Read on Twitter
OK, I've run a quick @Botometer analysis on 72 accounts that liked this tweet (sorry, that's all I could collect due to Twitter API limitations).
Short answer: yes, there is some bot activity, but not much.
Ping @mjkendall @TheRealPBarry @DanielBleakley https://twitter.com/mjkendall/status/1323026511008419840
Short answer: yes, there is some bot activity, but not much.
Ping @mjkendall @TheRealPBarry @DanielBleakley https://twitter.com/mjkendall/status/1323026511008419840
I used the @Botometer API to collect bot scores for these 72 accounts and set a Completely Automated Probability (CAP) threshold to 0.9 (in other words, 90% of the time we get it right).
The following accounts are flagged as *suspected* bots: @MalkawiYaz, @Msa87079318, @1RAMROY
The following accounts are flagged as *suspected* bots: @MalkawiYaz, @Msa87079318, @1RAMROY
This diagram shows the distribution of botometer CAP scores for the 72 accounts analysed.
Key takeaway: most accounts are not bots, but around one fifth of accounts have a CAP score greater than 0.8. So, 80% of the time the model would be correct in classifying them as bots.
Key takeaway: most accounts are not bots, but around one fifth of accounts have a CAP score greater than 0.8. So, 80% of the time the model would be correct in classifying them as bots.
Conclusions I've recently tweeted results of similar bot analysis, and based on this I would say there's a bit more bot activity here than usual. *But* some caveats:
- it's only a small and non-representative sample
- @Botometer is by no means perfect
- more work is needed
Addendum: I'm collecting all the replies to that tweet now (it takes a while), and I'll run analytics on it later. This includes sentiment analysis, to confirm if the majority of replies are negative (as @mjkendall and others observed). Sadly, it's not possible to collect likes.
