2:03:39unreasonable this is something that
2:03:41facebook can totally provide and i think
2:03:43the thing that is exciting to me is i
2:03:45think actually having more transparency
2:03:47on data would likely bring more
2:03:49employees to facebook
2:03:50when i worked at yelp
2:03:52um i was really sh
2:03:54struck by how did yelp staff itself so
2:03:57yelp is one of the only companies that
2:03:59releases a real user data set it shows
2:04:02you reviews from i think it's like 15
2:04:04college towns in the united states
2:04:06and it's used across the united states
2:04:09across the world
2:04:10as a data set to teach people machine
2:04:12learning to teach them how to analyze
2:04:14language computationally and a huge
2:04:16fraction of yelp's employees said the
2:04:19reason i decided to work at yelp
2:04:21why i did an internship at yelp was
2:04:23because i took a class where i used
2:04:25yelp's data and i was so intrigued by
2:04:28what i found that i chose to work here
2:04:31i think that one of the top challenges
2:04:32for facebook is that people don't want
2:04:34to work there right now they are in a
2:04:36feedback cycle where because they've cut
2:04:38corners they have scandals which makes
2:04:41it harder to hire which makes them
2:04:43cut more corners which causes more
2:04:44scandals
2:04:45this is just getting worse and
2:04:47unfortunately what i my disclosure will
2:04:49likely make it worse too
2:04:50i think there's a real thing that if you
2:04:52know more young people were exposed to
2:04:54these data streams in college classes
2:04:56and i guarantee you college classes will
2:04:58be taught about these data streams
2:05:00especially if they come out weekly
2:05:02monthly whatever is the appropriate
2:05:04frequency to still protect privacy
2:05:06and i think it might light a generation
2:05:08on wanting to work on facebook i think
2:05:10there's a real opportunity there
2:05:12with regard to improving control control
2:05:15is really hard in isolation so if we
2:05:18just let people opt into having a
2:05:20different ranking there's major
2:05:21limitations with that because for
2:05:23example i often endorse what i call
2:05:25chronological plus plus so ordering by
2:05:28time but having intelligible
2:05:30modifications like if you post a lot if
2:05:32you spam demoting people that's the
2:05:34thing that we can
2:05:35can actually understand by looking at as
2:05:37humans um if you just do that change but
2:05:41you don't change any of the other things
2:05:42about the product
2:05:43it won't work right because facebook has
2:05:46pushed people into million person groups
2:05:48and they will get spammed by the group
2:05:50right it's about changing features in
2:05:52concert and i think when you have a risk
2:05:55assessment that's like a one two three
2:05:57style
2:05:58facebook will have to think more
2:05:59holistically about how to solve these
2:06:00problems i also think it's
2:06:03dangerous to expect individuals to
2:06:06understand these issues enough when
2:06:07they're so complicated that they would
2:06:09adequately be able to protect themselves
2:06:11if you just have a thing where if people
2:06:12choose between two different rankings
2:06:15and like and one's a lot more fun
2:06:17like people will opt for
2:06:19the hyper-optimized one that has all
2:06:21facebook has all the data and all the
2:06:22control on
2:06:24and so i don't think users will get
2:06:26the product that they deserve
2:06:29if we just expect you know the market to
2:06:31solve for it
2:06:34thank you next i would like to ask marco
2:06:36camponosi to take the floor but i have a
2:06:39bad news
2:06:40so it seems that we will not have time
2:06:43for cure for catch the eye the session
2:06:47if those questions will be shorter and
2:06:49answers will be also shorter then
2:06:53maybe but now it seems that no catch in
2:06:56the eye
2:06:56so
2:06:57marco please
2:07:01thank you very much thank you to be here
2:07:03[Music]
2:07:05i think that the most important question
2:07:07is also
2:07:08who is
2:07:10entitled to define what is a fake news
2:07:13or when we are in front of a violation
2:07:15and i'm convinced that the public
2:07:17authority has to take this role
2:07:20uh
2:07:21but it's important also that you know
2:07:23which is the situation in europe because
2:07:27for fiscal reasons the major
2:07:31platforms are all based in ireland
2:07:33and the irish authorities cannot
2:07:37give an answer to all the procedures
2:07:39started 99 percent of the appeals are
2:07:43not prosecuted
2:07:45and the eu member states national
2:07:47authorities have to find back doors in
2:07:50order to solve this question
2:07:52the italian authority on data protection
2:07:56had to intervene on tick-tock regarding
2:07:59an important issue which was uh the the
2:08:02way
2:08:03uh of the consent given by children with
2:08:07less than 14 years old and they had to
2:08:10find the so-called utensil procedure
2:08:12given by article 66 of a gdpr regulation
2:08:16a german authority did the same on
2:08:18another topic and also the belgian
2:08:20authority has been authorized in order
2:08:22to solve the question because if not it
2:08:23was not possible to do elsewhere
2:08:27of course i told you i i think public
2:08:29authorities had to intervene but of
2:08:31course we have also to intervene on on
2:08:33the platforms
2:08:35we have on one side the risk in having
2:08:38an intervention which is too strict
2:08:42there is to have on one side over
2:08:44blocking
2:08:45the creation of filter bubbles
2:08:48or the
2:08:49complete cut of a political debate
2:08:53facebook does not
