Analysing the filter bubble – The Polarisation of our Online Communities

 

‘Analysing the Filter Bubble – The Polarization of our Online Communities’

Introduction

The Data Journalism course gave an introduction to conducting data journalism and journalistic research by explaining various methods of collecting data as well as the numerous different ways to analyse and showcase it. Due to reading about various aspects that had been researched and reported, it gave us an example of how to carry out research of our own. When exploring ideas for the research project within the topic of the filter bubble, our group decided to create a fake Facebook profile with the intention to create a profile with extreme one-sided views. To achieve this, we befriended strangers to increase the strength of our intentional filter bubble. Adding to this, we decided to compare our new bot to a much older (in terms of the lengthier time period the account had been in use) account, which was less active and was created for personal and social reasons, with no political intentions, unlike our bot ‘Scott Campbell’. We were expecting a noticeable difference in the timelines of these two accounts as well as their filter bubbles, as one (our bots filter bubble) was created to be prominent from the start and the other one was self-made over years of personal activity built up. Through the process of our research project, we managed to gather and analyze a sufficient amount of data, although we eventually concluded with were not those that we expected or hoped for. This is further elaborated in our research report.

 

Method

To start off with this journalistic research, we began with downloading and installing the software ‘Fbtrex’ on a Firefox browser on one of our computers ‘Fbtrex’ is software which collects and registers data taken from the Facebook platform. It was great to interact with this software, as it was something that none of us had the opportunity to work with before. The blue/green colour that the facebook bar became while using ‘Fbtrex’ made it easy to differentiate when one was using the software and when the data was being recorded and when it was not. Additionally, it was a way of differentiating between the private posts and the posts that could be collected for data; It made it simpler to understand what type of data would be recorded for our project and excluded those that weren’t automatically. After becoming familiar with the software, we began to create a fake account on Facebook with which we would use ‘Fbtrex’ to collect data. Daily liking and sharing posts helped us gather information for our final report as well as growing our filter bubble As well as this, we added people on our fake account, those who shared like-minded views. We joined groups, which we hoped would help us grow our filter bubble to a great extent. We were hoping for the filter bubble of our fake account to strongly oppose the one of our real account, especially politically, in our order for us to formulate a strong comparison. Through conducting the research over the space of a week we realized that it would not be possible to come to an accurate conclusion in comparing the political and social filter bubbles of our two facebook accounts given such a short period of time, as well as having two disparate accounts to compare, with a noticeable difference in an accumulation of a natural filter bubble and social sphere from our real account and the lack thereof in our fake account.

The first account we created had been deleted, due to Facebook registering our excessive activity as suspicious. We liked too many pages and added too many people in a short period of time. The second attempt at creating a fake account, however, was a success. Our fake account was a man named Scott Campbell who was born in 1971 in Champagne, Illinois, The United States. He then moved to Oregon and considers himself politically right-winged. He is a Trump supporter, thinks climate change is a hoax, is anti-gun reform and is anti-immigration. For the first attempt, we used a profile picture, but for the second one we decided against this, so Scott didn’t end up using a profile picture on Facebook. Given the made-up personal information of Scott, we decided to follow pages such as the ‘Donald J. Trump’ Facebook page. We sent some requests to join anti-climate change Facebook communities and other communities that believed climate change is a hoax. We also started to follow and like right wing related Facebook pages called ‘Right Wing Memes’,’ I Am Pro-Life’ and ‘Right Wing News’. We liked and shared some posts from the Donald Trump Facebook page and befriended some like-minded people we found through joining these Facebook groups. To be able to enter certain facebook communities we had to answer some questions like, “Why do you believe that climate change is a hoax?” or “Is the left using Global Warning to further their own agenda?”. After being accepted into these Facebook communities, more friend requests came in and we added some more people. We followed some right wing news facebook pages like Breitbart, Fox News and to create some diverse content, we also liked the facebook page ‘Fox Sports’. We logged onto Scott’s Facebook account daily, actively liking posts and accepting friend requests. At one point, he had over a hundred friend requests, which we found bewildering. We added only the people that had mutual friends and declined some requests. Interestingly, because we accepted more friend requests, Scott’s timeline showed more personal posts than posts from public Facebook accounts. We liked and shared some more posts from the public Facebook communities, however, to find these posts we needed to go to the communities page because Scott’s timeline wasn’t showing any, which is an important note to consider when analyzing the results we concluded.

