Artificial Intelligence and Social Media

Summary

This transcript explores the influence of social media platforms, emphasizing how companies utilize notifications and artificial intelligence to keep users engaged and drive profit. It delves into the role of machine learning in shaping content recommendations, sentiment analysis, and micro-targeting, raising concerns about privacy, bias, and misinformation. The script also touches on the emotional impact of social media, especially on teens, and discusses potential risks, including cyberbullying and manipulated perceptions. It concludes with a call for better regulation and management of social media's rapidly evolving landscape to safeguard users.

Takeaways

• 🐣 Social media companies constantly grab users' attention through notifications and emails that encourage content creation or site visits.
• 🤖 AI, particularly machine learning and sentiment analysis, is used by social media platforms to interpret and categorize user-generated data based on emotions and opinions.
• 😊 Emoticons help train AI engines by adding emotional context to messages, especially when understanding humor and sarcasm.
• 📈 Recommendation engines on social media platforms use user data, like content interactions, to suggest targeted ads or posts based on personal preferences.
• 📷 Facial recognition technology identifies and tags people in photos, contributing more data to social media algorithms about user behaviors and social connections.
• 💼 LinkedIn's AI tools match users with job opportunities, while YouTube uses algorithms to recommend videos, leading users down 'rabbit holes' to increase engagement.
• 😟 Social media's unregulated environment contributes to cyberbullying, teen anxiety, and the spread of misinformation, which can have serious real-world consequences.
• 🧠 Social media triggers reward centers in teen brains, encouraging addictive behaviors, while also intensifying feelings of anxiety and exclusion.
• 💻 Social media shapes our reality by filtering the information we see, reinforcing confirmation biases and sometimes manipulating perceptions for political or commercial gain.
• 📊 Predictive analytics on social media use vast amounts of data to forecast user behaviors, decisions, and even election outcomes, potentially influencing opinions and beliefs.

Q & A

What are the two main types of notifications used by social media companies?

The two main types of notifications are: 1) Notifications encouraging users to create more content, such as reminders about sharing updates when engagement is low, and 2) Notifications that drive users to revisit the platform, such as alerts about likes, mentions, or events.

How do social media companies use machine learning to analyze user content?

Social media companies use machine learning, particularly sentiment analysis, which combines natural language processing (NLP) and machine learning to categorize user content as positive, negative, or neutral, helping AI tools understand user emotions and opinions.

What role do emoticons play in sentiment analysis on social media?

Emoticons help AI sentiment analysis tools understand the tone of messages, especially in cases of humor and sarcasm, which are difficult for machines to interpret accurately.

How do recommendation engines on social media platforms work?

Recommendation engines collect data on user interactions, such as posts liked, content reviewed, and user activity, to display targeted ads or relevant posts based on the user’s preferences and patterns.

What are the ethical concerns around facial recognition technology used by social media companies?

Facial recognition algorithms often have biases, particularly toward underrepresented groups, such as people with darker complexions. These biases stem from unequal representation during the development stages, raising concerns about fairness and accuracy.

How does social media contribute to the spread of fake news?

Social media facilitates the rapid spread of false information because there is no verification system in place for self-published content. Studies show that fake news travels faster on platforms like Twitter, often due to its sensational nature.

What are some of the psychological effects of social media use on teenagers?

Social media can intensify feelings of depression and anxiety in teens. Factors include pressure to post appealing content, fear of missing out (FOMO) when seeing others' posts, and the stress of getting likes or positive feedback on their own posts.

How do social media platforms create echo chambers that influence user perception?

Social media algorithms often show users content that aligns with their views, reinforcing their opinions. This creates confirmation bias, where users primarily see posts that support their beliefs, contributing to polarized perspectives on controversial topics.

What is micro-targeting and how is it used in political campaigns?

Micro-targeting is a strategy that uses detailed data about users’ preferences, demographics, and behaviors to deliver personalized political ads. It is often criticized for its lack of transparency and potential for misuse in manipulating voter perceptions.