2:08:54want to anymore to have a political
2:08:56debate it's not business for them on the
2:08:59other side stricter rules could
2:09:02could be
2:09:03and become a barrier
2:09:05for the possibility for startups to
2:09:08enter in this special market
2:09:11and this is also as legislator something
2:09:13on which we should intervene and i hope
2:09:16that this will be done i don't know if
2:09:17you have an opinion on this topic thank
2:09:19you very much
2:09:21thank you miss hogan please
2:09:26i'm a strong proponent for having tiered
2:09:29regulations
2:09:31because you're right like startups have
2:09:32less resources to execute
2:09:34in systems like this but in a world
2:09:37where facebook will be the only form of
2:09:39personal media like large-scale personal
2:09:41media social media for the foreseeable
2:09:43future
2:09:44i think having exemptions or like having
2:09:47higher standards for facebook or like in
2:09:49the case of google for advertising i
2:09:51think that's appropriate
2:09:53so having tiers for regulations i think
2:09:55is really valuable for that reason
2:09:57with regard to the challenges of having
2:10:00enforcement be all
2:10:02giving ireland the burden of having to
2:10:04enforce all these rules because you're
2:10:06right most the tech companies are out of
2:10:07ireland um i think there's a real real
2:10:10need
2:10:12for there to be some kind of centralized
2:10:14authority in europe because like i said
2:10:17before like if there's only maybe 200 or
2:10:19300 people in the industry who have
2:10:22enough
2:10:23experience and insight around how these
2:10:25systems work and what the consequences
2:10:26of them are
2:10:28if we expect to
2:10:31spread them across 27 agencies
2:10:34um i think that's it's going to be very
2:10:36ineffective like it's not you're not
2:10:37going to get the critical mass in any
2:10:39one place
2:10:41i think there is also that conflict of
2:10:42interest factor that ireland benefits a
2:10:45lot from having those tech companies
2:10:46there and i think it puts them under
2:10:49unfair pressure to have to uh
2:10:51hold those things in tension because i'm
2:10:53sure ireland cares about the safety of
2:10:55our children ireland cares about our
2:10:57democracies being threatened but they
2:10:59also have
2:11:01extreme pressures being exerted against
2:11:03them which shows the need for a central
2:11:05central oversight
2:11:08thank you uh next i would like to ask
2:11:11maria manuel
2:11:12marquez to take the floor please
2:11:15thank you thank you francis for what you
2:11:18have done and for being here today with
2:11:21us and let me go straight to my
2:11:23questions
2:11:24facebook papers prove facebook now meta
2:11:28knows its causing harm and choose not to
2:11:32act
2:11:33therefore we must have art regulation to
2:11:36guide innovation away from hurtful
2:11:39results as you are doing in europe last
2:11:42week i was in washington having meetings
2:11:45with the congress and the white walls
2:11:48and i felt a large gap between the us
2:11:52and eu approach to regulation in your
2:11:56view how should we bridge this gap and
2:11:59create a democratic model to the digital
2:12:03ecosystem who has to move closer to the
2:12:07other
2:12:08i also want to talk about business model
2:12:12we can't expect different outcomes
2:12:14without change to the business model
2:12:17as zuckerberg said senator we sell x so
2:12:21it's not surprised that i'm referring to
2:12:24ads and micro targeting and behavioral
2:12:27advertising in particular
2:12:29social media cells as an utopia given
2:12:32enough diet it will perfectly pared
2:12:35predict the behavior of users and match
2:12:38it with the products being sold to them
2:12:41to the benefit of users and business
2:12:44alike
2:12:45however
2:12:47what we have is a dystopia
2:12:50not only there is little evidence of
2:12:53benefits for businesses
2:12:55but we ended up with an industry of
2:12:59surveillance and mass data collection
2:13:03that gives rise to vulnerabilities for
2:13:06individuals groups and even political
2:13:09systems
2:13:10you say
2:13:11you
2:13:13would like to give people choices i
2:13:16agree but it's more subtle than that
2:13:19first i don't think people really are
2:13:22able to consent to write a collection
2:13:25and use it
2:13:27and use with full knowledge of what that
2:13:31means
2:13:31what data who is collecting how they are
2:13:35using it and second even if i choose to
2:13:40share my own data
2:13:42that will contain information about the
2:13:45people i interact with
2:13:47so i shouldn't
2:13:49it shouldn't really be only my choice
2:13:53some of hazarian parliament think we
2:13:55need to put a stop to micro targeting
2:13:59and behaviorally advertising and the
2:14:02data collection that it implies
2:14:05i would like very much to hearing from
2:14:07you about this proposal thank you very
2:14:10much
2:14:11thank you miss hogan please
2:14:16i think there's a large opportunity to
2:14:18bring the united states and the european
2:14:21union closer together um in that you
2:14:24know what i've said repeatedly is that i
2:14:27don't support having a regulator that
2:14:29comes in and says we know how to do
2:14:30social media better than facebook but i
2:14:33do believe that consumers have the right
2:14:35to know what they're actually using
2:14:38and the thing about risk assessments
2:14:39especially risk assessments that are
2:14:41publicly published and having data where
2:14:44we can independently verify that they're
2:14:46actually making progress is it allows
2:14:48people to make choices on the products
