1 of 21

Data Science and Social Media

SPCP ComMonth 2021: How to Criticize the Media

Amber Teng

SPCP Batch 2013 �angelamarieteng@gmail.com

2 of 21

AGENDA

INTRODUCTIONS
WHY IS IT IMPORTANT TO THINK CRITICALLY ABOUT THE MEDIA WE CONSUME?
WHAT IS DATA SCIENCE AND HOW DOES IT AFFECT SOCIAL MEDIA?
RECOMMENDER SYSTEMS, INFORMATION RETRIEVAL, AND SOCIAL MEDIA
FILTER BUBBLES AND ECHO CHAMBERS
DATA PRIVACY AND ETHICS
ACTIVITY
Q&A

3 of 21

INTRODUCTIONS

*CAVEAT: this is a very big and widely discussed topic. During today’s talk, I aim to start a discussion and share preliminary resources rather than to comprehensively speak about all the implications and technicalities of data science, recommendation systems, and social media.

SPCP Batch 2013�Brown University, BA Economics, Archaeology�NYU, MS Data Science �Author, The Data Resource�Teaching Assistant, Data Science for Everyone�Research Assistant, Data Science Software & Services�Co-Founder & Co-President, NYU Women in Data Science�

4 of 21

Why is it important to think critically about the media we consume?

Technology advancement and data generation have skyrocketed--along with this, our reliance and dependency on media for information and increased
At a societal level, we generally trust media as an authority
In a world where we’re inundated with news and data, media literacy is key.
“Data is the new oil of the digital economy.”
In the age of globalization and hyperconnectivity, data and social media have become helpful drivers of efficiency, but they can also be equally harmful when misused

Source: https://www.wired.com/insights/2014/07/data-new-oil-digital-economy/ ; https://ecowarriorprincess.net/2020/10/key-takeaways-the-social-dilemma-documentary-netflix/

“In the past 50 years, as technology has made tremendous advancements, the power of the media has also gained increased influence as more individuals are able to access real time information with greater rapidity.
Increasingly, we live in a society dependent on this information and communication to perform our daily activities.
We all make decisions based on the information that we gather.
We live in an age where there are myriad media sources and, more often than not, these sources carry a bias on each issue.
As a society, we have put a tremendous amount of trust in the media as an authority on a broad range of topics, and we rely on them for accurate information on current events, entertainment, and education.
The media can be a helpful, but equally harmful, force.
The inundation of repetitive messages can develop and perpetuate negative perceptions and assumptions, as well as shape our opinions and beliefs—sometimes correctly, other times incorrectly.
https://diversityjournal.com/1645-media-diversity-and-negative-perceptions-and-assumptions/#:~:text=The%20media%20can%20shape%20our,%2C%20sexual%20orientation%2C%20and%20age.

5 of 21

Why is it important to think critically about the media we consume?

“In the past 50 years, as technology has made tremendous advancements, the power of the media has also gained increased influence as more individuals are able to access real time information with greater rapidity.
Increasingly, we live in a society dependent on this information and communication to perform our daily activities.
We all make decisions based on the information that we gather.
We live in an age where there are myriad media sources and, more often than not, these sources carry a bias on each issue.
As a society, we have put a tremendous amount of trust in the media as an authority on a broad range of topics, and we rely on them for accurate information on current events, entertainment, and education.
The media can be a helpful, but equally harmful, force.
The inundation of repetitive messages can develop and perpetuate negative perceptions and assumptions, as well as shape our opinions and beliefs—sometimes correctly, other times incorrectly.
https://diversityjournal.com/1645-media-diversity-and-negative-perceptions-and-assumptions/#:~:text=The%20media%20can%20shape%20our,%2C%20sexual%20orientation%2C%20and%20age.

6 of 21

Why is it important to think critically about the media we consume?

MENTAL HEALTH

DEMOCRACY

DISCRIMINATION

The # of countries with political disinformation campaigns on social media doubled in the past 2 years.

The New York Times

A 5,000 person study found that higher social media use correlated with self-reported declines in mental and physical health and life satisfaction

American Journal of Epidemiology, 2017

64% of the people who joined extremist groups on Facebook did so because the algorithms steered them there.

Internal Facebook report, 2018

Source: https://www.thesocialdilemma.com/

7 of 21

What is data science and how does it affect social media?

8 of 21

What is data science and how does it affect social media?

Data science lies at the intersection of computer science, mathematics, and domain applications
“Data science is an inter-disciplinary field that uses scientific methods, processes, algorithms and systems to extract knowledge and insights from structured and unstructured data, and apply knowledge and actionable insights from data across a broad range of application domains”
We all make decisions everyday, and each of these decisions involves selecting among a huge space of choices
Increasingly, our choices are informed by algorithms and data

Source: https://hackernoon.com/the-ai-hierarchy-of-needs-18f111fcc007 ; https://www.youtube.com/watch?v=r2I3IDKwyMw ; https://en.wikipedia.org/wiki/Data_science

9 of 21

What is data science and how does it affect social media?

