What the Facebook Whistleblower JUST Said

okay folks here is a summary of part one

of the testimony of francis hogan 37

years old product manager former product

manager at facebook she's got a degree

from harvard and this is a summary of

what she mentioned to the senate uh

today or the senate hearing committee

so francis hogan mentions that a

facebook repeatedly misled the public

about what its own research reveals

about the safety of children and the

efficacy of its artificial intelligence

systems

as a role in spreading divisive and

extreme messages she says that facebook

has the potential to bring out the worst

of us but could also bring out the best

of us if we had proper controls for

facebook she believes that facebook

harms children stokes division and harms

democracy she says that facebook puts

profit before people and that they will

not solve the problems they have without

the help or compelling of congress she

says that facebook resolves conflicts

within the company in favor of its own

profits rather than

uh what is best for society later she

mentions that facebook pays for their

profit with our

safety she says the results are more

profit for facebook but more lies and

more combativeness in our society and

sometimes sometimes this actually leads

to violence

she mentions that the company

intentionally hides vital information

that could potentially lead regulators

to

pay attention to facebook more

she says that facebook purposefully

hides behind walls that keep researchers

out this is different from what google

does for example where researchers can

access data and do more research

comparison she made

these uh problems she says are very

solvable and she gives some suggestions

for how to solve these issues which

we'll talk about in just a moment

she also mentions that children most

children who use facebook probably don't

have the self-awareness to recognize the

true issues of what facebook is actually

doing to them and how facebook is

actually hurting them

she mentions that facebook uh likely

she mentions that facebook says that

children lie about their age but the

reality is statistics within facebook

show that most children tell the truth

and you can figure out most children's

actual ages by doing some backward

analysis

and

she mentions that the real problems that

facebook beyond

people potentially lying about their age

because this was a question that came up

were really a staffing related that

there aren't enough humans involved in

decision making or problem solving at

facebook

and so

ultimately when there is understaffing

groups within facebook are discouraged

from actually trying to solve the issues

because that could potentially lead to

more work

when they're already overwhelmed so for

example francis mentioned that she was

on a counter espionage uh group or cell

or whatever

and uh at any given time she said they

could only deal with one third of the

caseload they had if they would work on

building a better

detector to find more problems then they

would just be even more overwhelmed and

even more behind and so she cites a lack

of

humans and a lack of staff as some of

the problems of the issues that we have

at facebook

she says that

her goal is to fix facebook however

facebook should declare moral bankruptcy

she mentions that some things that could

help are things like the kids act the

kids act by the way are things or is an

act that would limit things like

autoplay push alerts and like buttons or

even follower accounts for children

there was a mention that facebook's q2

revenue per user is 35 and 58 cents

annualized that would be over

uh 135 dollars

which is pretty incredible

actually probably over 140 dollars

senator klobuchar says that uh

algorithms are likely to promote uh

outrageous content to which base or to

which uh francis says absolutely

facebook uses engagement based content

algorithms that basically amplify

extremes and in reverse engineering

facebook found that it could actually

lead people in studies

to

to

do or act a certain way so for example

when the question was asked can

facebook's algorithm lead people to

anorexia she said the answer was yes

that research had shown that facebook

has that power which is kind of

incredible to think about that facebook

could essentially manipulate you by

catering certain messages to you

facebook invests more in its users that

make them more money she says uh

specifically in the english language she

talked a little bit more about the

algorithm the engagement based ranking

algorithm

she counters what facebook might say she

believes that facebook would say that

facebook uh says you won't like facebook

as much if we didn't cater content for

you

she has more of a preference

saying that hey maybe we don't need

algorithms to cater content for us maybe

maybe we could go time based uh on on

the way we view content in a feed but

we've also seen that by the way like

youtube

remember back in the day and i'm making

this reference here back in the day

youtube was very subscription-based if

you subscribe to a channel you would see

a list of all the videos from your

subscriptions

but youtube found that that didn't

actually increase viewership instead

providing you content that wasn't

necessarily from your subscriptions list

but what was most deemed to get the most

view time from you

ended up growing the platform

substantially more and so that's why

they switched to that sort of platform i

imagine facebook has found

very much the same thing to be true that

catered content

increases engagement and that's why they

use that the problem is

the

content that facebook is likely to

promote is content that's likely to

quote elicit an extreme reaction from

the user

because in that case it's more likely to

generate likes and shares

leading other people not only to view

and consume that content and then

potentially share that content but also

promoting other people to create content

and so on youtube it's worth noting that

when youtube suggests a video to you

it's because each video has a

predetermined or pre-estimated

watch time per impression and so every

time they show you a thumbnail they

assume that on average you're going to

spend five minutes here or 10 minutes

here eight minutes here but that watch

time per impression is really important

because it also calculates how likely

you are to share that video because see

if you watch a video for eight minutes

and don't share it that's eight minutes

