Advertiser profiles: by the Numbers

A lot of people seemed to gain insight from the advertiser profiles activity last week.

I had already seen how Facebook profiles my interests while poking around my account settings one day, but it didn’t startle me because I associate Facebook with a distinct lack of privacy. I go there to talk to family and scroll past people’s endless baby pictures; Facebook isn’t exactly the cool kid hangout it was the year I joined (2009, if you’re wondering). So I planned to just write about that. Only out of curiousity did I look at my profile on Twitter, which turns out to be shockingly in-depth and accurate for a site I barely use. My profile contained 60 Interests. I didn’t really start using Twitter regularly until last summer when I took a course called Digital Storytelling that required its use.

I decided it was possible that it just looks like they’ve profiled me well because of confirmation bias—basically, I was looking to see if they knew a lot about me, and therefore I found that they seemed to— which led me to the idea of using math to show how much they got right.

Overall, I saw a lot of trends to consider before moving forward. I grouped the interests accordingly.

ENTERTAINMENT

  1. Action and adventure
  2. Celebrity fan and gossip
  3. Comedy (inexplicably appears twice on Twitter’s list)
  4. Humor
  5. Independent
  6. Indie spotlight
  7. Literature
  8. Mobile
  9. Movie news and general info
  10. Movies
  11. Music news and general info
  12. Reality TV
  13. Rock
  14. Rock/Alt
  15. Television

 

NEWSERTAINMENT

  1. Books news and general info
  2. Business and finance
  3. Business and news
  4. Business news and general info
  5. Financial news
  6. Politics and current events
  7. Foreign
  8. Gov Officials & Agencies
  9. Government
  10. News

 

BEAUTY & FASHION

  1. Dresses and skirts
  2. Fashion
  3. Shopping
  4. Women’s accessories
  5. Women’s bags
  6. Women’s tops

 

ARTISTIC HOBBIES

  1. Arts and crafts
  2. Drama
  3. Music
  4. Painting
  5. Photography
  6. Jewelry
  7. Jewelry making

 

CAREER

  1. Biotech and biomedical
  2. Physics

 

SCIENCE FICTION & FUTURISM

  1. Computer networking
  2. Computer programming
  3. Computer reviews
  4. Documentary
  5. Entrepreneurship
  6. Sci-fi and fantasy (appears twice)
  7. Science news
  8. Space
  9. Space and astronomy
  10. Startups
  11. Tech News (appears twice)
  12. Technology
  13. World

 

HEALTH

  1. Health
  2. Health news and general info
  3. Health, mind, and body

 

INCORRECT

  1. Dance/Electronic
  2. Electronic

 

The breakdown is as follows:

 

TOPIC Count % of Predicted Interests % of Interests Minus Repeats & Incorrect Predicted Interests Average
ENTERTAINMENT 15 26% 27% 26%
NEWSERTAINMENT 10 17% 18% 18%
BEAUTY & FASHION 6 10% 11% 11%
ARTISTIC HOBBIES 7 12% 13% 12%
CAREER 2 3% 4% 4%
SCIENCE FICTION & FUTURISM 13 22% 23% 23%
HEALTH 3 5% 5% 5%
INCORRECT 2 3% 4% 4%

 

I have to admit, I breathed a little sigh of relief when I noticed that three of my interests appeared twice in the same wording, and pretty much everything fell into basic groups. Twitter doesn’t seem to know that much about me. On the other hand, it shows the depth with which they know me that they can figure out how much I like things. The two incorrect interests appeared only once, while the most accurate ones appeared multiple times.

After deleting the incorrect interests, my initial reaction is that I’m not necessarily uncomfortable with the whole world knowing this information about me. Most of it is fairly innocuous, but it’s a little disconcerting nonetheless. I’m most uncomfortable with how gendered my interests may seem to an outsider. I do not consider myself stereotypically feminine, yet you would likely guess that I am cisgender, heteronormative female based upon these interests. Of 60 interests listed and 57 unique interests (that is, discounting those that were repeated), I found 13 were either stereotypically feminine or overwhelmingly interests of women (or at least interests not of men, if you want to consider their prevalence with other gender identities). Many of these were inaccurate, and their prevalence was disproportionate to my interest in the subjects. It made me wonder whether the information in my feed is curated by Twitter purely based on assumptions about people who check the little box that says “female” on their profile or seem to be female, without consideration to my actual interest in them, and therefore possibly biasing me toward certain interests and worldviews. To be fair, books and other traditional media have also been strictly gendered, causing the same possible problem, but the Internet supposedly existed to create a space that exposed everything instead of catering to people’s biased notions, as discussed previously on my blog.  

DISTINCTLY FEMININE Average of 22% of my interests.

