,
24 tweets,
15 min read
Read on Twitter
Today a 21-pages document was sent to at least 22 members of the European parliament. The authors claim to „have clear evidence of US meddling in this EU law making process“ and that a full release will happen in the next days.
They are wrong and here is why. /1
They are wrong and here is why. /1
Good: They are transparent about how they came to their conclusion and what tools they used. Else this thread would have ended like the last one.
The false core claim: 23%-47% of mentions of anti copyright directive campaign hashtag mentions came from the US. /2
The false core claim: 23%-47% of mentions of anti copyright directive campaign hashtag mentions came from the US. /2
Biggest mistake: They presented location data by @Talkwalker without questioning it.
The used the full version, but the free version is enough to prove my point that the location data is not trustworthy.
Let's look at #SaveYourInternet as an example. 63k mentions in 7 days. /3
The used the full version, but the free version is enough to prove my point that the location data is not trustworthy.
Let's look at #SaveYourInternet as an example. 63k mentions in 7 days. /3
For the last 7 days only 10% of the mentions (this can be Tweets, blogposts and similar stuff; but most of it are Tweets) are from US accounts. The authors claim that most of the manipulation happened in June 2018, so I trust them that those numbers are true. /4
By clicking on the US, we can see the most popular mentions by that country.
First 14 results are Tweets by @Senficon. Member of the European parliament. Very influential, but definitely not an US account.
There is clearly a problem with the data. /5
First 14 results are Tweets by @Senficon. Member of the European parliament. Very influential, but definitely not an US account.
There is clearly a problem with the data. /5
@Senficon But the authors don't only claim, that US accounts took part in the discussion, but specifically point out that the majority of the US mentions came from Washington D.C..
In talkwalker it's possible to limit the search to a region. I choose Washington D.C.. /6
In talkwalker it's possible to limit the search to a region. I choose Washington D.C.. /6
Once I limited the results to D.C., I switched to the Influencer tab.
@Senficon MEP
@creativecommons US? not DC
@communia_eu Not DC
@YourMarkLubbers Unlikely
@createrefresh Nope
@fyeg Young European Greens
@edri No
@bitchute No
@yanatoom MEP
@EuropeanPirates Nope /7
@Senficon MEP
@creativecommons US? not DC
@communia_eu Not DC
@YourMarkLubbers Unlikely
@createrefresh Nope
@fyeg Young European Greens
@edri No
@bitchute No
@yanatoom MEP
@EuropeanPirates Nope /7
Maybe it's broken. What happens when I choose Vienna. Looks okay. Several entries from Austrian accounts.
Let's check Berlin. What is @Senficon doing here? I thought she was from Washington D.C.. Her German Tweets are shown.
I don't fully understand how it works. /8
Let's check Berlin. What is @Senficon doing here? I thought she was from Washington D.C.. Her German Tweets are shown.
I don't fully understand how it works. /8
Where are those Washington D.C. Tweets that had such an impact? In the document are examples of other Tweets, but I didn't see any that supposedly cam from D.C.. /9
Smaller mistake: Selection of hashtags.
They looked at the following ones:
#censorshipmachines
#saveyourinternet
#linktax
#stopacta2
#savetheinternet
#art13
But those aren't the most popular hashtags. At least not recently. /10
They looked at the following ones:
#censorshipmachines
#saveyourinternet
#linktax
#stopacta2
#savetheinternet
#art13
But those aren't the most popular hashtags. At least not recently. /10
I looked at over 400k Tweets.
Most popular hashtags: #artikel13, #uploadfilter, and #saveyourinternet.
The ones in the document may have been more popular last year though. /11
Most popular hashtags: #artikel13, #uploadfilter, and #saveyourinternet.
The ones in the document may have been more popular last year though. /11
There definitely have been US accounts using those hashtags. The same way Europeans use hashtags when talking about US politics. The internet has no borders (only firewalls) and people talk with each other globally.
That's not proof of „meddeling“. /12
That's not proof of „meddeling“. /12
Clarifications:
- The document was created by two authors.
- I decided to not name them because the document isn't publicly available yet.
- They are not associated with the EU.
- They are not associated with Talkwalker.
- They are part of a German lobby-group.
- The document was created by two authors.
- I decided to not name them because the document isn't publicly available yet.
- They are not associated with the EU.
- They are not associated with Talkwalker.
- They are part of a German lobby-group.
I am still trying to full understand the Talkwalker stats.
I entered the first Tweet of this thread to be able to compare the data.
They even display demographics, but I didn't find an explanation how they determine it. (I assume based on bio.)
I entered the first Tweet of this thread to be able to compare the data.
They even display demographics, but I didn't find an explanation how they determine it. (I assume based on bio.)
It seems like they count Retweets and Quote-Tweets, but don't display them. Which is understandable for Retweets, but not Quote-Tweets. Maybe Quote-Tweets aren't included. Then the numbers are even more off.
Clicking on a country shows an empty results box. With the exception of Germany, where my Tweet is displayed.
With the exception of Germany all countries either have mention count of exactly 10 or 20, which is another indicator that the data isn't reliable.
With the exception of Germany all countries either have mention count of exactly 10 or 20, which is another indicator that the data isn't reliable.
@HerkkoHietanen is categorized as a lawyer. Probably because he has the word 'lawyer' in his profile.
@V_Arakawa is categorized as a parent. Probably because she has the word 'daughter' in her profile. As a human I question that categorization.
@V_Arakawa is categorized as a parent. Probably because she has the word 'daughter' in her profile. As a human I question that categorization.
The authors published the content of the document on their website. You can now read it for yourself:
English version:
I didn't think they would publish it. They even kept the ridiculous population-to-tweets-ratio calculation.
I didn't think they would publish it. They even kept the ridiculous population-to-tweets-ratio calculation.
I believe, I now understand how @Talkwalker calculates locations.
- If an account filled out the location field, they try to match it to a region.
- If the field is empty or can't be matched, they use the language of the Tweet.
- If English it's Washington DC, if German Berlin.
- If an account filled out the location field, they try to match it to a region.
- If the field is empty or can't be matched, they use the language of the Tweet.
- If English it's Washington DC, if German Berlin.
Test by searching (talkwalker.com/social-media-a… free signup) and filtering by Region.
Filtering by Berlin shows accounts with German Tweets but no matchable location info. (Altenbeken probably not in their db)
Filtering by Nordrhein-Westfalen shows accounts with matchable locations.
Filtering by Berlin shows accounts with German Tweets but no matchable location info. (Altenbeken probably not in their db)
Filtering by Nordrhein-Westfalen shows accounts with matchable locations.
Thanks @DerNivel for pointing out that the Tweet language seems to be relevant.
@DerNivel It works for English Tweets the same way. I had to use a different search term, because the hashtags in the report didn't have enough volume in Virginia.
Washington DC: All accounts with English Tweets, but no proper location.
Virginia: Only accounts with matchable locations.
Washington DC: All accounts with English Tweets, but no proper location.
Virginia: Only accounts with matchable locations.