What concerns are there regarding AI's role in shaping the future of social media content?

There is concern that AI could be used to manipulate information, influencing user beliefs by curating the content they see. This could lead to AI controlling narratives, shaping public opinion in ways that are not transparent or democratic.

Outlines

00:00

📲 Social Media's Attention Economy

Social media platforms use notifications to grab our attention, either encouraging content creation or prompting visits to the platform. These notifications are designed to be vague, making users curious enough to click through, which often leads to increased exposure to advertisements. Social media companies leverage AI, particularly machine learning and sentiment analysis, to interpret user-generated content and emotions. Emoticons are a key factor in helping these systems understand human expressions like humor and sarcasm.

05:02

📊 AI and Social Media Recommendation Engines

Social media platforms employ machine learning to power recommendation engines, tracking user behaviors such as posts, clicks, and likes to serve relevant content or ads. Facial recognition technology also aids in identifying individuals in photos. LinkedIn, for example, uses AI to match job seekers with employers based on user data. YouTube's algorithms are adept at keeping users engaged, pushing content to lead users down 'rabbit holes.' These strategies generate profit for platforms like LinkedIn and YouTube while increasing user engagement.

10:02

🚨 Social Media's Influence on Mental Health and Behavior

Social media platforms have a significant impact on teen mental health, with studies showing links to increased depression and anxiety. The pressure to receive likes and positive feedback on posts can contribute to these issues. Social media also creates an addiction-like effect by constantly adapting to user preferences. This leads to problematic behavior, similar to how drug dealers cater to users, reflecting the deep psychological influence that these platforms can have.

15:03

🔍 Social Media Misdirection and the AI Matrix

Just like magicians use misdirection to divert attention, social media distracts users with posts and content while subtly influencing their perceptions. AI filters the information we receive, shaping our worldview. With many relying on social media for news, the content tends to reinforce existing beliefs, creating a feedback loop of confirmation bias. Social media has unprecedented influence, as demonstrated by how fake news spreads faster than facts and how platforms like Facebook have been used for political manipulation and social discord.

💥 The Real-World Consequences of Social Media Misinformation

False information spread on social media has led to dangerous real-world consequences. One notable example is 'Pizzagate,' where a man acted on conspiracy theories by attacking a pizzeria. Similarly, in Myanmar, a false report on Facebook incited mob violence. These incidents highlight how social media algorithms can amplify harmful narratives. While platforms may not intentionally cause harm, their design allows misinformation to spread rapidly, with dire effects.

🛑 AI Bias and Misrepresentation in Social Media

AI tools used by social media companies can carry hidden biases, especially in areas like facial recognition, where dark-skinned individuals are often misrepresented due to inadequate training data. Public perception can also be skewed, as people consistently overestimate or underestimate certain demographic statistics. These biases affect broader societal behaviors and decisions, including voting patterns and consumer preferences.

🎯 Social Media and Election Manipulation

Social media's influence extends to political campaigns through micro-targeting, where user data is analyzed to deliver specific ads and messages. This covert form of advertising can sway public opinion without transparency. Platforms like Twitter have attempted to mitigate these issues by banning political ads and addressing the dangers of deepfakes, which are becoming more sophisticated and deceptive with advancements in AI.

📉 Predictive Analysis and the Future of Social Media Manipulation

At its deepest level, social media utilizes predictive analysis to forecast user behavior and preferences. This enables platforms to shape public opinion and even alter election outcomes by tailoring content. The concern is that these predictive tools might evolve to the point where they manipulate beliefs en masse. The rise of social media during COVID-19 has made it an even more powerful communication tool, but like early automobiles, which were eventually made safer through regulation, social media must also be better managed.

Keywords

💡Social Media Notifications

Social media notifications are alerts sent to users, encouraging them to engage with the platform. These can include prompts to create content (e.g., 'You had no new likes this week') or to revisit the site ('Your post has been liked'). The video explains how these notifications are carefully designed to provoke user interest and draw them back into the platform, where they are exposed to advertisements.

