MeMo:KI – Twitter-Analysis
Who is tweeting under the hashtag #ArtificialIntelligence? Who is networking with whom? Every six months, we analyze Twitter communication with a special focus on the network structures between actors from business, politics, journalism, civil society organizations and private individuals.
Analysis of Twitter network structures on the topic of artificial intelligence
Artificial intelligence communication on Twitter
Who is talking about AI?
Who is networking?
Publication of the tweets
01.01.2022 – 31.12.2022
Number of tweets in the dataset:
151.352
1
What can I find here?
We show which actors (i.e. organizations and individuals) on the topic of AI are most frequently mentioned by other actors in Twitter discourse.
2
What was searched for?
German-language tweets (including retweets) containing the terms “artificial intelligence” or “AI” were searched for.
3
How was the data collected?
The Twitter Developer account for academic research was used for the download. The data was downloaded using the statistical software “R” (package RT Twitter V”).
4
Data protection
For data protection reasons, we have anonymized private individuals when presenting the results; the content of the tweets cannot be retrieved.
5
Do you know the actors?
Have fun analyzing the data!
On Twitter, actors (organizations or private individuals) set up so-called accounts with which they can disseminate content via the short message service and read messages from other Twitter users. Different voices can also be found on Twitter on the topic of “artificial intelligence. Some actors express themselves regularly, others only rarely, but they reach many people. There is often a direct exchange between the actors. Twitter offers individuals and organizations the opportunity to “tag” other users in their tweets and thus address them directly. We call this form of contact between different users networking, and the resulting entities are networks.
The complex networks that emerge in such discussions can be analyzed and mapped using network graphs. Depending on whether only strong or also weaker relationships are analyzed, more or fewer actors are included in the graphics. The representation thus becomes finer and more differentiated or remains coarse, but clearer.
The following applies to all illustrations: Connections are created through the function of the so-called @-Mentions (this means that accounts are directly mentioned or tagged) or a retweet. In Figure 1, you can explore a dense network. Here, we have mapped those accounts that were mentioned by at least 20 different accounts in tweets containing the terms “AI” or “Artificial Intelligence” in the six-month period.
Why is this important?
Via this narrowing down, we have filtered out such accounts from the representation that are assessed by others as not being that important. This is done in particular for reasons of clarity. This means that the graphics only show a small section of the actors that are addressed on the topic of AI.
The color of the accounts in the graphics indicates the type of actor involved, e.g. whether it is a political account (e.g. from a party) or a journalistic account. However, there can also be individuals behind an account who tweet privately and do not share information about their professional or occupational role. We have anonymized these for data protection reasons. However, there are also accounts of people who do not express themselves privately, but are understood as public figures. These can still be found in the network graphs if they are frequently mentioned by other accounts.
The complex networks that emerge in Twitter discussions can be analyzed and mapped via network graphs.
Network of accounts mentioned by at least 20 different accounts on the topic of AI.
Legend
leer
What is represented in the network?
In a network graph, two elements are represented: Nodes and edges. The colored dots represent the Twitter accounts, the so-called nodes. The larger a node is, the more often this account was mentioned or retweeted with an @mention. These references of different accounts to each other are represented by lines between the nodes: These are called edges. The wider such an edge is, the more often there are references between the accounts. In addition to the size of the nodes and the width of the edges, the position in the network is also significant:
The more often an account is mentioned by others of the mapped accounts or refers to them itself, the more central this account is in the network. However, if an account that is referred to frequently (large node) appears further out in the network, this indicates that the references come from accounts that are not mapped in the network.
What do the colors stand for?
The colors of the nodes indicate the type of Twitter accounts:
- Politics
- Economy
- Civil societies/NGOs
- Private individuals
- Journalism
- Science
- Public figures
- Other
Wann wurden die Tweets erhoben?
1. Halbjahr 2021
Zeitpunkt des Downloads:
30.07.2021
Veröffentlichungsdatum der Tweets:
01.01.2021 – 30.06.2021
Anzahl der Tweets im Datensatz:
79.505
2. Halbjahr 2021
Zeitpunkt des Downloads:
18.01.2022
Veröffentlichungsdatum der Tweets:
101.07.2021 – 31.12.2021
Anzahl der Tweets im Datensatz:
71.848
To make it a bit clearer, we have reduced the number of accounts shown in the following figures. In contrast to the figure above, the following network graphs appear less dense, as the network is now only represented by accounts that have been mentioned by at least 50 other accounts. In the figures on the left, you can see the current network structures from the 2nd half of 2022. Next to them, correspondingly, are the network structures from the 1st half of 2022.
Network of accounts mentioned by at least 50 different accounts on the topic of AI.
How do I read the network graph? What do the network structures tell me?
Who is talking about AI and how visibly?
The collected Twitter data can be used to depict network structures, i.e. to uncover networks of relationships between accounts.
