group 1.1

Roses are still red, Violet is only more blue…

Two weeks since the birth of Violet Kracht, she’s begun to grow her roots in Facebook’s algorithmic soil. We’ve created this persona to essentially investigate Facebook’s algorithm, and as we’ve previously established, the outcomes were less than surprising.  Given our persistent dedication in maintaining her character and interacting primarily with content that would reflect that, it feels as though we’re the ones who tricked the algorithm into doing just what we wanted.

We’ve collected the data, analysed the data, and visualised it in a way that might help us understand the algorithm somewhat better, although (spoiler alert!), there’s no such thing as full data transparency. Although we’re working with raw data, as media students we know better than to see this as a perk, since “raw data is an oxymoron”. That means that this ‘pre-cooked’ data was already biased (by Facebook’s algorithm), to fulfil a specific purpose (namely, to categorise Violet’s personality in ways that mould her growing filter bubble), which our analysis only confirmed. Another spoiler alert: the filter bubble is real.

Our story on the existence of the filter bubble starts by looking at both Violet Kracht and our contributor, alias: Rose(’s), Facebook profile, and looking at what their most common form of content was (that is; posts, events, groups, photos and videos). Our first data visualisation (see Violet’s Content Type) very clearly shows that posts, at 72.7%, were the most common form of content that appeared on Violet’s Facebook newsfeed. This is confirmed by the second visualisation (see Violet’s Source Type) that outlines what the most common sources of content on Violet’s timeline were over the course of two weeks. Looking also at our third visualisation (see Violet’s Post Sources), it is evident that even the most common sources on Violet’s timeline posted content primarily in the form of posts (with the second most common source also having posted a video on Violet’s feed), thus supporting our first finding about posts being the most common form of content to appear on the feed. Hence, it can be argued that there’s a correlation between the most common form of content Violet interacted with, and the form in which the most common source posted on her timeline. This leads us to the assumption that because Violet regularly interacted with posts (be it in the form of liking or sharing them), the sources that most frequently posted on her timeline corresponded to her preference in form (posts). Thus we can assume that because Violet mostly interacted with posts, the algorithm also delivered content in the form of posts mostly.

Violet's Content Type: This circle packaging visual demonstrates the highest type of content presented on Violet Kracht's Facebook news feed.

Violet’s Content Type:
This circle packaging visual demonstrates the highest type of content presented on Violet Kracht’s Facebook news feed.

Violet's Source Type: In this sunburst distribution, the frequency of content provided on Violet’s Facebook feed can be identified.

Violet’s Source Type:
In this sunburst distribution, the frequency of content provided on Violet’s Facebook feed can be identified.

Violet's Post Sources: This circular dendrogram presents a visualisation of the post sources as seen on Violet Kracht's profile.

Violet’s Post Sources:
This circular dendrogram presents a visualisation of the post sources as seen on Violet Kracht’s profile.

Moreover, the most common sources; LGBTQ Nation, RTL Nieuws and Washington Post somewhat reflect Violet’s personal and political leaning. As a transgender woman, it’s no surprise that the most common source on her timeline was from LGBTQ nation, at 19.7%. Since Violet interacted copiously with LGBTQ community-related events, videos, photos and posts, it comes as no surprise that subsequent posts reflected the same interest— a red flag calling attention to the presence of the filter bubble. The second most common sources, RTL Nieuws (a predominantly central news source) and the Washington Post (a right-centre biased news outlet) somewhat reflected Violet’s leftist leaning political preference, mostly in terms of her interests, but it didn’t entirely reflect her politics. That could either be due to the fact that she’s a Christian (which may have influenced the way in which Facebook’s algorithm read her political leaning), or because Violet is not a very politically active person.

Looking at our contributor Rose’s data, it is evident that she mostly interacts with groups on her Facebook feed, as the grand majority of her posts came from groups (see Rose’s Post Sources). Either from male or female people, our contributor Rose’s Facebook interaction was primarily from, or with, content sourced by public groups. What we can deduce from this data, is that because Rose has a longstanding Facebook network (her account has been active for over 10 years now), it is reasonable that the majority of her posts involve groups and people she voluntarily associates with (through liking, sharing and joining). Regarding the filter bubble, it is much more disguised amongst the variety of posts Rose has interacted with over the years, that contributed to the formation of her personal filter bubble, thus making it harder to distinguish between the content that was simply presented to her due to serendipity, and which were the results of the filter bubble. It is furthermore very difficult to discern the motive behind posts because we did not log into her account in a controlled manner and made interactions for research purposes. 

Rose's Content Type: This circle packaging visual demonstrates the highest type of content presented on contributor Rose's Facebook news feed.

Rose’s Content Type:
This circle packaging visual demonstrates the highest type of content presented on contributor Rose’s Facebook news feed.

Rose's Source Type: In this sunburst distribution, the frequency of content provided on Rose’s Facebook feed can be identified.

Rose’s Source Type:
In this sunburst distribution, the frequency of content provided on Rose’s Facebook feed can be identified.

Rose's Post Sources: This circular dendrogram presents a visualisation of the post sources as seen on contributor Rose's profile.

Rose’s Post Sources:
This circular dendrogram presents a visualisation of the post sources as seen on contributor Rose’s profile.

In light of the different circumstances under which Rose and Violet’s Facebook profiles were created and maintained, a direct comparison of how the filter bubble influenced their separate newsfeeds would not be accurate. However, as our visualisations show, it is undeniable that the filter bubble exists, as it maintains the persona we created for Violet, and reflects Rose’s most popular interactions.

Authors Group 1: Elodie Behravan, Naina Parasher, Paula-Lilli Stahmer, Dogasu Sitil, Maya Spangenberg

References -

Go to page 3 to read our other blog-post : (Re-)Reading the Riots: A Statistical Analysis

Go to page 2 to read an analysis of the filter bubble through data collected via a fake facebook profile : Roses are red, Violet(s) are blue, Is your filter bubble really you?