💡Sentiment Analysis

Sentiment analysis is a branch of machine learning that processes and interprets user-generated content to categorize it as positive, negative, or neutral. The video highlights how social media platforms use this tool to understand emotions and opinions within posts, often aided by emoticons, which help algorithms grasp nuances like humor or sarcasm. This forms the foundation of how AI engines learn about users' feelings.

💡Recommendation Engines

Recommendation engines are AI-powered tools that collect and analyze users' behavior—such as the content they interact with—and predict what they might enjoy next. Social media platforms use these engines to suggest posts, advertisements, or videos that are relevant to users. The video points out that these systems are central to platforms like YouTube and Pinterest, creating personalized experiences that keep users engaged.

💡Facial Recognition Technology

Facial recognition technology is used by platforms like Facebook to automatically identify and tag people in photos. The video explains how this technology works by feeding AI with more data about users' locations and social connections, which can then be used to refine advertising or content recommendations. It's an example of how personal data is collected and utilized without direct user involvement.

💡Artificial Intelligence (AI)

Artificial Intelligence (AI) refers to the computer systems that perform tasks traditionally requiring human intelligence, such as understanding language or recognizing images. In the context of the video, AI is the engine behind features like sentiment analysis, recommendation systems, and facial recognition, driving many of the user experiences on social media platforms and shaping the way content is delivered.

💡Cyberbullying

Cyberbullying is the use of digital platforms to harass or intimidate individuals, often with serious emotional consequences. The video discusses how social media platforms have become a breeding ground for this behavior, linking it to rising rates of teenage self-harm and suicide. It highlights the responsibility that social media companies bear in moderating harmful content, a task they often neglect due to profit motives.

💡Fake News

Fake news refers to the spread of misinformation or deliberately falsified stories via social media platforms. The video provides examples like Pizzagate, where false information spread rapidly and led to real-world consequences. It emphasizes that the algorithms driving content distribution often amplify sensationalized or false content because it engages users more effectively than the truth.

💡Machine Learning Bias

Machine learning bias occurs when AI systems make flawed decisions based on biased or incomplete training data. The video uses the example of facial recognition algorithms struggling with dark complexions due to underrepresentation in training data. It warns that AI systems can perpetuate and exacerbate existing societal biases if not carefully managed.

💡Microtargeting

Microtargeting is a marketing strategy that segments users into small, specific groups based on their data—such as interests, demographics, or behavior—and then delivers personalized content or ads. The video explains that social media platforms use microtargeting extensively, especially in political campaigns, to influence users' opinions and actions by tailoring messages to resonate with their specific preferences.

💡Confirmation Bias

Confirmation bias is the tendency for people to favor information that confirms their pre-existing beliefs. The video explains that social media platforms often reinforce confirmation bias by showing users content that aligns with their views, especially in controversial areas like politics. This creates echo chambers where users are exposed only to perspectives they already agree with, reinforcing divisive or extreme viewpoints.

Highlights

Social media companies use notifications and emails to drive users to create more content or visit the site.

Artificial Intelligence and machine learning tools are used to review user data and understand sentiments through sentiment analysis.

Emoticons help AI better understand the nuances of human emotions, humor, and sarcasm, essentially training the AI engines.

Social media recommendation engines gather data on user interactions and preferences to suggest relevant posts and advertisements.

Facial recognition technology is commonly used on social media platforms, such as Facebook, to tag people in group photos.

LinkedIn uses AI to match users to potential employers by analyzing hiring patterns, skills, and job descriptions.

YouTube’s recommendation engine uses user behavior to push more content, often leading users down content 'rabbit holes' for increased engagement.

Studies link social media to increasing rates of cyberbullying, teen depression, anxiety, and even suicides.

Social media can create anxiety by fostering comparison culture, where users feel pressure to post and get likes or comments.

Machine learning algorithms, despite their complexity, may reinforce biases, as demonstrated by facial recognition technology's lower accuracy on dark-skinned individuals.

