Group 1.4

Final Research Report

Link to our final research report


Written by:

Maaike de Lang – 11869402

Guansong Zou - 11752041

Bianca Lucini – 11631066

Terence Moore - 11652004

Asia Sidokhine – 11649542

Blog 3: The Filter Bubble of Jack Winslow

Our bot, Jack Winslow, is a 36-year-old American man who recently moved to Amsterdam and works as a restaurant manager. He is an avid Trump supporter based on his economic values. In order to reflect that on his Facebook profile, we made him join numerous groups such as ‘Donald Trump Supporters’. He made friends with other members from such groups and was soon connected with people who share the same interests and ideas as him. At the end of the first week, we realised his interests on Facebook were too politically-oriented and did not necessarily reflect what a real profile would look like in terms of variety of liked pages and joined groups. Due to this, we spent the past week trying to make his profile more complete, adding interests regarding sports, food and science, while decreasing our concentration on politics only. This has resulted in some changes which allowed for interesting observations.

Firstly, we noticed how hard it is to get out of a bubble once you’ve been placed in it. Jack had shown his support for Trump by liking numerous pages and joining groups, however after undoing some of these actions, we noticed his feed was still flooded in political content compared to the other elements he had expressed an interest towards. This can be seen in the graph below, politics was more than twice as present as any other topic in Jack’s feed.

Since we had added friends from Facebook groups such as Donald Trump Supporters, Jack was still receiving related content. We inferred two different ideas from this, the first being that since political views are more or less consistent throughout time, the algorithm may weigh politically oriented posts more than for example opinions on technology, as it keeps evolving at increasing speeds. And second, that it takes time for the algorithm to adapt to changes in interests and to forget previous activities, meaning that once one is in a bubble, it takes real effort on behalf of the user to leave it completely. Such an effort would include showing more varied interests and an openness to different opinions, which may sound simple but may be of great difficulty for people who are not aware of these bubbles, as they are invisible, making it close to impossible for them to not be affected by Facebook’s algorithm.

When comparing Jack’s profile to our personal one, we were able to reach additional conclusions. Contrasting to Jack’s Facebook activities, the personal profile we used for comparison had shown no interest in politics whatsoever and was therefore never exposed to any political content. From this it could be inferred that, although political content may be weighted more by the algorithm, it does not present such content unless there is an explicit interest. It is therefore not very likely to get stuck in a specific political bubble unless the user shows a clear position in their views.

Furthermore, although it is clear from our experiment with Jack that filter bubbles do exist on Facebook, we have also noticed how much harder it is to notice them on profiles that have been active for a longer time. Since Jack’s profile is only a couple of weeks old, there is less information about him and his interests but there are also less friends and pages that he could receive content from which affects not only the quantity but also the diversity of posts that can be presented on the feed. By contrast, the personal profile that it was compared to, is four years old. Not only has the algorithm had more time to adapt to the person’s interests, it also has a bigger diversity of content to choose from, as the user will have more friends, more liked pages and more groups he is a part of. As a consequence, although the user may find himself in a filter bubble, he is also presented to a greater variety of content, making it even harder for him to notice.

Finally, while this research has given us the chance to witness at first hand the power of algorithms, giving us some curious insights on the topic, it is still too early to draw any real conclusions. Jack’s profile is too young and too limited to allow us to make accurate deductions about the way filter bubbles work and how they may affect us.

Blog 2: The Online Life of Jack Winslow

Screen Shot 2019-03-12 at 8.04.29 PM

The bot we created on Facebook is called Jack Winslow (36) a true American man who just moved to Amsterdam eight months ago. As a manager of an American restaurant, he also is an avid Trump supporter based on his economic values, hence he joined Facebook groups such as ‘Donald Trump supporters’ (‘DTS’) and met friends who share the same interests with him. Besides following groups related to Trump and Tesla Motors, Jack also liked pages related to badminton and The Doors and Deep Purple. Personally, Jack is a divorced bachelor who, unlike his distant and troubled relationship with his father, tries to properly invest time in and raise his two children according to the values he deems most important for today’s society.

When we created Jack Winslow we kept in mind the traits we wanted to emphasis. After a few active hours online “Jack” joined the DTS group, he got several friend requests very quickly (4 hours in). We noticed that all the people who friended Jack were from this group and were very eager. Moreover, the first added friends caused a snowball effect of other mutual friends who want to add Jack showing a certain perpetuated filter bubble in which Jack was stuck in. Compared with my Facebook account where I only interact with close friends or talk to people when it is absolutely necessary, it is noticeable that people like Jack do not have a problem with talking to strangers who share the same (political) interests. Also, we expected that people who joined the group would be redneck, white, middle-aged American men. Surprisingly, we noticed that people who friended Jack were half male and half female, and the majority were rather colored and from outside of United States, but from countries that had positive connections to trump administration.