1
What do I see?
Network structures! Namely, networks of relationships between individual Twitter accounts. This makes it clear who is talking to whom on Twitter via “AI.
2
How is a relationship established?
An account addresses another account directly (using @mention) or retweets it.
3
What are @mentions and retweets?
The quality of the interaction cannot be tracked, i.e., praising as well as critical words are conceivable in the tweets. Retweets, however, can be understood as an affirmation of the actor’s content.
4
What is an Indegree?
Indegree is the total number of accounts that refer to an account (retweet and/or mention) during the survey period. If an account was mentioned by 25 different accounts, then the indegree is 25.
5
What is a node?
What is an edge?
The accounts shown are called nodes in the network graph. Edges result from retweets or mentions. An account A addresses an account B and thus gives it visibility.
6
Who explains more to me?
All scripts for data collection and data analysis are available on request.
Who tweeted the most? Which accounts were mentioned most often? Who referred to others most often?
In addition to the network graphs, the following tables show which accounts published the highest number of tweets per half-year, which actors were referred to most often, and which actors referred to others most often. It is important to note that we have anonymized private individuals here again for reasons of data protection. They were not removed from the representation because this would mean that important information would be lost for the network. Private actors are a relevant group for discussions on the topic of “AI”.
Who tweets the most?
2022, 2. Half Year
Who tweets the most?
2022, 1. Half Year
Who is addressed?
2022, 2. Half Year
Who is addressed?
2022, 1. Half Year
Who addresses?
2022, 2. Half Year
Who addresses?
2022, 1. Half Year
Who tweets the most?
2021, 2. Half Year
Who is addressed?
2021, 1. Half Year
Who tweets the most?
2021, 1. Half Year
Who addresses?
2021, 2. Half Year
Who is addressed?
2021, 2. Half Year
Who addresses?
2021, 1. Half Year
How do I read the tables?
MeMo:KI in AI Discourse
Our project is also part of the AI discourse. Our goal? We bring our research results into the discourse via the discourse. Follow us: @_MeMoKI
1
Who tweets most often?
- Here you will find a list of the accounts that tweet the most on the topic of AI (incl. retweets).
- These tweets do not necessarily mention other accounts.
- The “number of original tweets” helps to see how many of the tweets are not retweets, but provide new content on the topic.
- The “average likes” and “average retweets” figures map how much (positive) attention the account’s tweets have received on average.
- We have sorted the table by the number of posts. You can re-sort the table according to your interests by clicking on the categories.
2
Who is addressed?
- In this table you can find the accounts that have been mentioned most often by other accounts.
- Important for this is the number in the column to “Indegree”, which shows how many different accounts refer to the represented account.
- The “Weighted values” indicate the total number of incoming and outgoing connections. This means that if one account refers to another multiple times, this is included.
- This table is sorted by “Indegree”, again you can sort by your own interests.
3
Who addresses?
- In the last two spreadsheets you can see which accounts are particularly active in networking activities by addressing other accounts.
- The number under “Outdegree” shows the outgoing relationships of an account.
- Accordingly, the table is sorted by “Outdegree”, but can again be explored and sorted interactively.
What insights do the network analyses from 2022 provide?
1
No fragmentation
There are hardly any groups that can be clearly demarcated; most of the actors are in direct contact with each other and there are few groups that discuss the topic separately from others.
2
Networking of different actors
In Twitter discourse, actors from different fields usually refer to each other. This means that scientists not only network with other scientists, but are often also addressed by actors from business, civil society or journalism. In 2022, however, it can be observed that individual actors are increasingly networking within their own group. This applies in particular to journalistic actors, as well as to the groups of civil society/NGOs and politics.
3
No actor:inside type dominates excessively
This is often different in media reporting. There, it is primarily economic actors who dominate. All types of actors can be found in the Twitter discourse. If the most dominant type had to be identified, it is most likely to be journalistic actors in 2022, as in the previous year. The tendency is for journalistic actors to continue to increase in Twitter discourse.
4
Actors particularly present for the communication of AI competence
These actors are also very well networked with each other, as can be seen from the strength of the connection.
5
Specialist media are addressed
Although trade media, which primarily deal with technology, are still the most addressed journalistic media, it can be seen that established media such as the WELT, FAZ or Der Spiegel are also increasingly participating in the AI discourse on Twitter.
6
Private individuals are part of the discourse
These often tend to be on the outside because they usually send out a high-reach tweet but have only been retweeted once by other accounts. There has been an increase in the active participation of private individuals in the Twitter discourse on AI.
7
Important note
In our analysis, we can only make statements about the network structure and no statements about the content of the tweets. Due to the current developments of the platform, we will probably not be able to make any further analyses.
Do you have further questions about the analyses and evaluations?
Research partnership
Sponsored by