2:14:50they use
2:14:51that's really really important and i
2:14:53think that idea the idea that the
2:14:55regulator's job is to make sure that
2:14:57facebook's telling the truth
2:14:59and like has a process where they have
2:15:00to they actually have to reflect on
2:15:02what's happening and you know the thing
2:15:04that's great is if facebook doesn't
2:15:06exhibit progress then people actually
2:15:08get a choice on whether or not they want
2:15:10to keep using it or they can put other
2:15:11forms of pressure on facebook facebook
2:15:14has hidden this data because they know
2:15:16it's bad
2:15:17right they know that if more people knew
2:15:19it was actually happening on the
2:15:20platform they would put more pressure to
2:15:22change facebook and i think that kind of
2:15:24lens where you say hey this is not so
2:15:26much about a regulator telling you what
2:15:28to do and more about you should just
2:15:30have to be honest about what you're
2:15:31doing and we're going to make sure we
2:15:32have a process where every six months
2:15:34you have to come clean i think that's a
2:15:36thing that a lot of people in the united
2:15:37states could sign on for
2:15:39right um
2:15:40because you know there's more of a bias
2:15:42in the united states of like at least
2:15:44you know you can't have a free market if
2:15:46people don't know what they're buying we
2:15:47have for example nutritional policies
2:15:49where people have to publish the
2:15:51ingredients on their food it's the same
2:15:53thing we're just asking for the
2:15:54information ingredients in this case
2:15:56with regard to choices um
2:15:59you know in the case of
2:16:01tracking on apple products so apple
2:16:03recently forced um advertise or forced
2:16:06apps to disclose like do they do
2:16:08tracking and make people opt into it
2:16:10it's something like 80 percent of people
2:16:11choose not to opt in
2:16:13so the question is around making sure we
2:16:14don't have dark patterns that force
2:16:16people into systems and i do agree with
2:16:18you on this question of like often my
2:16:20actions might bleed information about my
2:16:23friends
2:16:24but i think if we required
2:16:26platforms like facebook to disclose what
2:16:28goes into their ad targeting then we
2:16:30would know whether or not they're using
2:16:32factors that would cause that kind of
2:16:33bleed over
2:16:35um with regard to micro targeting
2:16:37i think there are specific
2:16:39things in ads that really need to be
2:16:41regulated so for example given the
2:16:44current system subsidizes hate you know
2:16:46it's five to ten times cheaper to run a
2:16:48political ad that's hateful than a
2:16:50non-hateful ad i think you need to have
2:16:52flat rates for ads i think that's
2:16:54definitely a thing
2:16:55but i also think there should be
2:16:56regulation on targeting ads to specific
2:16:59people so i don't know if you're aware
2:17:01of this but right now you can target
2:17:03specific ads to an audience of 100
2:17:05people
2:17:06and i'm pretty sure this is being
2:17:08misused because i did analysis on who is
2:17:11hyper exposed to political ads
2:17:13and unsurprisingly the people who are
2:17:15most exposed are in washington dc
2:17:17and they are radically overexposed we're
2:17:20talking thousands of political ads a
2:17:22month
2:17:22and so i do think having mechanisms to
2:17:24target specific people
2:17:27without their knowledge
2:17:28i think is is unacceptable
2:17:31thank you next speaker is nicolas peksa
2:17:36nicholas
2:17:39thank you
2:17:40miss hogan uh first let me thank you for
2:17:43you coming here today and for your
2:17:45bravery in standing up against the
2:17:47practices that encourage hate and
2:17:49division in our society
2:17:50however we are not here only to listen
2:17:52and try to understand what went wrong in
2:17:54one of the most influential social media
2:17:56companies in the world
2:17:58we are also here to find a way forward
2:18:01so i would like to hear your opinion on
2:18:03the following questions uh first in your
2:18:05view what can the european union do to
2:18:08better protect legally with the blowers
2:18:11especially in big multinational
2:18:13companies like facebook is
2:18:16secondly
2:18:17the current proposal for the digital
2:18:18services act as i put forward in the
2:18:21economic and monetary affairs committee
2:18:23such as the direct commander system
2:18:25should have a clear option to be used
2:18:27without personal profiling
2:18:29how do you see this change will it help
2:18:31to alleviate the issues connected with
2:18:33the current practice of driving for
2:18:34higher engagement through conflict uh
2:18:37between the users
2:18:39certainly
2:18:40regarding data profiling and data
2:18:41retention platforms are profiling users
2:18:44collecting and analyzing their every
2:18:46action
2:18:47building up their profiles and then
2:18:49creating algorithms to make them stay
2:18:50longer on the platform
2:18:52in order to make them more emotionally
2:18:54invested
2:18:55do you believe this is appropriate or
2:18:57even ethical
2:18:59and lastly
2:19:01your public statements have revealed a
2:19:02lot of inner working of facebook
2:19:05however we cannot solely rely on the on
2:19:08vista blowers we also need much more
2:19:10systemic transparency
2:19:12do you consider the proposed provisions
2:19:14on transparency of algorithms for public
2:19:17and ngos
2:19:18open research and independent oversight
2:19:20sufficient
2:19:21or should europe be more ambitions uh i
2:19:24mean in sense of legal uh law making
2:19:27thanks for for any opinions you couldn't