 

Data from the accounts and analysis


Figure 1: Scott Campbell’s graph on his data collected from Facebook

 

Calculation:

Content

Count

Percentage

Event

0

0.00

Post

9

10.84

Videos

10

12.05

Photos

30

36.18

Groups

34

40.96

Recorded

83

100


Figure 2: Karoliina’s graph on her data collected from Facebook

 

Calculation:

Content

Count

Percentage

Groups

8

8.33

Events

10

10.42

Videos

20

20.83

Posts

43

44.79

Photos

15

15.63

Recorded

96

100

 

In order to get a set of data to compare, the Fbtrex extension allowed us to download all of the information it had gathered from these two timelines. This data was then categorised and made into tables and graphs in order to show the distribution of content on the fake account’s and the personal profile’s Facebook timelines.

Scott’s data from his Facebook page (Figure 1) showed that most of his content was group related, followed by photos. He was a member of two ‘anti’ climate change Facebook groups and the data showed that the posts that were shared within these groups, were dominating his timeline. However, we cannot tell if these posts were pushed because of the algorithm since private posts are not generated by Fbtrex. On top of that, Scott did not have many friends on Facebook, which showed on his timeline as less diverse posts were displayed. As seen in figure 1, the difference between the two highest percentages (posts including photos and group posts), and the three lowest percentages is around 20. However, when we were looking at Scott’s timeline, his timeline almost exclusively showed personal and private posts rather than posts from political right wing Facebook he was following. This data alone didn’t allow us to conclude whether Scott found himself in a filter bubble with a bias towards right-wing social and political views, as not all the content we were trying to get on the timeline showed up.

We then decided to compare Scott’s data to one of our personal accounts (Karoliina’s) data, who considers herself left-leaning but not politically active. We wanted to compare the difference in filter bubbles by having used the software Fbtrex. The personal account differed largely in the content it received on its timeline. Karoliina could be considered a ‘lurker’, a person that looks at what other people are posting, liking and sharing, but doesn’t post anything themselves. As it was an account that had been used for almost 10 years, Facebook had time to gather various personal data from different services and web pages, creating the opportunity for general and personalised ads and content. The advertisements seen on the personal account were mostly related to food, beauty, school, and shopping, all relevant topics for the owner of the account. In comparison to Scott’s account, this one had a lot more events on the timeline, as the friends on the personal account were people that attended events more often, while Scott’s friends were less in terms of numbers and not as diverse. Out of the “posts” section, about 60 per cent of it was promoted content, then content from liked pages, and only then posts from personal accounts, showing how Karoliina’s account had friends that tended to not upload basic posts on Facebook, but rather share or comment on content. The general view of Karoliina’s timeline was also a lot more diverse, as it had friends and pages accumulated from longtime Facebook usage, meanwhile, Scott’s account was made to focus on only one type of content/viewpoint. However, the personal account still had a filter bubble of its own, as these were the people and pages that Karoliina had chosen to befriend and like, therefore it was still a timeline filled with content that went with her personal interests and beliefs. We were not able to make a comparison regarding the political aspects as we had planned since no political posts showed on Karoliina’s Facebook timeline.

 

Ethics

Throughout this course, we were taught ethics in the realm of journalism. The do’s and don’ts when investigating and writing, and how one can remain unbiased and present the facts as they are, free from personal opinion or misrepresentation to favour one’s own stance.  Week by week, we have been shown examples of how ethics should be approached, from case studies like ‘what data can and cannot do’ (The Guardian, 2012) and the talk given to us by journalist Coen van de Ven from the ‘Groene Amsterdammer’.

