Stage 1: Understanding the Situation

We had a Skype briefing and a rough introduction to the South Sudan issue.

Some keywords have been given:

SPLA, SPLM, Salva Kiir, Riek Machar, SPLM-IO, TGNU, Juba, Lost Boys of Sudan, #defyhatenow, #peacejam, r0g, valentina wol

There’re also some links shared on Slack for us to look at, such as a campaign to reduce hate speeches , a report of inciting words on social media, and so on.

The topic is very difficult for me as I don’t have knowledge about this issue and the resources are hard to understand.I tried to draw diagrams to help myself memorise and understand.

note1

Stage 2: Concept Development

fake_real note2

From the research, I found that fake news causes a lot of problems, and my initial idea is to show a fake news about South Sudan situation and compare it with the fact and look at the hate speech around that fake news.

However, this idea has been doubted as there is no concrete evidence to support any News report about South Sudan. Even one of the recommended readings from the design studio, “How To Use Facebook And Fake News To Get People To Murder Each Other”, is commented as not trustworthy.

So I come up with the second idea: to find an inciting news/post on Facebook and track how many times it has been shared. However, this proved to be extremely difficult and time-consuming as posts on Facebook update frequently, those posts were either deleted or been pushed down by uncountable new posts about irrelevant topics.

Stage 3: Technical Workshops & Implementation

In the studio session, Oliver showed the class how to use webscraper.io plugin to get data from a website, and showed us a few data visualisation tools. Since my idea about “fake news spreading network” is not working out, I tried to install the plug-in to my Chrome browser and started to look for the big influencers’ Facebook accounts and see if I can do any data visualisation about them.

note3

I realised although it is possible to make Facebook show the number of common friends between two people by manipulating the URL, it is difficult to get the exact data and it is not straight forward. As shown on the screenshots, there’s no way to see the list of the common friends.

The data extracting is again time-consuming. The maximum number of friends for one Facebook account is 5000, and Facebook will only display around 20 at a time, I have to keep scrolling and waiting to get the full list.Another problem is that one of the big figures Riek Machar Teny Dhurgon doesn’t have any Facebook friend, therefore the final result will not look like what I wanted. Due to the time limitation, I only manage to get 3 people’s full friends lists.

facebook_code

facebook_webscrapper

I then tried to find a tool to visualise the data as a networking map of thousands of people. I have chosen to use Gephi as it is free to download. I stuck for a few days and got Oliver to explain the logic to me. At the end, I used an online tool to remove duplicate user IDs and make the friends list files in the correct format.

note4

removeduplicate

However, special characters and non-English languages would not display properly in Gephi.

excel_allnodes excel_friendslist

gephi2

But at least I got the visualisation of how these people connecting to each other, and from the graphic, the audience can see how influential they are, and an inciting post shared by one of them can be viewed by thousands of people.

screen-shot-2017-02-06-at-16-48-42

I also tried out different settings to see how to better providing the information.

gephi1

screen-shot-2017-02-06-at-17-33-13

commonfriends

I noticed there are a few common friends for them, I tried to find out more about them on Facebook, but the same problem happened as before: there are lots of irrelevant posts and shares every day, I couldn’t find anything interesting to add value to this visualisation

Stage 4: Start all over again

I was told the networking map doesn’t fit the brief about “hate speech”, therefore I started again focus more on the end users’ interactions. After checking lots of posts, I found one post on Facebook that has over a hundred comments, and the original post on the website only got less than 10 relevant comments.

newsonfacebook newsonweb

I decided to look into those comments and to find out the difference between two platforms in terms of hate speech ratio.
The final outcome is a web page with visualisation and the data used.

Click here to see the website (password: lange)

web_3  web_5

Through working on the visualisation, I found that although there are more hate speeches on social media platform, the number of rational comments is also larger, and there’s no advertisement. This may due to anonymous commenting is not allowed on Facebook.

screen-shot-2017-03-05-at-22-49-56 screen-shot-2017-03-05-at-22-50-25 screen-shot-2017-03-05-at-22-50-38

Update:

On 6 March, when I check again on Gordon Buay’s Facebook page, his friends list is disabled. The data I collected may become more useful now if anyone would like to analyse his social network connection.

facebook_gordonbuay