Machine learning models can alter societal behaviors
Broader societal impacts of search engines and recommender systems
Data ethics, data privacy, consent, confidentiality
Fairness, polarization, diversity
Models and datasets are inherently biased

Source: https://www.thesocialdilemma.com/; https://dataresponsibly.github.io/documents/EDBT2017.pdf; https://dl.acm.org/doi/abs/10.1145/3306618.3314288

Machine learning models can alter societal behaviors - feedback loops, basically get the user to do what the technologist w
Broader societal impacts of search engines and recommender systems - what does it mean for the people using and building these products?

Individual-level monetization and “influencer” status via more clicks and popularity

Data ethics, data privacy, consent, confidentiality - who has access to our data, what are they doing with it? Cambridg analytica etc, do we know what we are putting out there in the interwebs
Fairness, polarization, diversity - how do these algorithms work and how do they affect our perception of society?
Models and datasets are inherently biased - in terms of building these models, how are we ensuring that we are contextualzing the data and models that we’re using
Machine learning models are not objective--they are built by people for a specific purpose.

10 of 21

Recommender systems, information retrieval, and social media

Basic Information Retrieval Problem:

A collection of documents
Queries
A notion of relevance (How well does each document match a given query?)
Objective: help users retrieve documents to satisfy their information needs (Eg. web search)

Basic Recommendation Problem:

Collection of documents
Collection of users
Observed interactions between documents and users
Objective: Estimate relevance of each unobserved interaction

Source: Search and Discovery Course taught by Professor Brian McFee, TA-ed by Guido Petri and Amber Teng, Fall 2020 https://newclasses.nyu.edu/access/content/group/cacae473-9f49-419d-9163-674e9d75a323/Week%2001/Week%2001_2%20-%20Information%20retrieval.pdf

Modern approaches use machine learning to optimize performance given past user behavior
These are based on ALGORITHMS
In the broadest sense, an algorithm is simply a set of calculations to be carried out, often to perform a mathematical function. Artificial intelligence (AI) is a very broad term that applies to the many different advanced uses of algorithms to mimic and/or replace the need for human intelligence. Unlike a simple fixed algorithm, AI uses a system of algorithms and can create or modify algorithms without human intervention through the process of continually optimizing for better and better results – often referred to as machine learning.
As the surplus of data generated by our digital life grows, corporations are increasingly building AI algorithms that draw upon this data to model our behavior, target us, and make complex business decisions. The promise of an algorithm’s objectivity has engendered our trust in these data-driven approaches, however when used outside of a purely mathematical context, algorithms reflect an instance of logic programmed by a human – logic that frequently reflects the individual bias or the interests of the company they represent.
IR: Objective: help users retrieve documents to satisfy their information needs (Eg. web search)

A collection of documents
Queries
A notion of relevance (How well does each document match a given query?)

Recommender: Objective: Estimate relevance of each unobserved interaction

Unobserved → meaning data you haven’t seen yet ; eg netflix of the shows you haven’t seen, which ones will you like

11 of 21

Recommender systems, information retrieval, and social media

Information Retrieval

Focuses on short-term user needs
Applications derive from search engines
Queries are assumed to be independent

Recommender Systems

Focus on longer-term user needs
Applications primarily in entertainment
Adapt to a user’s behavior over time
Queries are sequential and interdependent
Also uses other users’ behavior

Source: Search and Discovery Course taught by Professor Brian McFee, TA-ed by Guido Petri and Amber Teng, Fall 2020 https://newclasses.nyu.edu/access/content/group/cacae473-9f49-419d-9163-674e9d75a323/Week%2001/Week%2001_2%20-%20Information%20retrieval.pdf

The core ingredients of a recommender system:

12 of 21

Recommender systems, information retrieval, and social media

Questions to consider when building or using recommender systems:

Does the evaluation accurately reflect how the system is used in practice?
Do we have enough interaction data to reliably evaluate a model?
Do we have enough interaction data to reliably fit a model?
What sources of bias exist in the content?
What sources of bias exist in the users?

Source: Search and Discovery Course taught by Professor Brian McFee, TA-ed by Guido Petri and Amber Teng, Fall 2020 https://newclasses.nyu.edu/access/content/group/cacae473-9f49-419d-9163-674e9d75a323/Week%2001/Week%2001_2%20-%20Information%20retrieval.pdf

13 of 21

Filter bubbles and echo chambers

Benefits of Recommender Systems:

Can help users efficiently navigate large collections of items
The recommender can adapt to the user to be more efficient

But...