2:19:30share with us
2:19:32thank you miss hold on please
2:19:39with regard to
2:19:41whistleblower protections uh i i believe
2:19:44that there is a need for stronger
2:19:46whistleblower protections across the eu
2:19:48because not all member states have
2:19:49whistleblower protections um and i want
2:19:52to and even in the united states for
2:19:54example in the united states i was
2:19:56protected under the the securities and
2:19:58exchange commission because i worked for
2:20:00a public company had i worked for tick
2:20:03tock and had the same information i
2:20:05would have had no
2:20:07i would have no legitimate way to bring
2:20:09forward that information
2:20:11or wouldn't have the same protections
2:20:13there's a need for protecting both
2:20:15public and private employees
2:20:21there's also a need for having more
2:20:23on-ramps to disclosure if i had driven a
2:20:26bus in the united states or worked at a
2:20:27hospital there would have been a phone
2:20:29number that was legally mandated to be
2:20:31in my break room that said did you see
2:20:34something that endangered public safety
2:20:36call this number someone will take you
2:20:38seriously
2:20:39well and you'll be you'll be safe
2:20:42nothing like that exists in europe and
2:20:43nothing exists like that in the united
2:20:45states i had concerns about national
2:20:47security because i was working on
2:20:49counter-espionage i think facebook has
2:20:51radically under-invested in major
2:20:53national security issues and i'm happy
2:20:54to talk to any of you about this later
2:20:56um and i had no i had no idea on who i
2:21:00could talk to and i think that's really
2:21:03i think the human cost of that is
2:21:05something that's hard to understand
2:21:07where when you have to hold a secret
2:21:09that you know is jeopardizing people's
2:21:11lives and you have no way to reconcile
2:21:14that that destroys people and i watched
2:21:16that happen to my co-workers
2:21:18with regard to profiling and data
2:21:20retention i think it shouldn't you
2:21:22shouldn't be allowed to take third party
2:21:24data sources something facebook does
2:21:26they work with credit card companies
2:21:28other forms um and it makes their ads
2:21:30radically more profitable i don't think
2:21:33you should i think you should have to
2:21:34consent every time you hook up more data
2:21:36sources because i think people would
2:21:38feel really uncomfortable if they knew
2:21:40facebook had some of the data that they
2:21:42do and there's people who aren't even on
2:21:43facebook who i've i i don't know if this
2:21:45is a current thing but i've watched
2:21:46documentaries and read articles where
2:21:48they say they have shadow profiles for
2:21:50people who aren't yet on the platform
2:21:52where because they have these sources
2:21:54like credit card data and things like
2:21:55that they can form profiles on people
2:21:58and lastly on this question of can you
2:22:00be more ambitious with data i think
2:22:02having a process that is dynamic where
2:22:04we change our data asks over time and we
2:22:06match them to problems
2:22:09people often get fixated on the idea
2:22:11that we need raw data we need access to
2:22:13all the raw data
2:22:14and as someone who has done a lot of
2:22:16data analysis of social media networks
2:22:19raw data i think opens a lot of a can of
2:22:21worms around privacy
2:22:23um and is very hard to work with like
2:22:26the scale of facebook's data is is just
2:22:29amazing
2:22:31having a system where people we assess
2:22:33problems both on by facebook and by the
2:22:35community that's the one and the two
2:22:38and we have to publish aggregate data
2:22:40that would show progress
2:22:42i i think a lot of learnings and and um
2:22:45our ability to understand how these
2:22:47systems work would happen as a side
2:22:48effect of the data we ask for because
2:22:51there's often many many questions you
2:22:52can answer with a set of data not just
2:22:54the original one that you asked
2:22:57thank you next i would like to ask rob
2:22:59ross to take the floor please
2:23:02thank you chair and a special thank you
2:23:05to you uh miss hogan for being
2:23:07here with us today
2:23:09thank you for your presentation and your
2:23:11courage to bring out the information you
2:23:12have revealed i know from my own
2:23:14experience that it's much easier to
2:23:16remain silent than to speak out
2:23:19miss hogan i have suffered several
2:23:21concerns and hope your answers will help
2:23:23us to tackle these very important
2:23:25challenges
2:23:26i listen to your presentation and your
2:23:28appearance in the media and i wonder
2:23:30what you see as the biggest problem
2:23:32is it misinformation or is the algorithm
2:23:35showing targeted misinformation to
2:23:37specific users making them addicted to
2:23:40the platform and further polarizing our
2:23:43society
2:23:44i ask this because regulating
2:23:46transparency about algorithm algorithms
2:23:49is probably easier to address than
2:23:50topics as misinformation and also hate
2:23:53speech i understand that these forms of
2:23:56information can be a real problem but
2:23:58the definitions are at the end of the
2:24:00day somewhat
2:24:01arbitrary
2:24:03and i also fear
2:24:04a world where users can only take notice
2:24:07of the official truth
2:24:09should we include in the dsa that
2:24:12platforms give users an option to
2:24:14completely opt out of any content
2:24:16suggested by algorithms so by giving
2:24:19users the possibility to choose for a
2:24:21real
2:24:23timeline without any suggestion by