Misinformation spreads rapidly on social media; for example, fake news on Twitter spreads six times faster than the truth.

Pizzagate, a conspiracy theory, is an example where false information led to real-world violent actions.

Social media platforms are often accused of being complicit in amplifying false narratives due to algorithmic biases that promote engagement.

Micro-targeting in political campaigns leverages detailed personal data to craft covert and highly targeted advertisements.

Deep fakes, powered by AI, have raised concerns about the manipulation of audio and visual content, potentially deceiving large audiences.

Transcripts

like

chicks in a nest social media companies

clamor for our attention with their

emails notifications and beeps

these come in two varieties

notifications that encourage you to

create more content

you had no new likes this week share

updates about yourself and friends by

creating a new post or video

and notifications that encourage you to

go and visit the site

your post has been liked you were tagged

in a photo

you have been mentioned birthday

reminders

there's an event happening near you

someone viewed your profile

here are tweets that might be of

interest to you

the notifications and emails never tell

you the full story

they're designed to provoke your

interest so that you follow the link to

the social media site

and there you are subjected to more

advertisements

and the till's ring catching

at level two we're looking at artificial

intelligence

and more specifically machine learning

these tools review the large amounts of

data that us users post on social media

sites

and they have the ability to decipher

these data

and gain an understanding of what is

being said

here we use a branch of machine learning

called

sentiment analysis a process that uses

natural language processing

nlp and machine learning to pair social

media data

with predefined labels such as positive

negative or neutral

in this way machines can develop agents

that learn to understand the opinions

and emotions

underlying the messages we are asked to

help in this process

by adding the emoticons these help the

sentiment analysis tools better

understand the nature of our messages

particularly in the areas of humor and

sarcasm

which are concepts ai machines find

difficult to understand

so by using emoticons we are in effect

training the ai engines

having analyzed all the data that we the

users have posted

what do the social media companies do

with it

well at level two they do the following

they populate

social media recommendation engines the

recommendation engines collect data on

the content you engage with

from pinning a picture on pinterest to

commenting on an instagram post

and then they display material that they

anticipate you will enjoy

this material can be either targeted

advertising

or other people's posts that are

perceived to be relevant

the recommendation engines collect

information relating to

the content you post the content you

review and click on

and the number of likes you put on

certain posts

and in this way they can deduce your

personality

facial recognition technology is used to

identify people in photographs

you may have seen on facebook that when

you post a group photo

many of the faces if not all of them

will be identified and automatically

tagged

you can then enhance the knowledge base

by clicking on the

check-in icon and in this way you can

tell facebook

that all these people were in this

location at this time

this then feeds the machine learning

programs with more information on the

preferences of the individuals

linkedin has two facets a social media

platform

facebook for corporate workers which

uses its recommendation engine to

suggest people that you might like to

connect with

but it makes its money from recruiters

seeking talent

as a user you have kindly updated your

career history and skills

and linkedin uses an ai tool from

recently acquired bright

to perform intelligent matches for both

employers and job seekers

it takes into account the user's hiring

patterns work experience

and similarities in job descriptions

kaching

money for linkedin youtube is

particularly adept at analyzing your

usage patterns and making

recommendations

for what you should look at next these

are designed to tempt you down

rabbit holes so that you continue

watching videos

and in so doing you're subjected to more

advertisements

catching again national laws run way

behind

the environments that technology

companies are creating

social media companies are commercial

organizations and are driven by the

pursuit of profit

therefore they are not incentivized to

regulate themselves

this leads to at best a degree of

negligence in reviewing the content that

they publish

unregulated self-published content is

causing

increasing concern imperial college

london

published a report in the lancet

outlining the harm

caused to teenagers by cyber bullying

teen suicides and self-harm have

dramatically increased over recent years

there is evidence that supports a link

with social media

youtube instagram and snapchat are the

most common social media platforms for

teens

in a study by researchers at the ucla

