Are you there Twitter? It’s me, Nina.

Nina Lutz
10 min readFeb 26, 2020

This was originally done for an assignment for Ethan Zuckerman’s class Fixing Social Media.

Hi, I’m Nina. I was born and raised in Arizona, USA. I am a cisgender, white passing woman in her twenties who primarily dates men. I speak English and Spanish and have a smart phone. I have no visible disability and at a US ladies size 12 I am not actively discriminated against for my size. I have no children and I am a graduate student at MIT.

Basically, I am very privileged and especially from a data and media collection standpoint — modern media is basically catered towards me.

I can look on almost any mainstream social network and see people who look like me. If you are in the United States, I encourage you to look at your Instagram Discovery. I bet you can find plenty of white girls my age on there. Look at the trends on Twitter. There are a lot of white and hypereducated trends and voices on the national level (after all, it is election season and Elizabeth Warren is affiliated with Harvard). Listen to Tik Tok — you’re going to hear English and Spanish quite often.

In entertainment I can count on actors and actresses that look like me and have similar conveyed values and aesthetics to me. In music I have a huge variety of artists that are singing narratives I could be a part of.

And while representation has improved, we are far from any sort of level playing field. More representative and diverse images are still not the default in a lot of cases.

The truth is: I can scroll to the point of blurring through any media and still see myself. Not everyone can say that.

Maybe I spend so much time on social media because I am attracted to these filtered representations of myself. Maybe I like curating my own chambers of digital space. Maybe my own crutches of cosmetics are something I constantly need. Maybe I just mindlessly scroll.

Either way, I do spend a lot of time on social media. Staring at stuff like this:

My Twitter and Instagram are both public @ninalikespi so if you need to find me there you can to see yourself what I am following.

I would consider myself a fairly average social media user for my demographic. Above average of some of my friends but definitely on par.

I consume more than I post — by a lot. Between February 12th and now I have only posted: 3 instagram photos (personal), 4 instagram posts (my lab’s account), 1 facebook post, and 49 tweets.

But from Febuary 12th until now I have liked over 2000 social media posts between Facebook, Instagram, and Twitter.

On average I like a total of 149 things a day. For every 100 items I interact with I might post about 2.

This is aside from the time I spend being mediated in general, which is also a lot. What can I say, I am a loyal millennial.

I am constantly glued to my computer and phone. I am a graduate student who does a lot of code and design work. I live on a public transit route. I listen to music. I message my friends. I am constantly emailing and slacking and all the other communication types. I share Instagram stories.

I kind of pride myself on being super responsive. It’s become a personality trait and something I think about a lot. This creates a bit of a positive feedback loop when I am on my devices.

According to tracking apps I used during the time period of Febuary 11th till February 25th, I spend about 15 hours a week on social media between my computer and phone. I spend an additional 8 hours a week on these devices on communication and messaging.

So, looking at my usage, it’s time to ask: is this really a bad thing?

I think some folks want to say yes. I think the narrative of these social media addicted millennials are attractive. Or some awkward MIT grad student who can’t socialize in real life and relies on their computer to navigate personal and professional life. Or maybe some basic Instagram girl who does fad diets and photoshops all her pictures.

But there is no conclusive evidence that this mediation is the entire issue. There are countless studies, but like any complexity, there is no one axis that you can normalize around.

There is a lot wrong with social media. There is definitely no denying that. From radicalizing young people to promoting unhealthy beauty standards with editing.

In looking at how much I use social media, I wondered, who am I following as a cumulative picture?

I chose to focus on my Twitter, since I find myself most active on it and I actively follow folks in the Academic Twitter realm. I considered my Twitter feed fairly diverse. I follow a lot of people who do interesting art and advocacy about diversity and inclusion in STEM.

So I decided to do an average of all the profile images that I follow on Twitter. This is done by computationally averaging images. It’s based off composite photography, which we will talk about later.

Image average of every profile avatar on profile.

Well, that wasn’t successful. So I took out all the people and found the average logo of organization profiles.

Average image of all the organizations I follow on Twitter.

After doing that and not wanting to explore more organization profiles, I decided to do the average person/face avatar. I will admit, I use this term loosely. I decided that any photo an individual is using as an avatar to represent themselves.

I find avatars fascinating. How we chose to obscure our faces with filters and props and cartooned version. Here are some rather abstract avatars I considered:

Given this definition, I took the average of all these avatars:

Average avatars

This felt strange to me, so I decided to separate out photograph and rendered images. These were the results:

Average avatars of photo based avatars vs rendered avatars (drawn, etc)

Photographs of course dominated the avatars so I started splitting that into different categories. First I considered photos that were obscured (ones where you couldn’t see the whole face) and photos that were black and white.

Average avatars of obscured photos and black and white avatars.