if you watch a video and share it with a

person who then shares it to another

person who shares it to another person

that could be 24 or 32 minutes of watch

time per that one impression they gave

you right so they're looking for that

sort of outrageous or extreme content

youtube does pretty much the same thing

facebook says

it's uh artificial intelligence can find

bad content uh francis says but she

argues that you can't rely on artificial

intelligence

to solve these issues or or to find bad

content because in her research

they found uh and and on sort of groups

that she's worked on they found that

ai was only about 10 to 15 percent

effective at actually finding content

that violated uh the terms of services

or was bad content

on facebook and so she believes that's

very important for not only research to

be shared with congress but congress to

provide oversight to congress maybe even

facebook and cr create suggestions for

how that oversight should look uh and

congress can collaborate on that sense

she was also a big fan of what twitter

is doing with link sharing when twitter

when somebody posts a link and somebody

wants to retweet that twitter has a

little notification that pops up and

says you haven't read this aren't do you

want to read this article before you

re-share it facebook

she says doesn't do this because that

would reduce shares and facebook likes

shares so they don't want to stand in

the way of of of um people sharing

content she doesn't believe that you

having to click on the article before

sharing it is actually a form of

censorship instead she says hey like

this is a good way to help prevent miss

or disinformation

now uh let's see here she also mentions

and i thought this was very interesting

uh another senator came up and brought

up some examples of how facebook's ad

review had failed a few times i actually

have that up right here

and here's an example

so

this is by the

techtransparencyproject.org the ttp

referenced by a senator during the

hearing

and they submitted this ad

targeted to 13 to 17 year olds they

never ran these ads they just wanted to

see if these ads would get approved

and so here's one it says it's time for

some skittles are you in and it's a

picture of a bunch of drugs throw a

skittles party like no other and that

was approved

here's one

that shows a picture of a skinny girl

and says when you're craving a snack

visit pro anorexia sites to feed your

motivation and reach your goal except

they shorten anorexia and just call it

ana and then they call it an antitip and

how this got approved or here's some

other dating or gambling or alcohol or

smoking ads that got promoted to 13 to

17 year olds

and they show

how quickly these were approved a

potential reach of 9.1 million 13 to 17

year olds and they submitted these ads

and got basically approval within uh at

exactly the same time

within about an hour without within an

hour and one minute all of these ads

were approved and look at that all of

the ads were approved at exactly the

same time see that here this

41 pm approve approve approve approval

proof all of them at exactly the same

time now

francis

it doesn't work on these particular ad

teams but she believes that it's likely

that algorithms are approving these and

that it's very unlikely that humans are

looking at at all of these ads and if

facebook is relying on algorithms to

review ads then it's likely that only 10

to 15

of bad ads are actually caught

which is a problem

and especially when she believes that

facebook creates mental health issues

for not just

women but also children in general which

we've seen

through the wall street journal expose

where

there's ample evidence of mental health

deterioration in individuals who are

children especially girls but also boys

and

one of the issues that she cites is

suggesting that

parents

don't really know how to help their

children because they didn't really grow

up with social media they might say

something like hey just stop using

facebook but that would be bad advice

because

people have a temptation to continue to

resort to social media because of the

way our brains are wired the dopamine

release of every time we check our phone

and so she says this is an issue because

when you wake up in the morning you see

people who are cruel to you when you go

to sleep at night you see people who are

cruel to you through facebook

and this is very different from how

bullying for example used to work

bullying in schools didn't used to

follow you home now it follows you home

wherever you are

and you don't get a break from that sort

of bullying and and that continues to

unfortunately perpetuate

negative mental health

for for children

she uh she's a big fan of promoting

oversight and regulation to

section 230 the communication decency

act she mentions that

platforms like facebook have 100 control

over their algorithms and algorithms

should not prioritize growth at the sake

of or at the expense of public safety

uh she also mentions that 80 to 90 of

sensitive content in terms of swear

words somehow still makes it through

facebook algorithms and facebook

algorithms are especially bad at

identifying

inappropriate content in other languages

but they're also very bad at identifying

inappropriate content in english

then we have uh let's see we talked

about the ads already

and she mentions that facebook has a

focus on scale how can we do things

cheaply and fast it's very possible that

a lot of content on facebook is never

reviewed by uh that submitted like for

ads for in this case was for ads she

believes it's very likely and very

possible that most ads are not reviewed

by humans

we touched on that already she believes

we need to quote human scale social

media not computer scale social media

this kind of goes back to not having

enough staff

and i you know my sort of takeaway was

really that in her opinion

artificial intelligence sucks at

preventing the negative from exploding

on facebook that

algorithms have too much biases or have

too many biases and that left unchecked

you end up with like she mentioned in

her opening remarks more divisiveness

more potential violence more bullying

more mental health problems and so this

gives you a summary

of the first and longest part of the

senate testimony from francis hogan i

expect the second part to be relatively

similar but uh this gives you

part one so thank you so much for

watching if you found this helpful

consider subscribing and folks we'll see

you in the next one

[Music]