What happens when users adapt to the recommender system instead?
How does the recommender affect users over time?

Filter Bubbles: intellectual isolation resulting from personalized searches when a website algorithm selectively guesses what information a user would like to see (user information, such as location, past click-behavior and search history)

Source: Search and Discovery Course taught by Professor Brian McFee, TA-ed by Guido Petri and Amber Teng, Fall 2020 https://newclasses.nyu.edu/access/content/group/cacae473-9f49-419d-9163-674e9d75a323/Week%2001/Week%2001_2%20-%20Information%20retrieval.pdf

Recommender systems and search engines work by focusing attention on some items at the expense of others
In the ideal case, you only get relevant results
In the less-than-ideal case, the recommender creates a bubble around each user, limiting their exposure to certain items
A filter bubble is a term coined by the Internet activist Eli Pariser to refer to a state of intellectual isolation^[1] that can result from personalized searches when a website algorithm selectively guesses what information a user would like to see based on information about the user, such as location, past click-behavior and search history.^[2]^[3]^[4]
As a result, users become separated from information that disagrees with their viewpoints, effectively isolating them in their own cultural or ideological bubbles.^[5]
The choices made by these algorithms are not transparent.^[6]
Prime examples include Google Personalized Search results and Facebook's personalized news-stream.
The bubble effect may have negative implications for civic discourse,
The results of the U.S. presidential election in 2016 have been associated with the influence of social media platforms such as Twitter and Facebook,^[9]^[10] and as a result have called into question the effects of the "filter bubble" phenomenon on user exposure to fake news and echo chambers,^[11] spurring new interest in the term,^[12] with many concerned that the phenomenon may harm democracy and well-being by making the effects of misinformation worse.^[13]^[14]^[12]^[15]^[16]^[17]

14 of 21

Filter bubbles and echo chambers

Why is this a problem?

“users become separated from information that disagrees with their viewpoints, effectively isolating them in their own cultural or ideological bubbles” (Ie, you only hear what you want to hear)
Increased ideological and political polarization, extremism, cultural tribalism
Google Search, FB Feed / Newstream, 2016 US Election, 2016 Philippine Election

Echo Chambers: situations in which beliefs are amplified or reinforced by communication and repetition inside a closed system and insulated from rebuttal

People are able to seek out information that reinforces their existing views without encountering opposing views, potentially resulting in an unintended exercise in confirmation bias.

Source: Search and Discovery Course taught by Professor Brian McFee, TA-ed by Guido Petri and Amber Teng, Fall 2020 https://newclasses.nyu.edu/access/content/group/cacae473-9f49-419d-9163-674e9d75a323/Week%2001/Week%2001_2%20-%20Information%20retrieval.pdf

Echo chamber: when a recommender influences a user’s preference by repeated exposure
Filter bubble: The effect of limiting content delivered to users
a feedback loop refers to an endless loop of instructions that can be given to a computer that has no final step.
An internet user's past browsing and search history is built up over time when they indicate interest in topics by "clicking links, viewing friends, putting movies in [their] queue, reading news stories", and so forth.^[22] An internet firm then uses this information to target advertising to the user, or make certain types of information appear more prominently in search results pages.^[22]
This process is not random, as it operates under a three-step process, per Pariser, who states, "First, you figure out who people are and what they like. Then, you provide them with content and services that best fit them. Finally, you tune to get the fit just right. Your identity shapes your media."^[23]
Part of this is because big tech and social media companies are incentivized to keep you on platform
Confirmation bias is the tendency to search for, interpret, favor, and recall information in a way that confirms or supports one's prior beliefs or values.^[1]

People display this bias when they select information that supports their views, ignoring contrary information, or when they interpret ambiguous evidence as supporting their existing attitudes.
The effect is strongest for desired outcomes, for emotionally charged issues, and for deeply entrenched beliefs. Confirmation bias cannot be eliminated entirely, but it can be managed, for example, by education and training in critical thinking skills.

