3 ms·

The standard of evidence you're requiring to call something a bot compaign is high enough that nothing counts. So sure. It's theoretically possible that all th

by ikeboy 6y ago

The standard of evidence you're requiring to call something a bot compaign is high enough that nothing counts. So sure. It's theoretically possible that all the apparent hints of impropriety are just coincidences and lots of people just forgot what they did and so on. But then that goes both ways. It's not like the evidence is significantly stronger on one side or the other; if you're inclined to explain away everything, you can do so. If you'd like, read my initial comment as saying something more along the lines of "there's some evidence of massive bot compaigns on both sides", or replace that with a weaker claim if you think none of this counts as evidence. But motivated skepticism of the evidence on one side only doesn't help. So: what evidence do you see that there was bot behavior on the anti-NN side that's so much stronger than any of the evidence on the other side? Everything I've seen can be easily explained away by the same standards you're applying.

geofft 6y ago
The standard of evidence I'm applying is simply that evidence for X being true must not also be plainly compatible with X being false. Here, for instance, is some evidence that meets that standard, from that very same article you love: > He found a near-constant rate—1,000 every 10 minutes—punctuated by periods of zero comments, as if web robots were turning on and off. He determined many were from hacked accounts. This pattern is either server-side (e.g., a server-side rate limit and an extremely flaky server) or client-side. If it's server-side, we'd expect to find it independent of the contents of the comment, we'd expect to have stories of people being unable to submit comments, etc. Unless we have those stories, then this pattern is plainly incompatible with actual humans submitting comments, even with the aid of some automation or pre-written letters. That is evidence that these comments were almost certainly not from actual humans. No similarly strong evidence exists on the pro-net-neutrality side. "A bunch of comments claimed an address from Russia" is certainly suspicious and weird, I agree, but it is not plainly incompatible with humans (if it's IP address, bad geolocation; if it's e-mail address, Russian-run mail provider whose accounts are used by Americans; etc.). I am happy to grant with more detail about what's is meant there - which neither the FCC nor the WSJ has provided to us - there may be strong evidence, but it's not there yet with what you've posted. If you want to revise your initial complaint to "Friendly reminder that some weak evidence exists in favor of small, isolated bot campaigns on the pro-network neutrality side and very large bot campaigns on the anti-network neutrality side, such that if all suspicious comments were disregarded in proportion to their likely untrustworthiness, the overwhelming weight of public sentiment as measured by this exercise would be firmly on the pro-network-neutrality side," then I have no remaining objection.
- ikeboy 6y ago
  >Unless we have those stories, then this pattern is plainly incompatible with actual humans submitting comments, even with the aid of some automation or pre-written letters. A site is set up to collect names for form letters. They don't submit it in real time; instead, they submit it in batches every few minutes. This is at least as plausible as the various explanations you gave above. In fact, if you read the linked blog post, they explain that this is how the API worked - you'd upload a CSV and everything would be uploaded in bulk. And the sizes still don't add up. As mentioned, a single campaign for the pro-NN side had 7 million suspicious comments. That's bigger than any of the alleged bot campaigns on the other side. Overall, there's more suspicious comments for the pro-NN side. Whether the evidence is enough to definitively call them bot campaigns or not, that's still true.

by ikeboy 6y ago

geofft 6y ago

The standard of evidence I'm applying is simply that evidence for X being true must not also be plainly compatible with X being false. Here, for instance, is some evidence that meets that standard, from that very same article you love: > He found a near-constant rate—1,000 every 10 minutes—punctuated by periods of zero comments, as if web robots were turning on and off. He determined many were from hacked accounts. This pattern is either server-side (e.g., a server-side rate limit and an extremely flaky server) or client-side. If it's server-side, we'd expect to find it independent of the contents of the comment, we'd expect to have stories of people being unable to submit comments, etc. Unless we have those stories, then this pattern is plainly incompatible with actual humans submitting comments, even with the aid of some automation or pre-written letters. That is evidence that these comments were almost certainly not from actual humans. No similarly strong evidence exists on the pro-net-neutrality side. "A bunch of comments claimed an address from Russia" is certainly suspicious and weird, I agree, but it is not plainly incompatible with humans (if it's IP address, bad geolocation; if it's e-mail address, Russian-run mail provider whose accounts are used by Americans; etc.). I am happy to grant with more detail about what's is meant there - which neither the FCC nor the WSJ has provided to us - there may be strong evidence, but it's not there yet with what you've posted. If you want to revise your initial complaint to "Friendly reminder that some weak evidence exists in favor of small, isolated bot campaigns on the pro-network neutrality side and very large bot campaigns on the anti-network neutrality side, such that if all suspicious comments were disregarded in proportion to their likely untrustworthiness, the overwhelming weight of public sentiment as measured by this exercise would be firmly on the pro-network-neutrality side," then I have no remaining objection.

ikeboy 6y ago
>Unless we have those stories, then this pattern is plainly incompatible with actual humans submitting comments, even with the aid of some automation or pre-written letters. A site is set up to collect names for form letters. They don't submit it in real time; instead, they submit it in batches every few minutes. This is at least as plausible as the various explanations you gave above. In fact, if you read the linked blog post, they explain that this is how the API worked - you'd upload a CSV and everything would be uploaded in bulk. And the sizes still don't add up. As mentioned, a single campaign for the pro-NN side had 7 million suspicious comments. That's bigger than any of the alleged bot campaigns on the other side. Overall, there's more suspicious comments for the pro-NN side. Whether the evidence is enough to definitively call them bot campaigns or not, that's still true.