group 4.3

Group 4.3 Final Research Paper

Data Journalism: Blog Post Assignment
Fon Somers
Lily Wong
Kenza Sabri
Tino Vanhanen
Sebastiaan Eckmann

Analysis of the statistics in the article Fake News and False Flags by The Bureau Of Investigative Journalism


Fake News and False Flags, a journalistic project conducted by the Bureau of Investigative Journalism and written by Crofton Black and Abigail Fielding-Smith, details how a British public relations company, Bell Pottinger, was involved in the creation of propaganda videos in Iraq. In this operation, the Pentagon financed the majority of the funds received by the PR company, over half a billion dollars, to create television commercials and news items that appeared to be created by Arabic TV. Although the main focus of this article is the interview with Martin Wells, the report also included several statistics and findings in regards to the subject matter which we will assess and critique the use of in this blogpost.

Several statistics and sources are included within this article:

The Bureau has discovered that between 2006 and 2008 more than 40 companies were being paid for services such as TV and radio placement, video production, billboards, advertising and opinion polls.
A 2015 study by the Rand Corporation concluded that “generating assessments of efforts to inform, influence, and persuade has proven to be challenging across the government and DoD.”
Bell Pottinger’s work in Iraq was a huge media operation which cost over a hundred million dollars a year on average.
A document unearthed by the Bureau shows the company was employing almost 300 British and Iraqi staff at one point.
The Bureau has identified transactions worth $540 million between the Pentagon and Bell Pottinger for information operations and psychological operations on a series of contracts issued from May 2007 to December 2011.
Lord Bell told the Sunday Times, but the firm would have made around £15 million a year in fees.
According to Tunnicliffe, the contract, which totalled $5.8m, was awarded after the CPA realised its own in-house efforts to make people aware of the transitional legal framework ahead of elections were not working.

Examples of areas that could have included some statistics:

Firstly, “Bell Pottinger’s output included short TV segments made in the style of Arabic news networks and fake insurgent videos which could be used to track the people who watched them, according to a former employee.”
Half a dozen former officials and contractors involved in information operations in Iraq were interviewed. However, from these interviews, no exact statistics have been concluded.
Moreover, the article does not provide any statistics on how many videos, documents or commercials Bell Pottinger actually produced. It is mentioned that it was a “huge media operation”, but without any numbers, it is hard to understand how many people were influenced by this media content.
The article is full of vague sentences such as “But the largest sums the Bureau was able to trace went to Bell Pottinger.” For instance, in this case, we are not given any sums and even if it is mentioned earlier that the Pentagon gave Bell Pottinger more than five hundred thousand euros, it is not compared to any other projects funded by the Pentagon, and thus it is not possible to infer how important this project actually is for Pentagon.
“The same month he arrived there were five suicide bomb attacks in the city, including one a suicide car bomb attack near Camp Victory which killed 14 people and wounded six others.” A lot of irrelevant information and statistics is provided. Not to say that these statistics would not matter, but they just do not add any value to this project!

In conclusion, although this article includes many numbers about the different operations that were held, no comparison or context is given which strips the information from any value that it may have. Even though many financial figures are listed and repeated, they are sort of just thrown in there without any further explanations. Moreover, no statistics about the actual media operation are presented, which makes the article largely superficial and unofficial.


Fielding-Smith, Abigail, and Crofton Black. “Fake News and False Flags.” The Bureau of Investigative Journalism, 2 Oct. 2016,


Assignment: Bot Facebook Account Experiment

Capture d’écran 2019-03-13 à 11.30.01

This week we created a bot account on facebook for the purpose of investigating the filter bubble. The aim of this experiment was to see if the interests, likes and shares of both our bot account and our personal accounts would have any influence on each other. To do this we first downloaded the ‘Facebook Tracking Exposed’ extension to track and collect our public data. We then created our bot account, George Fieldings. We wanted the bot’s likes and interests to be almost the polar opposite of our own so that any contrasts and crossovers we observe on either accounts will be blatantly clear. George Fieldings is a middle aged man from Maryland, USA. We chose to make a conservative, liking Facebook pages such as FOX NEWS and Breitbart as well as more general interests such as Thai food and basketball in order for him to seem as mundane as possible.

The first thing we noticed when comparing our own personal accounts and Georges account is that he definitely gets a lot more traction from seemingly random users. This issue might come from the fact that because we actually use our personal Facebook accounts, we have privacy settings in place in order to make sure that we are only communicating and sharing with the people that we intend to. However, Since George’s profile is completely public, he received a lot of friend requests and messages from random users who seem to have no connection to him, both interest wise and location wise. We can infer that some of these users seem to be bots themselves. Moreover, due to the fact that the account was just created, his timeline did not change regularly. Although he liked and shared publications and posts from pages that are updated frequently, that was not enough for him to have access to a dynamic or full timeline. Furthermore, the bot account does not have any sponsored posts or advertisements as we think that it has not been created for long enough and has not been active long enough to get a tailored and personal set of recommendations. However, we noticed that after liking a Thailand Travels page, Facebook has recommended a set of three suggested groups with Thailand as the main subject including Thailand Travel, Thailand, Your Pictures, and Excite Fan Club Pattaya. As for his friends suggestions, we noticed that it was a mix of random bot accounts and strangers as well as our mutual friends as we had added him as a friend on all our personal accounts.