The freedom with which we created a seemingly stereotypical rendition of a radical, white American male seemed to introduce an interesting opportunity to experience online life from his perspective, however, in turn also introduced ideological and ethical issues that had to be overcome or that hindered us to continue an even deeper look into our online persona. Ethically, the problems of operating a fake person onan online platform, where a certain form of societal decency has translated into the virtual community, had presented a more difficult challenge than expected. Knowing that you are deceiving (expectedly) real people who really are expecting to become friends with Jack seemed to persist. For example, personal one-on-one communication brought on a sense of discomfort and insecurity even behind the anonymity of the web. Furthermore, ideologically the issues remains in the fact that the objective task, to create a persona the opposite of yourself, had caused a certain unease in the connections made online, whether through friend requests, ‘friends’ personally contacting our persona or the radically political posts that dominated the timeline of Jack Winslow.

By working with fbtrex, we could distinguish if news outlets shown on Jack’s account were original or shared by others and their origins. The most original news-outlet on the timeline was the Remnant newspaper, which is a traditional catholic extreme right paper wholly supporting Trump and his policies. This type of news origin is very expected for our persona and further perpetuated the filter bubble in which we placed ourselves.

We believe that Jack remaining an active persona on facebook is a considerable accomplishment even with the issues we encountered. However, the short time with Jack did not allow us to thoroughly play out  his developed online character. With more time, we would have liked to improve failures that became evident throughout this experiment. For example, a big question which we set ourselves very early in the process had been whether the same people would have send friend requests to Jack Winslow if he had not joined the DTS group. Such a radically ongoing topic has immediately placed our bot into a bubble, which was only perpetuated further by the posts and people that appeared thereof. We hypothesis that perhaps, even if we had decided to proceed with Jack being an avid Trump supporter, that we might have held back from specifically joining any groups and first elaborated his personal information to see what other posts, videos, people would appear on the bots page. In terms of the persona we created, the content we expected and saw coincided, our feelings towards the content however were unexpected. We believe we limited the data we could have received from our persona by giving him a very active political view.

Blog 1: Cross-border Collaboration: An analysis of the Panama Paper



The Panama papers article published by the International Consortium of Investigative Journalists (ICIJ) aimed to shed light on the culture of criminal practices within the world of corporate tax havens through the collaborative analysis of a massive cache of leaked data from the law firm Mossack Fonseca.

In 2011, a reporter at Süddeutsche Zeitung received a USB drive containing over 2.5 terabytes of records from a confidential source and shared them with ICIJ which started the largest cross-border journalistic investigation to date.

The leaked data included corporate tax records, ownership deeds, personal financial information and emails that investigators collated and analysed in order to make sense of the volume of information they obtained.

The documents included information spanning nearly 40 years, providing extremely detailed accounts of how dark money flows through the global financial system. Not only did this data present new insights into the world of offshore companies, it also made the deep relationship between the offshore world and criminality finally visible to the public.

The aim of the article was to make the public aware of the glaring discrepancies of tax laws in many nations and to demonstrate that these firms and individuals’ actions were disrupting the global economy. The journalists were not only trying to tell readers about the impact of the global elite’s actions but also to demonstrate that are not above public scrutiny and ultimately, accountability.

Ensuring Integrity

As the ICIJ combined many different data types in order to initially evaluate the validity of the information, but also to better visualise the story, their methods indicate that their findings depict reality. In a media landscape where it is increasingly difficult to ascertain the truthfulness of a news story, the ethics and political agenda of a publication should be called into question. The ICIJ as an organisation strives to be as politically neutral as possible and aims to uphold high journalistic integrity. It is an independently funded organisation that relies on public donations to ensure it can conduct investigations of this scale.

It is clear that the ICIJ is committed to ensuring open access to information and this is demonstrated by the support received by one of its key donors, the Open Society Foundation chaired by the philanthropist George Soros, that works to promote transparency and democracy among many nations. Despite receiving funding from many non-governmental organisations, the ICIJ insist that they “maintain a strict firewall” between donors and their editorial independence.

Critical Analysis

The article uses vague terms such as “suspect transactions” to refer to the dealings of Mossack Fonseca but does not make clear whether they are illegal or not. The use of that kind of language makes the paper seem more like an opinion piece as they do not substantiate many claims with sources and when they do the link leads to an error message. Throughout the article there are numerous sharp statements being shared, together with numerical figures, about individuals and their criminal behaviour. These declarations are never cited and are only sometimes joined with hyperlinks that support their claims.

The article is quite long and very packed with information. As a consequence, the general message is lost between small details and minor stories that do not contribute to the bigger picture. This can be seen in the article’s reference to Jackie Chan when he is not necessarily involved in any illegal activities. Although the article mentions that Chan was not breaking the law, it still makes him lose his credibility as they cluster him with drug traffickers, ponzi schemers and even a convicted sex offender. This may also be interpreted as a way of defaming Chan or simply it is poor structuring on the part of the editor.

It is certainly a momentous task to coordinate an investigation of this scale but in some ways it felt that the article falls short in offering safeguards against the corruption it has revealed. It does note that many of Mossack Fonseca’s practices do not violate corporate laws but fails to address what seems like the major concern, the failure of global economic policy and legislation.

Written by: Bianca Lucini, Terence Moore, Guansong Zou, Asia Sidokhine and Maaike de Lang