brain mapping center

they found that certain regions of teen

brains became

activated by likes on social media

sometimes causing them to want to use

social media more

researchers believe social media can be

associated with the intensification

of the symptoms of depression the

harvard graduate school of education has

posted a research story

entitled social media and teen anxiety

with anxiety being triggered by seeing

people posting about events to which

they haven't been invited

feeling pressure to post positive and

attractive content about yourself

feeling pressure to get comments and

likes on your posts

and having someone post things about you

that you can't change or control

just take a moment and think do you

recognize any of these behaviors in

yourself

can you control your addictions one of

the most compelling features of social

media

is that it responds to your choices and

so adapts to suit your preferences

there are two groups of people that call

their customers users

one is the technology companies and the

other is drug dealers

now let's move to level three where we

look at the less

obvious behavior patterns facilitated by

social media companies

magicians are able to use misdirection

because our brain automatically

categorizes people's motions

by interpreting their intentions we see

somebody push their spectacles up the

bridge of their nose and assume

that their glasses have slipped but the

magician uses this motion

to hide something in their mouth

similarly

whilst we're being distracted by the

posts photos and friendly surveys we

partake in

we are being misdirected away from what

is happening beneath the surface

the threat from artificial intelligence

is not taking the shape of terminators

being managed by skynet it is much

closer to the matrix

where we are living in an artificially

constructed reality

by this i mean that the view we have of

the world is through

our computers and mobile devices and the

information we receive

is being filtered by ai engines these

are regulating the news articles that we

see

which of our friends posts we see and

the advertisements in our news feeds

a recent pew survey found almost

60 percent of us regularly use social

media for our news

but we are complicit in constructing our

artificial world

by creating a confirmation bias

by this i mean when it comes to

controversial topics

including politics if you're like most

people the majority of your friends and

followers on social media

probably share your outlook this means

that the vast majority of tweets

facebook posts pins or other content you

read on these sites

tend to express the same point of view

as your own

here are some july 2020 figures world

population

7.8 billion unique mobile phone users

5.2 billion internet users

4.6 billion active social media users

4 billion social media companies have

unprecedented audiences facebook youtube

and whatsapp

dominate the league table twitter

clearly punches above its weight with a

comparatively small

326 million this creates

a lot of influence any advertiser will

tell you

the most powerful form of advertising is

word of mouth

we believe what our friends tell us

social media

provides a vehicle for forwarding and

liking messages

so that they become word of mouth

endorsements

because social media sites provide the

vehicle for self-publication

there's no verification of the content

this leads to fake news

and we have a natural bias towards false

and potentially more exciting

information

in 2018 an mit study showed that fake

news on twitter travels six times faster

than the truth

each person with marginal views can see

that they're not alone

and when these people are introduced to

each other on social media

they collaborate and share information

here are a couple of examples

where social media postings have been

interpreted as genuine nudes

pizzagate here conspiracy theorists

falsely claimed in posts made on social

media that a number of emails linked

high-ranking democratic party officials

within an alleged human trafficking and

child sex

ring managed out of the comet ping pong

pizzeria

in washington dc a man from north

carolina

believed these stories to be true and

went to the comet ping pong

armed with a rifle which he fired inside

the restroom in the belief he was there

to rescue children

the restaurant owner and staff received

death threats

from other conspiracy theorists in

myanmar

previously known as burma on the evening

of july 2014

a mob of hundreds of angry residents

gathered around the sun tea shop

in the commercial hub of mandalay

myanmar's second largest city

the tea shop's muslim owner had been

accused falsely

of misdeeds against a female buddhist

employee

the accusations against him originally

reported on a blog

exploded when they made their way to

facebook

is facebook complicit in these events

well certainly

they allowed false accusations to be

published

but to be fair that's simply what they

do and the way their algorithms work

their ai tools identified the nature of

the posts

and direct them to people showing

similar interests

and so facebook became the wind behind

the fire

you may think handing decisions over to