  1. Arts and crafts
  2. Celebrity fan and gossip
  3. Dresses and skirts
  4. Fashion
  5. Jewelry
  6. Jewelry making
  7. Literature
  8. Painting
  9. Reality TV
  10. Shopping
  11. Women’s accessories
  12. Women’s bags
  13. Women’s tops

 

I also found that I am uncomfortable being pigeonholed into being interested in certain topics. For instance, while I love crafts and could very easily have looked up Jewelry DIY projects, I would not classify that as one of my priorities. Meanwhile, I saw nothing relating to my interests in social and economic justice, just what I would classify as “newsertainment”— general appeal news that relies on shock value and suspense to capture your interest, not the kind of informative news that raises awareness of real issues (e.g. CNN-type as opposed to ProPublica). Although I decided to remedy the problem by taking my example to heart and just following ProPublica and the related groups NPR and PBS, I am more disturbed by the possibility that Twitter is influencing me to be more interested in entertaining things that I will reliably click on than what I really should be reading.

Based on my actual priorities, I decided that the actual breakdown should have been something like this:

CAREER 30%

ARTISTIC HOBBIES 20%

BEAUTY & FASHION 15%

SCIENCE FICTION & FUTURISM 15%

HEALTH 10%

ENTERTAINMENT 5%

NEWSERTAINMENT 5%

Based on that, I deleted some interests. Here’s my updated list:

ENTERTAINMENT

  1. Comedy (inexplicably appears twice on Twitter’s list)
  2. Humor
  3. Independent
  4. Indie spotlight
  5. Literature
  6. Rock
  7. Rock/Alt

NEWSERTAINMENT

  1. Politics and current events
  2. Gov Officials & Agencies
  3. Government
  4. News

BEAUTY & FASHION

  1. Fashion
  2. Shopping
  3. Women’s accessories
  4. Women’s tops

ARTISTIC HOBBIES

  1. Arts and crafts
  2. Music
  3. Painting
  4. Photography

CAREER

  1. Biotech and biomedical
  2. Physics

SCIENCE FICTION & FUTURISM

  1. Computer networking
  2. Computer programming
  3. Computer reviews
  4. Documentary
  5. Entrepreneurship
  6. Sci-fi and fantasy (appears twice)
  7. Science news
  8. Space
  9. Space and astronomy
  10. Startups
  11. Tech News (appears twice)
  12. Technology
  13. World

HEALTH

  1. Health
  2. Health news and general info
  3. Health, mind, and body

And my updated breakdown:

 

TOPIC Count % of Predicted Interests % of Interests Minus Repeats & Incorrect Predicted Interests Average Ideal % Breakdown % Difference New Count % Breakdown New % Difference
ENTERTAINMENT 15 26% 27% 26% 5% -136% 7 19% -116%
NEWSERTAINMENT 10 17% 18% 18% 5% -111% 4 11% -74%
BEAUTY & FASHION 6 10% 11% 11% 15% 35% 4 11% 32%
ARTISTIC HOBBIES 7 12% 13% 12% 20% 48% 4 11% 60%
CAREER 2 3% 4% 4% 30% 158% 2 5% 139%
SCIENCE FICTION & FUTURISM 13 22% 23% 23% 15% -41% 13 35% -80%
HEALTH 3 5% 5% 5% 10% 62% 3 8% 21%

 

I plan to do more, but it’s a start. I hope it will make Twitter a more meaningful and productive experience for me, along with the changes I outlined in last week’s blog post.

The “Interests from Partners” info made me a lot more uncomfortable, as it contained some sensitive information, such as a very good guess of my parent’s ages, the groceries I buy, and the used cars I’m looking at, which I really don’t want out in the world on Twitter, much less on this post! Eek. Here’s where you can adjust those settings:

 

Then there’s the “Tailored audiences” for me, which blew me away farther. Apparently I am part of 3535 audiences from 740 advertisers. As I browsed the list, I saw a lot of things I am completely not interested in such as Wells Fargo (a company I despise) and the Koch brothers (people who I despise personally). It seems as if Twitter sends a poorly formatted list in small print just to make it a pain to read and analyze. They also provided the list in the form of usernames, which makes it that much less convenient to find out who is behind some of these names. It was a lot more than I dreamed when I came up with the idea of analyzing everything in depth with numbers. The only thing I can do with these is give a list of the companies I am absolutely not interested in purchasing anything from or being involved with.

@aiginsurance

@amazonkindle

@amctheatres

@americanexpress

@americasnavy

@badmoms

@balancebar

@bananarepublic

@bankofscotbiz

@bbcarabic

@bbcsport

@bbcswahili

@bbcurdu

@capitalone

@capitalonecb

@chickfila

@directv

@directvnow

@disney_uk

 

I found almost 20 of the first 4 letters of my Twitter advertiser list were things I would never be interested in, of a little over 200 items that were in those first 4 letters, making for roughly 10%. It really bothers me now that I cannot change this.

Even if you aren’t interested in doing this activity, I think everyone should take a look at this article that talks about the subject: http://www.businessinsider.com/twitter-how-to-see-interests-advertisers-profile-privacy-settings-2017-5

P.S. Just for fun, I also looked at what else Twitter predicted before I went and edited my Interests.

I have to say, I feel the age is “accurate” in the sense that I have a very wide range of interests and often prefer the company and interests of older people!

Leave a Reply

Your email address will not be published. Required fields are marked *