2:24:25algorithms of the platform or the
2:24:27platform
2:24:28in addition i have an open question it
2:24:31is it has to do with the mental side
2:24:34effect of instagram and this is a major
2:24:36problem with even resulting in suicide
2:24:39by young girls do we have suggestions on
2:24:42what we should do to tackle that problem
2:24:44thank you very much
2:24:46thank you ms holden please
2:24:50i have repeated frequently that i don't
2:24:52believe the problem with facebook is bad
2:24:54people or bad content is about a system
2:24:58that progress like consistently has a
2:25:00bias towards amplifying giving the
2:25:03largest reach to the most extreme and
2:25:05divisive content
2:25:06that's that's the real problem here um
2:25:09facebook has known this since at least
2:25:112018 when they rolled out
2:25:13meaningful social interactions they know
2:25:15that it is influenced publishers its
2:25:17influence political parties and they
2:25:18know that there's a recurrent problem on
2:25:20their platforms so i think it's about
2:25:22the system and this is systematic
2:25:24choices
2:25:25i think this idea of allowing people to
2:25:27opt out of algorithms i i think people
2:25:30should have more control for example
2:25:32there was an independent person who
2:25:34wrote an extension that allowed you to
2:25:36unfollow things
2:25:38something that most people are not aware
2:25:39of is facebook right now if you are
2:25:41invited to a group
2:25:43and in the case of invites to groups
2:25:45because their rate limits are so high
2:25:48um the top person that invited people to
2:25:50q anon groups invited 300 000 people to
2:25:53qnon groups they did not consent to
2:25:55being invited but facebook
2:25:57then inserted content from those groups
2:25:59into their feed
2:26:01so imagine someone comes along and says
2:26:03we're going to help you unfollow things
2:26:05that you're following and it's me one
2:26:06click
2:26:07facebook threatened to sue that person
2:26:10right because they know they need to
2:26:11keep you in those systems if you're
2:26:13going to consume enough content they
2:26:15don't want you choosing what you look at
2:26:17they want to feed you that content i
2:26:20don't though support
2:26:22allowing people just to opt out of using
2:26:24algorithms because you can't disentangle
2:26:28product choices and algorithms so for
2:26:30example if people don't opt out of all
2:26:32those groups they're in or unfollow you
2:26:34know a variety of things um
2:26:37you'll you'll now get spammed by those
2:26:39mega groups
2:26:40right so
2:26:41uh if you are a member of a million
2:26:43person group you'll get a fire hose of
2:26:45all the content from that group every
2:26:47day and so just opting out the algorithm
2:26:50won't solve the problems that we're
2:26:51we're discussing right now
2:26:53lastly on the mental effects for
2:26:54teenagers um or mental
2:26:57mental health effects overall because
2:26:59something that i don't think has been
2:27:01talked about enough is the role of
2:27:02facebook on things like depression
2:27:04right one of the top risk factors for
2:27:06being overexposed to misinformation is
2:27:08being recently widowed or recently
2:27:10divorced or moving to a new city so
2:27:13facebook is showing the most
2:27:14misinformation to people who are already
2:27:16socially isolated
2:27:18and i think that's problematic they know
2:27:20that being depressed makes you consid it
2:27:22correlates with you consuming more
2:27:23content on facebook but in the case of
2:27:25teenagers um
2:27:28i think a lot of the problems around
2:27:29things like eating disorders are just
2:27:31repeat they're just more echoes of the
2:27:33same pattern that when you follow
2:27:36moderate interests engagement-based
2:27:38ranking pulls you towards extreme
2:27:40interests and they've seen this over and
2:27:42over again they can create a new account
2:27:44they can follow center left or center
2:27:46right content and you'll get pushed to
2:27:48the far left and the far right and
2:27:50that's another thing that will tear our
2:27:51democracy apart and so it really is
2:27:53about systemic problems and how do we
2:27:55address these systematic like these
2:27:57systemic risks
2:28:00thank you next speaker is timo welcome
2:28:02team of the floor is yours please thank
2:28:04you very much chair first of all i would
2:28:06like to thank you uh for your courage
2:28:09for standing up for what you believe in
2:28:11and for sharing your knowledge with us
2:28:13here in the european parliament and
2:28:15answering so many questions so i'm
2:28:18convinced that your insights can indeed
2:28:20change the course of
2:28:22the negotiations on the digital services
2:28:24act and help making it a true digital
2:28:27constitution for our european citizens
2:28:30so you made clear how important it is
2:28:33that we tackle social media algorithms
2:28:37there can be no place in europe for
2:28:38algorithms that promote disinformation
2:28:41and hate over sober political debate and
2:28:44high quality information for profit
2:28:46reasons i really believe that so we need
2:28:48to break the harmful monopoly of big
2:28:51platforms like facebook over our
2:28:53attention
2:28:54so in my view this monopoly rests on two
2:28:57aspects
2:28:58uh content recommendation and
2:29:01personalized advertising so the
2:29:03recommender algorithms are used to keep
2:29:05us on the platform for as long as
2:29:08possible at any cost so that the
2:29:10advertising algorithms can feed us the
2:29:13highly personalized personalized ads to
2:29:15finance the whole operation so my
