,
27 tweets,
10 min read
Read on Twitter
I frequently complain about people tweeting graphs with no context or omitting important details. And graphs tend to get lots of diffusion with zero fact-checking. Graphs are easily misinterpreted & can be presented deceptively. Without proper disclaimers it's very reckless.
For example: a thread claiming @RealAlexRubi was pushing some Russian narrative at #UniteTheRight just b/c some of his live tweets went viral so his node was quite large for a time.
He claims Alexander was "bloviating" about antifa & ignored Heather Heyer's death. Alexander witnessed Heather Heyer's death and continued reporting from the scene afterwards. How many tweets are normal after witnessing a murder?
But I digress...
But I digress...
Here are some more graphs presented without any info or context. Screenshots of normal twitter interactions are presented as "attacks"
Most people have no idea wth they're looking at, the onus is on us graph creators to interpret clearly and responsibly.
Most people have no idea wth they're looking at, the onus is on us graph creators to interpret clearly and responsibly.
It's not just graphs, people tweet spreadsheets with zero context or knowledge of the geopolitics surrounding tweets they collected & make wild assumptions based on very loose "links"
Monoglots don't realize how much they miss not speaking the language.
Monoglots don't realize how much they miss not speaking the language.
So I've been sorting through the Twitter graphs and various claims made in @eoinlenihan's thread about alleged "connections" between journalists and antifascist activists. And I found some problems.
.@Timcast made a 20 min video about it. Although Tim says various times "take it with a grain of salt" his video is titled "Verified Journalists Exposed Working With Antifa And Far Left Activists" -it has 106K views since May 17. Eoin's thread has thousands of retweets & likes. 
RedState published an article titled "Journos from Prominent News Publications Found to Have Working Relationships With Antifa" that fails to explain what the "connections" mean - probably b/c Eoin doesn't explain what the connections mean & implies something nefarious happening.
I'm not going to share any of the links but I made some visual aids to go over a few points that need clarifying and also show some common things that are easily misinterpreted in graphs so (hopefully) we can learn how to fact-check graphs when they are tweeted like this.
We don't know what timeframe the data was collected but he says it's a graph of "social interactions" on Twitter, so tweets, retweets and mentions.
2 important caveats missing from the thread:
1. Tweets, retweets and mentions do not indicate a profound relationship between accounts
2. For obvious reasons, antifascist accounts retweet news reported by journalists who report on far right extremism
1. Tweets, retweets and mentions do not indicate a profound relationship between accounts
2. For obvious reasons, antifascist accounts retweet news reported by journalists who report on far right extremism
In the first tweet he assumes there are relationships based on public Twitter interactions. He presents 1.65% of the dataset as "most connected" - the "minimum of 8 connections" means accounts that have retweeted these journalists 8 times or more (in an unknown timeframe).
Again he presents 1.65% of the original dataset as "most connected" (connections are public interactions on Twitter). He highlighted some accounts & seems to have enlarged & nodes/labels & colored them w/ contrasting color for emphasis. Also he tweeted 2 different graphs here:
If you've never used network visualization software you wouldn't know the scale and color of the nodes, labels and edges can be all changed. Tweeting graphs without clarifying that gives the optics of influence & importance.
(if all nodes are similar, it's not as eye-catching)
(if all nodes are similar, it's not as eye-catching)
Journalists need to understand context of the data being presented, the tools used to present it, the limitations of those tools and if graphs are presented with none of that info - be highly skeptical as to *why* it's omitted
Here in the middle of Eoin's thread, he demonstrates a fundamental lack of understanding in how to interpret clearly labeled data and he refuses to make a correction. All tweets with errors like this should be deleted.
First Vigil aggregates publicly available court docs, if he has a problem w/ that data being public, take it up with the court system that publishes case info before convictions.
Frankly I think it is a privacy issue that court docs are online for people w/o convictions.
Frankly I think it is a privacy issue that court docs are online for people w/o convictions.
Reminder that police have tweeted out mug shots of protesters exercising #1A rights, before anyone was convicted of a crime and most were never charged with anything. citylab.com/equity/2018/08…
Far-right trolls and right wing media had a field day w/ those mugshots 🙄
Far-right trolls and right wing media had a field day w/ those mugshots 🙄
Ok back to Eoin.. he repeats claims that verified accounts are "highly connected" & "closely associated" based on a retweet graph of public interactions. Again, for obvious reasons, antifascist accounts are gonna retweet news reported by journos who report on far right extremism.
He's using these graphs to add weight to his central claim that journalists are working closely w/ antifascist activists. His other "evidence" includes screenshots with no names & reporting by verified journalists. Reporting on leaked info is something every journalist does.
He references @letsgomathias's report on IE members in the US military - those military members *posted selfies* hanging IE propaganda. huffpost.com/entry/white-na…
So...ongoing investigations were opened based on very credible evidence which was newsworthy and reported on. 🤷🏻♀️
So...ongoing investigations were opened based on very credible evidence which was newsworthy and reported on. 🤷🏻♀️
Finally I'll point out that he has ignored my basic questions about the timeframe of the data collected & method used to collect it and evaded others' questions.
And it seems the Al Jazeera interview he bragged about 2 days ago has been removed. Judging by Eoin's LinkedIn profile, seems like he works in education? No idea what his qualifications are to call himself "extremism researcher" but his alt/troll account was banned in 2018.
He just replied and contradicted his original representation of his dataset. It's a following graph, not a graph of "interactions" as he claimed throughout his thread.
He completely misrepresented the data. So... some info about following networks...
He completely misrepresented the data. So... some info about following networks...
Following networks are even more tenuous "connections" than retweet networks. For obvious reasons antifascist accounts are going to follow journalists who report on the far right and journalists who report on the far right are going to follow antifascist accounts that leak info.
"Based on follows not followers" - that doesn't even make sense? The fact that he didn't communicate that basic detail and represented the graphs as "social interactions" shows he lacks a basic understanding of what he tweeted.