We learnt the power of ethical choices in journalism in the first few weeks of the course and stemming from those examples we were given the opportunity to put these ethical practices we learnt to our own devices in the form of our group research project, which consisted of researching evident filter bubbles on Facebook while using the software ‘Fbtrex’. As mentioned previously, we began to set up a fake Facebook profile or a ‘bot’. To make our bot both realistic and integrate into the general Facebook community, we had to source our profile photo from the web, using a stock image site. This, in its own right, would be considered unethical, and when this first bot was shut down by Facebook for suspicious activity, in the interest of ethics, we made the decision to use no profile photo for our second bot. Following this, we then proceeded to add people, create fake information concerning the background of our bot (Scott Campbell), and join groups we thought our bot would hypothetically support. All of these actions come into question in ethical terms, as the first principle of ethical journalism is to be truthful. As well as this, ‘real’ Facebook profiles began to add us and attempted to interact with us regarding our political and socialist views that we presented our bot as supporting. Reflecting upon this, it is perhaps concerning that we would lead others to believe we too were ‘real’ and held actual social and political views. Yes, our project was for the sake of research, but when we compromise others and their information for the sake of research, are we really keeping true to our ethics? True, the filter bubble is evident but in our opinion, it would have been possible to conduct this research without the need to involve strangers and their own personal data, which they have not consented to. Overall, we have learnt the principles of ethics in journalism and how to ensure that you do not stray from those ethics but in the practice turn, we did in fact, stray from these principles and if we were to conduct this research again we would have to reevaluate our actions whilst conducting research and stay true to the truth.

Conclusion

After conducting our research, we came to the conclusion that the gathered data and analysis didn’t show what we originally expected it to. Even though the bot’s account and the personal account differed when it came to their timelines, we weren’t able to create the filter bubble that we wanted for Scott. The two account’s timelines differed in which type of content was depicted, but we unable to compare them when it came to the political aspect. As we mentioned throughout the report, Scott’s account was created with the specific purpose of creating a certain type of filter bubble. Therefore, we consciously added friends with strong right-winged leaning views, we joined groups that would agree with Scott’s hypothetical views and liked pages with the same kind of content. This is unlike the filter bubble of Karoliina’s real account, as hers has accumulated over the space of years, naturally building as she liked pages that coincided with her own personal views and added friends she met as she went through life.  Scott’s account was used for a time period of a week, not allowing the filter bubble to develop to a full extent. For a more accurate comparison, the fake account would have to be kept active for longer to gather more data.

The personal account, although not as obvious, also had a filter bubble which was discovered through this research. Even though Karoliina was much less active on Facebook (no posting or sharing, occasional liking of posts), there was a filter bubble which was promoted by her Facebook friends, the pages she liked, and the sponsored content she interacted with; This content confirmed her own personal views and interests. Therefore she was only exposed to the views of her Facebook friends who are often similar to her in either personality, interests, or social background.

In conclusion, we did not get the filter bubble we expected with the fake account but we were able to compare it with a filter bubble that at first, did not seem to exist but was discovered through the data we gathered with the ‘Fbtrex’ tool, helping us categorise and analyse the content of the two opposing Facebook profiles. The research showed us how easy it is to create a specific type of filter bubble on Facebook, especially one with strong political views, but how this content is not always immediately displayed on timelines, and that almost every account has a filter bubble to a certain extent. Conducting this research has made us even more aware of the polarization occurring online everyday, affecting us all and making us blind to the fact that there are those with views other than our own, and contributing to the wide-scale shock felt when one’s own affirmations don’t line up with the outcome of major political events. To counteract this growing societal issue and create a space for public debate to thrive, Facebook itself must take responsibility for this and act by changing its algorithm and making conscious decisions towards a more transparent and democratically led platform.