As a group, we did not notice any particular change in our own timelines, publications or advertisements. The sponsored posts or ads remained the same using the ‘Facebook Tracking Exposed’ extension or not. We expected some posts regarding Trump and other conservative politicians to pop up on our timelines as those were the posts that our bot account interacted with. However, none of us found that anything related to the content that our bot account has interacted with showed up on our timelines. The only things we saw were the articles and videos that our bot account had shared as we were friends with him on facebook but no page suggestions, right-wing articles, advertisements etc. were observed.

Overall, this experiment was somewhat disappointing in the sense that we did not notice any crossovers between our bot account and our personal account. However we understand that this issue may come from the fact this account is very new and has only been active for a few days and we predict that if we had done this experiment for a longer period of time we would have seen better results concerning the exchanges and dynamics between the bot account and our personal ones as well as within the bot account itself when it comes to the sponsored posts and advertisements.

Assignment: Bot Facebook Account Results 

We have now been using the Facebook Tracking Exposed Extension to collect and analyse our data as well as our bot account as part of our experimental project for about two weeks.

So far we have not noticed any drastic changes. A few more posts with content that seem to relate to our bot’s interests and likes have popped up on our personal pages however, it was so far in between that we were not sure if they were correlated or not. We did notice that George Fieldings (our bot) received a lot more friend requests than our personal accounts. So far in our two weeks of setting up the account, we have received over 100 friend requests, 57 facebook friends and of those, 36 of them are females. What we found interesting is that 30 of them are Asian, which means that the algorithms noticed our activity concerning Asian restaurants and massage places. Furthermore, 15 of George´s Facebook friends are from the U.S. This is most probably a result of our interest in American politics and news agencies. We noticed quite a lot of friend requests from Thai women as well and can infer that this is probably the result of our bot liking Thai Travel pages. Additionally, we also noticed a change in the type of friend requests that our bot account has received throughout the duration of his time on Facebook. When we first created his account, we noticed requests from accounts that seemed like fake accounts, they were all made recently, and mostly consisted of one or two photos. However, as our bot account started interacting with the platform more and more and began sharing and liking content, we started noticing that the demographics of the friend requests changed into genuine accounts, mostly of women seeking interactions with George Fieldings.

After collecting our bot’s account data using the Facebook Tracking Exposed extensions, we proceeded to put them into a graph or chart in order to make them clear and readable. We used the website that allowed us to choose what characteristics we wanted to emphasize. In this case, we focused on two variables, the value which are the photos, videos, posts, events and groups, and the count which represents how many times they were used by our bot account. Furthermore, we added those numbers into a circle packing chart where the hierarchy was based on the values and the size of the circles on the counts. We noticed that there are no circles for the posts or events as our account has not shown interest in any events and has not posted anything. Overall, after breaking down George Fieldings’ activity log, we noticed that his interactions can be broken down into a set of percentages. 9.8% of his interactions revolved around photos, 61.1% were about videos and 29.1% involved groups activities.

Capture d’écran 2019-03-19 à 18.19.22

Capture d’écran 2019-03-19 à 18.18.53

After comparing the two accounts, one being the bot account and the other our personal account. The differences were quite evident as for our personal account we must take into consideration that we have certain privacy settings that we did not implement into the bot account. The influences of this are that more people can easily access and see the bot account and therefore it increases the activity and interaction levels. Unlike on the real account which is more filter towards preferences and personal (real) relationships. Secondly, in terms of activity – because the bot account is newer, it can be seen that it is more “on the radar”. Because Facebook wants to promote friends by suggesting things such as “just joined facebook”. Whereas, our personal account is more linked to people with things in common such mutual friends. Additionally, the bot account was in contact or more so accepted to more groups and more random groups. Whereas, the personal account was groups not recorded probably due to inactivity.


Capture d’écran 2019-03-19 à 18.19.33
Capture d’écran 2019-03-19 à 18.19.08
In conclusion, it can be affirmed that only a small development was noticeable on the Facebook feed of our bot George Fieldings. The types and amount of friend requests of our bot account changed during our experiment which led to a relation between the activities, likes, and demographic of the friend requests. With a perceptible difference, privacy settings and level of activity are the main factors explaining the contrast between the accounts as well as how ‘new’ the bot account is. It seems that the Facebook algorithm still needs to adjust the information processed by our bot in order to link the information to the correct people with the same interests. Our hypothesis is confirmed, however, a longer period of researching is needed in order to conclude a more concrete finding related to the Facebook algorithms in which more accounts should be taken into comparison.