15 of 21

Data privacy and ethics

To build these models and recommender systems, we need data…

Unsolicited data collection
3rd-party data sharing
Unsolicited access by employees / developers
Exposure of sensitive information
Targeted advertisement (e.g. the “Target story”)
(Pricing) discrimination and profiling
Leaking data to other users
Users may not be fully aware of what data is collected and how it’s used

While Netflix, Hulu, HBO, Amazon Prime and other services employ recommendation algorithms, there’s an important distinction between streaming platforms and the social media and search companies we critique in the film: on streaming platforms, users pay for the content, and therefore the business model is more aligned with our best interests.
- https://www.forbes.com/sites/kashmirhill/2012/02/16/how-target-figured-out-a-teen-girl-was-pregnant-before-her-father-did/
how Target tries to hook parents-to-be at that crucial moment before they turn into rampant -- and loyal -- buyers of all things pastel, plastic, and miniature. He talked to Target statistician Andrew Pole -- before Target freaked out and cut off all communications -- about the clues to a customer's impending bundle of joy.
Target assigns every customer a Guest ID number, tied to their credit card, name, or email address that becomes a bucket that stores a history of everything they've bought and any demographic information Target has collected from them or bought from other sources. Using that, they looked at historical buying data for all the ladies who had signed up for Target baby registries in the past.
As Pole’s computers crawled through the data, he was able to identify about 25 products that, when analyzed together, allowed him to assign each shopper a “pregnancy prediction” score. More important, he could also estimate her due date to within a small window, so Target could send coupons timed to very specific stages of her pregnancy.
So Target started sending coupons for baby items to customers according to their pregnancy scores
Target knows before it shows.
“My daughter got this in the mail!” he said. “She’s still in high school, and you’re sending her coupons for baby clothes and cribs? Are you trying to encourage her to get pregnant?”
The manager didn’t have any idea what the man was talking about. He looked at the mailer. Sure enough, it was addressed to the man’s daughter and contained advertisements for maternity clothing, nursery furniture and pictures of smiling infants. The manager apologized and then called a few days later to apologize again.
On the phone, though, the father was somewhat abashed. “I had a talk with my daughter,” he said. “It turns out there’s been some activities in my house I haven’t been completely aware of. She’s due in August. I owe you an apology.”

16 of 21

Data privacy and ethics

The Attention Extraction Economy: technology platforms profit from the monetization of human attention and engagement

Surveillance Capitalism: mass surveillance of our online activity in ways that we are often unaware, and the commodification of this data for commercial purposes

Biases in Data and Modeling: datasets are inherently biased; the key is understanding how to account for, contextualize, and mitigate the effects of those biases

Trade-offs to Consider:

Accuracy / efficiency vs privacy
Accuracy vs transparency and model interpretability

In particular, some issues that data privacy and ethics could cause are the following
The attention extraction economy refers to technology platforms that profit from the monetization of human attention and engagement. This includes, but is not limited to, Facebook (which owns Instagram), Twitter, TikTok and companies like Google (which owns YouTube) that profit from keeping users hooked on their platforms because more engagement means more advertising dollars.

Metrics
This also affects how people get paid → more views, SEO algorithm etc

Surveillance capitalism is a term popularized by film subject Shoshana Zuboff in her book The Age of Surveillance Capitalism. It refers to the mass surveillance of our online activity in ways that we are often unaware, and the commodification of this data for commercial purposes. The unprecedented scale of the data collected by these online companies and its use to predict and influence our purchases, behaviors, and thoughts has made them some of the richest companies in the history of the world.

Trade-offs to Consider:

Accuracy / efficiency vs privacy - random recommendations that aren’t useful, vs useful recommendations but companies have your data and klnow who you are
Accuracy vs transparency and model interpretability - on the governance side, should governments care more about how accurate these models are for users, or do they want to be able to regulate and understand these models? Ie regular model management; models don’t always perform the way you want them to

17 of 21

So, what now?

Some questions to ask when consuming social/media:

Who is this intended for?
What assumptions does this content make about society, or about its audience?
How real and accurate is this information?
Why did I get this social media post recommended to me?
What data could I have given or shared on this platform for this to show up on my feed?
Am I okay with sharing this personal data to the public?
What values are presented?
What is the commercial message of this?
Are there additional political and social messaging latent in this context?
Do I agree or disagree with this message?
What biases persist in this post?

Be vigilant: Discuss, listen, and research
Be cognizant of your use of social media, and be aware of where and with whom you are sharing your data
Rebuild the system

18 of 21

In today’s age of hyperconnectivity and increased dependency on social media, how can we think critically about the media we consume and the algorithms that curate this content?

ACTIVITY

19 of 21

QUESTIONS?

20 of 21

References

A list of the references used for this deck can be found on my notion page:

https://www.notion.so/ateng2507/3c9271f4309246ceaf085935052b6b3e?v=2ec7d38ed2aa424bb43c14d3bda5d220

21 of 21

Thank you!

Amber Teng - SPCP Batch 2013 �Email: angelamarieteng@gmail.com �Book: https://www.amazon.com/Data-Resource-Emerging-Countries-Landscape/dp/1641372524 �Twitter: @ambervteng �LinkedIn: https://www.linkedin.com/in/angelavteng/ �Instagram: hambergur_fries