Research Findings: Confirming the Filter Bubble Effect
To be updated with each new participant study
*continued from this research plan
Step 1: Prove the Filter Bubble Phenomena
Participant 1 confidently defined herself as a left-leaning Liberal. She said that she follows various different publications on Facebook, most of them Liberal nature in content.
When asked to pick 10 top posts that grab her attention, she was very vocal in her reasoning as to why she chose each. Some of her reasons for rejecting certain posts were: “this one just sounds like a gloat.” “this one’s written by an obscure publication, and the content headline indicates that it’s extremely alt-right.” “This one sounds like a lot of the things I’d read in the past and I probably won’t read it.”
As opposed to reasons for the posts she chose to read: “This policy direction sounds interesting, I’d read it.” “This is clearly a Republican POV, but it sounds like a smart move (the post was about supporting a Liberal ideaology),” etc.
The 10 posts that she chose were perfectly bipartisan, 50% being more Liberal in nature and the other 50% Right. But among these 10, the 4 posts she chose to actually read were all Left. She explained that: “I tend to read posts that I know would leave me feeling positive about… I don’t want to read posts that are gloating, sound stupid, or negative in any way to my beliefs.” Result: 100% more likely to actively contribute** to own Bubble.
Participant 2 had difficulty defining herself on the spectrum of Left or Right-leaning, but ended up identifying herself as a more Left-leaning individual than Right. She wasn’t as vocal in her thought process as the first participant, but said: “I rarely follow publications, but I read what my friends have shared.” She chose her top 10 posts to be also 50% Left-leaning and 50% Right-leaning, and contrast to the first participant, chose her top 4 posts to be 50/50; 0% likely to actively contribute** to own Bubble.
Participant 3 defined herself as a definitive Conservative, and marked herself almost all the way to the right side of the L-R spectrum. She was very vocal in her reasoning and had a very clear idea of what kinds of content she’d engage with, and why. She ended up choosing 13 posts out of the 20*, 6 of them Right-leaning, and 7 more Left.
As she walked me through her logic behind choosing these posts, however, it became clear to me that she was much more level-headed and unbiased than previous participants in the study: “I won’t read this article, because it sounds like a click-bait; it’s an obscure-sounding publication, too, and I don’t want to waste my time, unless it’s from a legitimate source.” “If the headline doesn’t mention any specifics and is vague about what the content will be about, I’ll pass.” “Well, I don’t know much about the US-Cuba relationship, so I guess I’ll read up on it,” et cetera. She applied these logics for both Left-leaning and Right-leaning posts. Of her top 5 picks, 3 were Right-leaning and 2 Left; 20% more likely to actively contribute** to own Bubble.
*I said that choosing 13 out of 20 (and not 10) was perfectly fine, but later realized that it would complicate the numbers when tallying up how much she was leaning one way or another; hence, I’ll work with the percentage instead of hard numbers with this participant.
** “Active contribution” indicates the participant’s likelihood to actively engage in magnifying their own Filter Bubble by clicking on the “Like” button, commenting, or sharing a post that originates from their side of the political spectrum.
Step 2: Unpack the UI of the article post
To discover what specific elements really affect the readership and the wish to engage, I took the same posts and separated each post into three parts: 1. the name of the source (i.e. publication or a friend) and their commentary; 2. the photo of the article; 3. the headline and a short quote from the article; and 4. the bottom-most social engagement bar (likes, comments, number of shares). On a simple scale system, I labeled one end “Would engage” and the other “Wouldn’t engage.” I then asked the participants:
- If you were to lay each of the elements out on this scale system, how would you organize them?
- How are you deciding what to engage with? What factors are being considered in your decision? (for a more qualitative data)
Participant 1’s engagement behavior
- The Social Element — if her friend had shared or liked the post, it was much more likely to affect her decision to engage, than anything else.
- The Photos — had the second-most impact on her decision to engage.
- The Source — (and their blurb second) had the third-most impact
- The Title of Articles (almost on par with #3)
- The Public “Likes” — she didn’t care much for the number of “likes” that the post received from the general public (outside of her friends)
Participant 2’s engagement behavior
- The Headlines (title) of the articles — she reads the headlines first
- The Social element — She then looks to see who has shared that post
- The Source (publication) — she judges by the publication, not so much the blurbs or commentaries.
- The Photos — she doesn’t pay too much attention to the photos
- The Public “Likes”
Participant 3's engagement behavior
- The Social Element — she recognized that friends sharing had the most impact on her decision to engage. But she also mentioned that she is selective about who: “I appreciate that this person shared the post, but knowing her, I could gauge what kind of article it’ll be. I know that she is knowledgeable about her stuff, but because I know that we don’t agree on most things, that affects my level of engagement.” “My other friend also often shares stuff, and I deeply respect them because he studied law and has very interesting opinions.”
- The Source — whether the article was from a reputable publication or not played a significant role in her decision to read an article; she vocalized time and time again her skepticism towards alt-right or alt-left publications (i.e. Addicting Info, Occupy Democrat, etc).
- Title of Article — this was ranked very closely after the Sources. “If a weird publication had an interesting title, I’d still give it a skim.”
- The Public “Likes” — She said that this element could also play a big role, but only when the content is viral; as in over 1 million likes, because she wouldn’t want to miss out on something viral.
- The Photos — If the photos matched the content or didn’t look biased (no frowning, unflattering headshots), I’d read it. But most of the photos on these articles look fake, like they’re stock photos that anyone from anywhere could’ve taken.
An interesting insight from participant 3 regarding her top pick — the Social Element — was that although it would take a lot for her to share something (i.e. she strongly believes in a subject or cause that she wants everyone in her facebook social circle to read up on), she does occasionally “Like” a post in order to show her support.
She also mentioned that when a post’s topic sounds interesting, she sometimes clicks into the comments and reads the discussions being held regarding that post. But that when she tries to share that post, “the comment thread attached to that original post doesn’t come with the shared post, and it kind of bums me out.”
I wonder if there are other Facebook users that feel this way, and if I could draw out a pattern from these qualitative feedback — it would certainly lead to very unexpected and interesting design implications.