4 ms·

I will say a lot of the examples of false-positives for bots are true-negatives for BotSentinel. Many of the handles are deactivated or suspended, but here are

I will say a lot of the examples of false-positives for bots are true-negatives for BotSentinel. Many of the handles are deactivated or suspended, but here are some examples of some live ones. - @ErikLuczak https://botsentinel.com/profile/1077274080984862720 https://botsentinel.com/profile/1077274080984862720 - @pdmonna https://botsentinel.com/profile/1200756272942829575 https://botsentinel.com/profile/1200756272942829575 - @CTO1ChipNagel https://botsentinel.com/profile/1411284846 https://botsentinel.com/profile/1411284846 Some accounts get a "disruptive" score - @cj_panirman https://botsentinel.com/profile/719017787805577220 https://botsentinel.com/profile/719017787805577220 But another question is, do we care if these bots are actually automated, or whether they are individuals that behave like them. Manually tweeting a single, political hashtag 854 times in a six day period is a spammer, regardless if they are a human. The result is little different than if they had programmed a bot to do so. Another was spamming using an IFTT trigger. If I was on the twitter team, I'd have no problem if that person was flagged as a bot.

mike_hearn 5y ago
Of course we care. Claiming an account is a bot is very different to claiming it's a real but disruptive person. Claims of botting are ultimately always claims of a conspiracy (by the bot master). Many of these conspiracies are merely commercial in nature and so we don't normally use the term, but social bot claims are insidious because they're invariably assertions of political or geopolitical conspiracy. If it's not true it's deeply corrosive to real personal relationships, trust in electoral integrity, international relations and so on. Even putting the wider social effects of a false classification to one side, a disruptive person is not necessarily misrepresenting anything and can possibly be made to follow the rules of the community with temporary timeouts, notification messages, etc. They may not have realized they were being disruptive. A spammer already knows they're breaking the rules of the platform and don't care, so the countermeasures need to be different. Twitter themselves have pointed this out quite eloquently, and made it clear that they think BotSentinel is no better than Botometer: "What about tools like Botometer & Bot Sentinel? .... Binary judgments of who’s a bot or not have real potential to poison our public discourse — particularly when they are pushed out through the media ... the narrative on what’s actually going on is increasingly behind the curve."
- jszymborski 5y ago
  I'm not associated with the Bot Sentinel project, but I think one of their major strengths is how they communicate their results. See the description for "disruptive" accounts, the class before "problematic": > Accounts that are classified as disruptive, often engage in malicious tweet activity. Some disruptive accounts harass other accounts and use offensive language. Disruptive accounts periodically share misinformation and can frequently spam hashtags. Inauthentic accounts and toxic trolls regularly receive a disruptive rating, so we recommend you exhibit caution when interacting with disruptive accounts. And from their FAQ > I am a human being, so why am I rated disruptive or problematic? > Bot Sentinel doesn’t just classify bots, we developed Bot Sentinel to classify several types of accounts including accounts operated by human beings. All this to say, a great effort is made to ensure people understand that what is being classified is behaviour, not human-ness. To your point re: "...social bot claims are insidious because they're invariably assertions of political or geopolitical conspiracy. Surely, many humans behind accounts classified as "disruptive" or "problematic" are statistically not going to be physically in some warehouse or being paid as part of some state-sponsored psy-op. But it's hard to extricate the political and propagandist nature of groups of, often loosely coordinated, individuals who consistently and totally disrupt non-political discussion with literal thousands of messages. It wouldn't be accepted in any physical public space at the scale and frequency it is done online. > "A spammer already knows they're breaking the rules of the platform..." I would contest this point, and further would say that ignorance of the rules is not much of an excuse. Bans are often preceded by a suspension, which provide users with an explanation of the offending behaviour. I'm not sure how things like Bot Sentinel do anything but educate users on whether their behaviours are disruptive or not. Good faith actors will take that as constructive advice to change, bad faith actors will wear it as a badge of honour. > Binary judgments of who’s a bot or not have real potential to poison our public discourse Bot Sentinel does not claim to identify bots full-stop. I think this is a bit of willful ignorance on Twitter's part. The bit about "the narrative on what’s actually going on is increasingly behind the curve" is something I feel is particularly rich, because Twitter is financially incentivised to be themselves "behind the curve" when it comes to moderating behaviour on their platform, as disruptive accounts often drive engagement.

mike_hearn 5y ago

Of course we care. Claiming an account is a bot is very different to claiming it's a real but disruptive person. Claims of botting are ultimately always claims of a conspiracy (by the bot master). Many of these conspiracies are merely commercial in nature and so we don't normally use the term, but social bot claims are insidious because they're invariably assertions of political or geopolitical conspiracy. If it's not true it's deeply corrosive to real personal relationships, trust in electoral integrity, international relations and so on. Even putting the wider social effects of a false classification to one side, a disruptive person is not necessarily misrepresenting anything and can possibly be made to follow the rules of the community with temporary timeouts, notification messages, etc. They may not have realized they were being disruptive. A spammer already knows they're breaking the rules of the platform and don't care, so the countermeasures need to be different. Twitter themselves have pointed this out quite eloquently, and made it clear that they think BotSentinel is no better than Botometer: "What about tools like Botometer & Bot Sentinel? .... Binary judgments of who’s a bot or not have real potential to poison our public discourse — particularly when they are pushed out through the media ... the narrative on what’s actually going on is increasingly behind the curve."

jszymborski 5y ago
I'm not associated with the Bot Sentinel project, but I think one of their major strengths is how they communicate their results. See the description for "disruptive" accounts, the class before "problematic": > Accounts that are classified as disruptive, often engage in malicious tweet activity. Some disruptive accounts harass other accounts and use offensive language. Disruptive accounts periodically share misinformation and can frequently spam hashtags. Inauthentic accounts and toxic trolls regularly receive a disruptive rating, so we recommend you exhibit caution when interacting with disruptive accounts. And from their FAQ > I am a human being, so why am I rated disruptive or problematic? > Bot Sentinel doesn’t just classify bots, we developed Bot Sentinel to classify several types of accounts including accounts operated by human beings. All this to say, a great effort is made to ensure people understand that what is being classified is behaviour, not human-ness. To your point re: "...social bot claims are insidious because they're invariably assertions of political or geopolitical conspiracy. Surely, many humans behind accounts classified as "disruptive" or "problematic" are statistically not going to be physically in some warehouse or being paid as part of some state-sponsored psy-op. But it's hard to extricate the political and propagandist nature of groups of, often loosely coordinated, individuals who consistently and totally disrupt non-political discussion with literal thousands of messages. It wouldn't be accepted in any physical public space at the scale and frequency it is done online. > "A spammer already knows they're breaking the rules of the platform..." I would contest this point, and further would say that ignorance of the rules is not much of an excuse. Bans are often preceded by a suspension, which provide users with an explanation of the offending behaviour. I'm not sure how things like Bot Sentinel do anything but educate users on whether their behaviours are disruptive or not. Good faith actors will take that as constructive advice to change, bad faith actors will wear it as a badge of honour. > Binary judgments of who’s a bot or not have real potential to poison our public discourse Bot Sentinel does not claim to identify bots full-stop. I think this is a bit of willful ignorance on Twitter's part. The bit about "the narrative on what’s actually going on is increasingly behind the curve" is something I feel is particularly rich, because Twitter is financially incentivised to be themselves "behind the curve" when it comes to moderating behaviour on their platform, as disruptive accounts often drive engagement.