What I Get Targeted On In a Day

What I Get Targeted On In a Day

What I Get Targeted On In a Day

By Sofia Nikolaeva

The starting point of Sofia Nikolaeva’s video work is the fusion of two datasets requested from her social media. The first dataset features an extensive list of keywords generated and utilized by Facebook for the purposes of engaging relevant advertisements. The second is a dataset of all photographs Sofia has posted as Stories on Instagram over the past years. Bringing these datasets together, the project questions on what basis or input might the AI algorithms of Facebook have come to recognize such an eclectic list of personal interests?

What I Get Targeted On In a Day explores the interconnection between these two datasets and shows that this connection is much more than just the activity shared on Facebook alone. By analysing the list of keywords and nostalgically traversing through her photos archives a clear link begins to emerge. Suddenly the marketing keywords Green Colour takes on a new significance. Only once reflecting on the masses of images shared — and the prominence of green in her photos over the years — does she learn this about herself.

Perhaps every single keywords can be traced back to her visual histories? What I Get Targeted On In a Day presents this interconnection of keywords and images as a speculative Vlog — a single day that attempts to keep up with algorithms that even at this very moment continues to generates more & more keywords for targeting.

image

Research

Key terms:

  • Facebook advertisement policy
  • Marketing policies
  • Targeting advertising
  • Digital Conscent
  • Data possession / Data request
  • Facebook's list of presumed interests

By removing yourself from a certain category of interests, you do not allow those advertisers who want to show ads to people in this category to include you in their target audience. This does not affect the total number of ads you see. We may still be able to show you ads related to these categories if we deem them relevant to you. *

* (The official policy of Facebook regarding advertisement https://www.facebook.com/policies/ads/)

Data

Datasets that are part of this project come from Facebook and Instagram

Folders from Facebook Dataset
Folders from Facebook Dataset
Folders from Instagram Dataset
Folders from Instagram Dataset

Early stage of the project: ANALYSING ALL TYPES OF DATA

As the first step, it is important to go through all the data to understand what information Facebook stores and what types of this information as of most value & can be part of an interesting narrative.

Below is the summary of information types that at early stages of the project were considered to proceed with:

  • No data for the section of 'Events'
  • My primary location is the city of The Hague
  • The average age range of my online connections is described as 'Early Adulthood'
  • It is predicted that I speak English and that English is the preferred language of my online activity, however the set language is Russian
  • There are keywords based on which the ads are displayed to me online: journey, pain, organic product, money back guarantee, asia, aluminium, vegan, sunday, begging for sweets, the hague, transportation, fashion design, dublin, europe, snakes, russia, jewellery design / 165 keywords overall
  • My last identified location is (52.088187519762, 4.2734517699439), which is my home location
  • image
image
  • My identified timezone is Amsterdam
  • No data for the section of 'Other Activity'
  • My search history is looking for the same things over and over: bedframe*3times, drawers, table, tafel, bedframe, drawers, bedframe, drawers, bedframe etc.
  • My online activity is associated with locations where I wasn't as time of being online: Amsterdam, Rotterdam, Arnhem, Nijmegen, Apeldoorn
  • No data for the section of 'Trash'
  • No data for the section of 'Voice Recording and Transcription'
  • No data for the section of 'Places'
  • My profile still lists Russia and Moscow as my country and city
  • My online signature is default 'Send from a mobile device' and is in Russian
  • The location identified as a 'Labeled Location' is my home address
  • There are areas on the map identified as 'Viewed Areas'

Further into the project: FACEBOOK ADS + VISUAL DATA FROM INSTAGRAM

FACEBOOK DATASET 'MY INTERESTS'

Folder 'ads_and_businesses' stores information about advertisement and targeting in relation to online activity
Folder 'ads_and_businesses' stores information about advertisement and targeting in relation to online activity
Folder contents
Folder contents

Below is the screenshot of the list of 165 keywords Facebook generated and considers them as 'My Interests'. Based on those 'Interests', Facebook decided which advertisements are the most relevant to me

image

As the list is mostly in Russian, here is the table of all keywords translated and arranged into categories

Categorisation_of_the_ad_interests_.pdf46.9KB

INSTAGRAM DATASET: ARCHIVE OF STORIES

The dataset consists of 21 folders with images that have been posted as 'Stories' in a spam of around 4 years

Below are screenshoots from some of the folders with images / folder contents are arranged by months

Folder contains 95 images posted in April-May 2021
Folder contains 95 images posted in April-May 2021
Folder contains 100 images posted in June-July 2021
Folder contains 100 images posted in June-July 2021
Folder contains 47 images posted in October-November 2021
Folder contains 47 images posted in October-November 2021
Folder contains 69 images posted in August-September 2021
Folder contains 69 images posted in August-September 2021
Folder contains 45 images posted in December-January 2021
Folder contains 45 images posted in December-January 2021

Prototypes & Experiments

An outcome of Part 1 of exploring the connections between the datasets and upcycling the image waste

  • The video presents the fictional conversation between me and Facebook, where it informs me of their advertisement policy and the categories of interests that it was generated and subscribed me into
  • Fictional conversation with Facebook
    Fictional conversation with Facebook
Sketch 1: thinking how to introduce Facebook
Sketch 2: thinking how to introduce Facebook
  • The video narrates from the past, as the image layer / 'IG Stories' have long time deleted themselves and have turned into an archive
The script for the video

Continuation into Part 2: digging deeper / narrating from the present + upcycling the image waste from the IG dataset

The format of a Vlog is inspired by titles 'What I eat in a day' / 'What I do in day'

Vlog cover image: Sketch
Vlog cover image: Sketch
  • The idea is to use the format of 'In a day...' to show a regular day on life - this is one narrative of guiding the viewer through the day ('Present Time' layer) with a casual content easy to consume
  • The second layer / MAIN NARRATIVE is rooted in an Instagram Dataset and is an upcyling action for the image waste from the past
  • Below are some raw footage fragments for the Vlog
  • Morning part
    Morning part
    Day part
    Day part
    Day part
    Day part
Morning part
Morning part
Day part
Day part
Evening part
Evening part
  • Below are sketches of the Vlog, where 2 layers of narrative are present:
    • Narrative from the present - the Vlog itself
    • Data waste recycling - layered images from the IG dataset
    • Right part of the screen - Vlog, image waste in the middle, also this is the part that presents the categories of interests Facebook put  me it
      Right part of the screen - Vlog, image waste in the middle, also this is the part that presents the categories of interests Facebook put me it
      image

Outcomes