a machine

would remove any human bias overtly tech

companies are politically correct

and do not display any segregation or

discrimination

however their machine learning

algorithms do have a covert bias

it is what they're designed to do they

categorize people and make decisions

based on these categorizations

an example is that social media facial

recognition algorithms

do not work well on people with dark

complexions

and the reason is that this group were

underrepresented in the machine learning

stages

of the algorithm's development

our perception doesn't always match

reality

here are some examples from surveys

people in saudi arabia were asked

what proportion of the population do you

think is obese

respondents replied 25

it's actually 75 percent

people in the uk were asked what

percentage of the population do you

think

is muslim they answered 24

it's actually 5 percent people in japan

were asked what percentage of the

population live in rural areas

they answered 56 percent it's actually

seven percent our perceptions affect our

behavior

perhaps even our voting preferences in

2010

facebook conducted an experiment by

randomly deploying

a non-partisan i voted button

into 61 million feeds during the u.s

midterm elections

that simple action led to an additional

340 000

votes a number that could swing an

election

facebook had shown itself to be an

influencer

the chinese government is cited as being

particularly prolific in the creation of

false accounts

and paying people to broadcast

pro-government messages

with the intention of changing their

perceptions

in september 2019 twitter removed over

900 accounts it believed were

established by the chinese government

which were deliberately and specifically

attempting to sow political discord

in may 2020 the bbc reported that

hundreds of fake

or hijacked social media accounts have

been pushing pro-chinese government

messages

about the coronavirus pandemic onto

facebook twitter and youtube

these were attributed to the chinese

regime's 50 cent army

these are hired professional trolls and

the nickname

suggests how much they get paid per post

micro targeting

is a marketing strategy that uses

people's data what they like

what they're connected to what their

demographics are so that they can be

segmented into small groups for content

targeting

leading up to elections detailed micro

targeting has been used to target voters

the institute of practitioners in

advertising the ipa

has called for a suspension of micro

targeting ads in political campaigns

due to the covert secret nature of these

ads that do not have to be listed for

public display

like other political advertising it

creates a culture

that lacks openness and can be

vulnerable to abuse

with the us 2020 election looming

twitter announced

we will not permit our service to be

abused around civic processes

most importantly elections any attempt

to do so

both foreign and domestic will be met

with strict enforcement of our rules

twitter also banned political ads and

deep fakes

while the actor faking content is not

new deep fakes leverage

powerful techniques from machine

learning and artificial intelligence

to manipulate or generate visual and

audio content

with a high potential to deceive

and an entertaining example that springs

to mind is the fabricated footage of

forrest gump

talking to john and f kennedy

level four is evolving but it's here

and at this deep level it is where the

vast amounts of data

collected from around our planet are

used to perform predictive analysis

the software mines and analyzes

historical data patterns to predict

future outcomes

by extracting information from data sets

to determine likely outcomes

many of our decisions are not based on

logic

rather emotions trust intuition

satisfaction and culture

all play a crucial role in persuading us

to buy a certain product

or make a particular decision this

provides the opportunity to say

commission a social media company to

predict the voting patterns from an

election manifesto

the manifesto could then be reworked so

that it derives the most positive

results

and the analysis could also identify the

key messages to be conveyed to

particular social groups

i do hope we don't reach the next level

which is where social media companies

are commissioned to change the beliefs

of populations

by manipulating the information we

receive and so

change our beliefs to order so in

conclusion i would say that

during the lockdown periods for covert

19 there has been a significant increase

in online activity

social interaction is part of our dna

and so social networking

is here for the foreseeable future when

cars were first introduced to our

highways

they featured body damaging vertical

shapes with sharp metal emblems

they were a new concept and it took some

years of public pressure to change the

laws

and make the cars safer changes included

having gently sloping fronts with

collapsible bumpers

and outlawing those fixed protruding

bonnet emblems

and also fitting airbags

social media is a powerful and rapidly

evolving

communications vehicle we cannot put the

genie back in the bottle

but we must learn to manage it better

you