2:29:18question would be
2:29:19and how far are those two sets of
2:29:22algorithms connected the ones used for
2:29:25recommendation of content and the ones
2:29:27used for advertising should we approach
2:29:30them together or do you think we could
2:29:32solve most issues by focusing on one of
2:29:35the other the papers
2:29:37another question on this information so
2:29:39the papers you shared
2:29:41included a striking figure so and you
2:29:43mentioned it today in 2020 facebook
2:29:46allocated 87 percent of its budget for
2:29:49developing its misinformation detection
2:29:51algorithms to the us and only 13 to the
2:29:55rest of the world this suggests that
2:29:57facebook prefers to focus on the english
2:29:59speaking world
2:30:01what can we in europe do to force
2:30:04facebook and others to pay more
2:30:05attention to what is happening here so
2:30:08do you think we need dedicated
2:30:10legislation to deal with that um yeah i
2:30:13have of course a lot of more questions
2:30:15but i'll keep it for for this for now
2:30:17and thanks again for being here with us
2:30:19and for helping us to make the digital
2:30:22europe safe
2:30:24thank you timo uh miss hogan please
2:30:28i strongly believe that europe has a
2:30:31critical role to play in regulating
2:30:33these platforms because you are a
2:30:35vibrant
2:30:36linguistically diverse democracy
2:30:39um
2:30:40there are real problems with having
2:30:42systems that
2:30:44are having a company that can overlook a
2:30:46language because there are only five
2:30:48million speakers of it
2:30:50um i guarantee you there are european
2:30:53languages or even within larger
2:30:54countries dialects and remember ai is so
2:30:57brittle it's so fragile a dialect i'm
2:31:00guessing uk english doesn't get anywhere
2:31:02near as much actual efficacy as u.s
2:31:05english because it's enough different
2:31:08that ai then breaks
2:31:09right
2:31:11so i think there's a deep deep need to
2:31:13make sure that platforms must disclose
2:31:16what safety systems they have what
2:31:18languages those safety systems are in
2:31:20and the performance per language and
2:31:23that's the kind of thing where you can
2:31:24put in the dsa you can say you need to
2:31:27be honest with us on is this actually
2:31:29dangerous for a large fraction of
2:31:30europeans because when you do
2:31:33legislation like that you actually end
2:31:34up protecting the rest of the world too
2:31:37because it will force facebook towards
2:31:40language neutral content neutral
2:31:42solutions
2:31:43um because that's how what will scale it
2:31:46also speaks up for people who live in
2:31:49fragile places in the world that don't
2:31:50have as much influence right the places
2:31:53in the world that have the most
2:31:54linguistic diversity are often the most
2:31:56fragile um places and and they need
2:31:59europe to step in because you guys have
2:32:01influence and you can help you can
2:32:02really help them
2:32:04um
2:32:05and that includes things like they
2:32:07should publish the content taken down
2:32:08per language it should be how many
2:32:10staffers are actually protecting german
2:32:12elections or protecting french elections
2:32:15um it should be things like uh how many
2:32:18staff like how many moderators are in
2:32:20each language what's the latency for a
2:32:22review in each language because that'll
2:32:24be the only way we can have an equitable
2:32:26system
2:32:27with regard to on the monopoly being
2:32:30driven by content recommendations or
2:32:31advertising
2:32:33i don't think you will be able to um
2:32:37like just having a lot of people to swap
2:32:38in and out recommendation systems i
2:32:40don't think will be effective because
2:32:42facebook's whatever one facebook
2:32:45controls
2:32:46because they control the rest of the
2:32:47product changes and they can have all
2:32:49the data
2:32:50they will have a much more enjoyable
2:32:53more addictive experience than anyone
2:32:54else will and so if you allow people to
2:32:57choose their system um also it's a
2:32:59question on business model it's likely
2:33:01the only way you can make a business
2:33:02that could have its own ranking system
2:33:04is if you charge a subscription
2:33:07as facebook has shown in places like
2:33:10the lesser developed world
2:33:12if you charge for other choices and
2:33:15facebook is free people overwhelmingly
2:33:17choose facebook and so i i think it's
2:33:20really a thing of we need to have
2:33:21accountability and transparency on how
2:33:23facebook works then assuming that we can
2:33:25have other choices and the choices are
2:33:27going to be with what helps us
2:33:30thank you
2:33:31next i would like to ask zipper to take
2:33:34to the floor please
2:33:37thank you very much chair and thanks mrs
2:33:39hogan for casting the light on the
2:33:41business
2:33:42model of facebook and you clearly
2:33:44advocate for accountability for
2:33:47transparency and public scrutiny as we
2:33:50do in the labor committee and in
2:33:52parliament and therefore allow me some
2:33:55questions first of all you already
2:33:57raised some concerns and criticisms
2:34:00regarding the use of automatic tools for
2:34:02content management
2:34:04management and content filters as they
2:34:08cannot reliably identify illegal content
2:34:11resulting in suppression of legal
2:34:13content do you believe in this context
2:34:16that the addition of human review could
2:34:18be a solution
2:34:20second
2:34:22one of the most important safeguards to
2:34:24all our digital communication of course
2:34:26is encryption
2:34:28and on the one hand the european