I then started to divide the profiles based off gender identity. I did the avatars for She/Her, They/Them, and He/Him identifiying individuals. Obviously there are errors as this is a homework assignment and not an advanced study.

But here are the results:

She/her, they/them, he/him identifying avatars

Here are some of the profiles I was averaging, for reference. These aren’t all of them, but just to give you a glance at my feed.

I then ran a facial recognition algorithm on the same datasets.

Results of the facial detection of she/her, they/them, he/him identifying avatars

Due to biases in the algorithm (and some pretty bad errors), but results were 77% of my data (plus the errors). If you want to know about that go to

An example of a very bad error…when algorithms think that Google is more of a face than 3 people. Oops?

I hand fixed this data (wouldn’t recommend for a larger data source but for my size it was doable). And then compiled the faces again.

Results of the facial detection with the additions of all the faces that the algorithm missed of she/her, they/them, he/him identifying avatars

The following diagram shows a good summary of all of these.

And well, let’s just say they all look a lot like this face:

It’s my face. My network looks a lot like me. And I would wager a bet that for a lot of you reading this, even if you don’t think it will, yours might too.

Of course, this method is flawed. A lot of profiles I follow weren't in this particular representation if there was no face representation. Furthermore, several people I follow, use renders and drawings that use a lot of black and white and bright colors that people just aren’t (no one I know has green skin, as far as I know). Or use photos where most of the picture is a bright light blue sky and just a small face.

But I refuse to explain away a trend. Even if the numbers below don’t encompass my entire feed, faces were still more than half of my feed (approximately 65%). So I did a breakdown of the face representation profiles that I used image averaging on.

Overall race breakdown of faces analyzed.

Most of my Twitter feed is white. By a lot.

There is also an interesting narrative in who is showing their face and who is obscuring and filtering it with renders and filters. But that is an entirely different study.

At the end of the day, I get where my white dominated feed comes from.

We are all products of our biases. I grew up in White and Latinx spaces, it makes sense these reflect in who I follow. I follow a lot of academia and a lot of folks who do things in the computer science and art/tech spaces. These spaces are predominately white.

Of course, diversity is multiaxis and intersection. I follow a lot more women than men. I follow a lot of gender nonconforming folk. I follow a lot of trans folk. But there is a major axis I am biased towards, and it’s an axis that determines so much of our systemic experiences in the world. And I cannot ignore this axis.

My feed doesn’t make me a bad person. It also wouldn’t make me a good person if it was this perfect pie chart. Performance is not action. And curation is only a form of performance.

But I think it’s important to confront and see what voices we are listening to. Who are we choosing to listen to, whether subconsciously or consciously.

When I was showing this to a friend, she laughed nervously “Omg I would never do this.”

Me: “Why not?”

Her: “Well, if your feed is this white, I don’t even want to know how white my feed is.”

Me: “Why not?”

She didn’t have a good answer.

I think that she should do this exercise or one like it (maybe not as in depth). I think everyone should.

We are never going to make strides in inclusion if we are unwilling to look at where you start. It’s kind of like starting to train for a marathon without knowing what your starting time is. Except this marathon can’t just be run on Twitter.

Because these biases predate Twitter. They predate the first photo, because photos themselves are rooted in them.

I can’t help but find the irony at my use of these average faces in this piece. A lot of people use these for similar veins of communication. But rarely is their history addressed.

Composite photography, particularly of faces, is largely credited to the cousin of Charles Darwin — Francis Galton. Look him up. He and subsequent photographers tried to use composite photography with faces and finger prints to identify a “criminal type”.

This echoes back to the fear that many share when it comes to profiling in facial detection and surveillance. And this fear has been a reality since the beginnings of photography.

I think our social media is a strange curation. The algorithms and recommendations reveal not only how we are being surveyed throughout our interaction but just the reality of our networks.

I like to think of data and computational methods as tools for telling narratives, not the narrative itself. And I think that social media could be a tool to tell us things about ourselves, but only if we are honest about the state of our feed.

This is an exercise and not a scientific study. Data is apt to error due to the amount of manual categorization. This is not meant to serve as a scientific source or any formal recommendations.

This data was collected by downloading publicly available profile pictures of the people I follow.

These profiles were sorted manually by me with the help of some Python scripts to speed up the process.

Gender identities and racial breakdowns are done manually based on profile feeds and personal knowledge. These are apt to error.

Nina M. Lutz is currently a graduate student at the Media Lab.

Where many computer scientists want to make computers think more like people, Lutz aims to use computers to remind us to think of other people. Especially people who may not look like us. Lutz’s methods in doing this reconsider the design and technology choices around the human face in exploration of human identity through technology.

As a first generation college student, she is passionate about education and combating inequities in STEAM and opportunities in academia. Lutz welcomes emails from students and mentees about research and education both in and out of the design and computation space.




Nina Lutz

Instead of making computers think like people, I want to use them to make us think about other people.