2:34:30commission announced that it generally
2:34:32does not want to weaken encryption but
2:34:34again and again we have debates about
2:34:37backdoors
2:34:38now we heard about facebook's wider
2:34:41plans regarding end-to-end encryption
2:34:45does it mean that facebook users can be
2:34:47sure that their privacy
2:34:49is actually protected and do you believe
2:34:52that the millions of users communication
2:34:55is safe and confidential
2:34:58now
2:35:00thirdly on the dsa a lot of debates are
2:35:03around the regulation of advertising
2:35:06and i think the current practices of
2:35:09personalized targeting should be
2:35:11replaced
2:35:12by contextual advertising
2:35:15and regarding the targeting for
2:35:17commercial advertising one idea is this
2:35:21should only be possible where users have
2:35:24freely opt-in with a limited time to
2:35:27ensure that the once given consent
2:35:30cannot be exploited what are your views
2:35:33what facebook's reactions will be if
2:35:35this were legally required
2:35:38and finally again and again we have
2:35:41debates about data retention meaning
2:35:43providers have to collect and store data
2:35:46for years and years
2:35:48having in mind facebook's dominance on
2:35:50the market and regarding all the debates
2:35:53we heard not only today do you think
2:35:56this is the right approach to give more
2:35:58powers to facebook and other providers
2:36:00thank you
2:36:02thank you but uh
2:36:04i would like also to ask
2:36:10to take the floor and then we will take
2:36:13also the last question and then i would
2:36:16like to ask
2:36:17ms hogan to respond
2:36:19thank you
2:37:16respect
2:38:53gracias
2:38:58thank you and the last speaker is iban
2:39:01garcia del blanco please
2:39:07the last but not the least uh thank you
2:39:10very much for being here uh ms hogan uh
2:39:13i'm going to speak in spanish from this
2:39:15moment so
2:39:16[Laughter]
2:39:18so please take the
2:39:20headphones um
2:39:56[Music]
2:40:12[Music]
2:40:55[Music]
2:41:01[Music]
2:41:40um
2:41:55the
2:42:00[Music]
2:42:09thank you it's 1931 already and it
2:42:13means i have to
2:42:15thank
2:42:16god early our interpreters thank you
2:42:18very much
2:42:19and now i would like to ask miss hogan
2:42:22to take the floor
2:42:27how much time do i have do i only get
2:42:29two minutes now for those three sorry
2:42:31time is over but uh yeah i'll do this
2:42:33i'll do it as fast as i can i'll do as
2:42:35fast as i can um
2:42:39okay
2:42:41um okay so with regard to human review
2:42:44because two of you asked about it um i
2:42:46human part of why facebook doesn't do
2:42:48enough human review is human review does
2:42:50not scale when they talk about having 40
2:42:52000 moderators they have 3.1 billion
2:42:55uh users like it's just not possible to
2:42:57have humans review enough of the content
2:43:00that's why i advocate for making these
2:43:02systems
2:43:03more human scale less opportunities for
2:43:05mass distribution to a million people
2:43:07and
2:43:09finding more ways that humans moderate
2:43:11each other
2:43:12versus having ai do it
2:43:14with regard to encryption
2:43:16i'm really concerned that facebook
2:43:18should publish what is it doing in terms
2:43:21to implement uh encryption because when
2:43:23you give people a false sense of safety
2:43:25they can actually put themselves at
2:43:27jeopardy and so forcing facebook to
2:43:29disclose like what are the safety
2:43:30systems how do they work because we saw
2:43:32with a whistleblower prior to my
2:43:35my sec filing just a week before that
2:43:38they felt that facebook had not
2:43:39adequately communicated how safety
2:43:42systems worked on whatsapp and that
2:43:44people would be surprised so i think
2:43:46having that level of transparency is
2:43:47really essential
2:43:49um with regard to
2:43:52uh in
2:43:53[Music]
2:43:54let's see
2:43:56i do worry about this concept of
2:43:58addiction or things that are are drawing
2:44:00people in
2:44:02there were two comments around the hours
2:44:03and hours that people spend on these
2:44:05platforms i think
2:44:06this is true across all platforms i
2:44:08think you should have to disclose by
2:44:10percentile
2:44:11and and not just by percentile for
2:44:13extreme users so this is the top one
2:44:15percent maybe by 10th of a percentile
2:44:18how many hours do people spend on these
2:44:19systems per week
2:44:21because the real problem on a lot of
2:44:23these things for the harms involved is
2:44:25that as people get exposed to targeted
2:44:27misinfo or the misinfo of these
2:44:29concentrated communities they get cut
2:44:31off from the rest of reality right that
2:44:34when you begin to believe beliefs that
2:44:36are not consent consensual like beliefs
2:44:38that are the consensus norms it makes it
2:44:40harder for you to reintegrate into
2:44:42society and you pull people into these
2:44:44these rabbit holes um so i think
2:44:46disclosing how much time people are
2:44:48spending is going to be important
2:44:49especially for the metaverse
2:44:51for those of you who are not familiar
2:44:53with the novel snow crash which is where
2:44:54the phrase the metaverse came from
2:44:56specifically talked about in that book
2:44:58i'm shocked they picked this name in
2:45:00that book the meta the metaverse is a
2:45:02dystopian thing that people's lives are
2:45:05so unpleasant that they need to hide in
2:45:07the system for half of their half of
2:45:09their day i think that's what's going to
2:45:11happen you're going to have people who
2:45:12are disconnected from society or
2:45:14marginalized and you're going to have
2:45:16people getting pulled into these virtual
2:45:18worlds because they are really addictive
2:45:20they are really compelling they give
2:45:22people a sense of agency when they might
2:45:23not have it in their real lives and i
2:45:25think requiring
2:45:27publishing or having people disclose
2:45:29like do they feel like they're in
2:45:30they're in control we have to have ways
2:45:32of getting
2:45:33research and governance on these things
2:45:36with regard to
2:45:40the quantity of data delivered in the
2:45:42case of the metaverse i'm very concerned
2:45:43about that we talked about before that
2:45:45these systems require having even more
2:45:47sensors in our life and the issue i'm
2:45:49even more concerned about is if they're
2:45:50trying to aim for the workplace people
2:45:52don't get to consent in their workplaces
2:45:54in the same way and so i worry about the
2:45:56idea of facebook forming profiles on us
2:45:58where people can't even opt out because
2:46:00it's between their livelihood and
2:46:02their privacy
2:46:04um
2:46:05[Music]
2:46:07i highly encourage this idea on for for
2:46:10opting in on personalized targeting
2:46:13there should be time limits
2:46:15because people should be able to make
2:46:16different choices over time and i
2:46:18strongly encourage having that be
2:46:19time-bounded and i think the last one
2:46:22i'm going to comment on
2:46:27data retention so something that
2:46:28facebook definitely takes advantage of
2:46:30is they no longer retain data beyond 90
2:46:33days
2:46:34um that actually has some pretty serious
2:46:36security implications so
2:46:38a thing that we found across when i was
2:46:41working with the threat investigators
2:46:43was that often it was hard to see how
2:46:46things like influence networks evolved
2:46:48or the impact of terrorism recruiting
2:46:50because we could only see back 90 days
2:46:52the only way you got to have data over
2:46:55time was if you uh you intentionally
2:46:57flagged that person and so it became
2:46:59difficult to actually track the
2:47:01evolution of some of these systems um
2:47:03and so i think there are uh there should
2:47:06be questions around
2:47:08how data retention works i'm not saying
2:47:09keep it for years but maybe more than 90
2:47:12days would be helpful
2:47:13maybe six maybe six months or something
2:47:16yeah that's what thank you so much for
2:47:17your time i'm incredibly grateful that
2:47:20the european union is taking this
2:47:21seriously i think you guys are really a
2:47:23light in the darkness like i feel
2:47:25confident that you guys are going to
2:47:27pass a law that has teeth
2:47:29that it has dynamic ways of adjusting
2:47:31over time
2:47:33because we need that we are going to
2:47:35need oversight on these platforms more
2:47:37and more with time they have far too
2:47:39much power in our society to be able to
2:47:41operate in the dark and facebook has
2:47:43shown they will not give us any data
2:47:45unless we force them to the last thing i
2:47:47want to drop in there before my six
2:47:48minutes expires is facebook has shown
2:47:51that they will lie with data and so i
2:47:53encourage you to put in the dsa
2:47:55if facebook gives you data they should
2:47:57have to show you how they got it
2:47:59because we have had multiple scandals
2:48:01where facebook says they've given out
2:48:02data but because every data analysis
2:48:05requires assumptions
2:48:06facebook makes assumptions and then
2:48:08doesn't disclose them so we think we
2:48:10know what data we're looking at but in
2:48:12reality they are misleading us they have
2:48:14the best data scientists in the world
2:48:16they can dance with data unlike anyone
2:48:18else and it's really really important
2:48:20that they should have to disclose the
2:48:21process the queries the notebooks they
2:48:24use to pull this data because you can't
2:48:26trust anything they give you unless you
2:48:28can confirm them
2:48:29thank you so much
2:48:31[Applause]
2:48:45thank you
2:48:46thank you very much for being with us uh
2:48:48to hear
2:48:50this
2:48:51evening
2:48:53i would like to thank all the speakers i
2:48:55would like to thank all the members of
2:48:57the parliament and our guests taking
2:48:59part in this hearing
2:49:02this fundamental piece of legislation on
2:49:04what our digital economy is based in the
2:49:07european union is e-commerce directive
2:49:10all those main principles country forage
2:49:13in principle limited liability
2:49:15liability principle uh
2:49:18prohibition of generator general
2:49:20monitoring or coming from e-commerce
2:49:22directive but we got to the sea uh
2:49:26commerce directive already 20 years ago
2:49:28at that time we didn't have the social
2:49:32media platforms we didn't have the
2:49:34sharing economy uh platform
2:49:37platforms and it's absolutely clear we
2:49:40have to update this e-commerce directive
2:49:43and our digital services act is dealing
2:49:46exactly
2:49:48with those issues so i'm absolutely sure
2:49:50that
2:49:51all our members of the parliament they
2:49:53got a lot of pros and contras and now
2:49:58they are
2:49:59more confident in in their decisions
2:50:01than before this hearing once again
2:50:04thank you very much for being here with
2:50:06with us today

About this transcript

This page contains the full transcript of LIVE: Facebook Whistleblower Haugen Testifies at an EU Parliament Committee Hearing by NTD News, generated from the public captions YouTube serves with the video. The transcript has 21,581 words across 3,680 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.