Thread

1. nomel · 1790637663 · focus · HN ↗

   > Thin accounts: few comments, low scores, no real standing in the subreddit.

   > Young accounts, or accounts with histories that are hidden or wiped.

   Nope, this is no longer an indicator of a bot account. Many that I've looked into will have active posts to local town/region subreddits (often times multiple!) and different sports subreddits, sometimes getting unusually high amounts of karma in them (which suggests a network).

   Turing test is looong dead, so no reason to assume the small fractions of a cent to make comments, to get a reasonable history, isn't being taken advantage of.

   1. 13415 · 1790669333 · focus · HN ↗

      All claims about bot accounts are flawed when there is no independent confirmation of the methodology. You'd need to have some sufficiently data based on verified bot accounts and evaluate your methods and claims on this data. Many people claim they can identify bot accounts intuitively. I do believe they can to some extent, but it seems far-fetched to insinuate that this ability is accurate enough to evaluate some statistical methods based on indicators.

      1. nomel · 1790892978 · focus · HN ↗

         > but it seems far-fetched to insinuate that this ability is accurate enough to evaluate some statistical methods based on indicators.

         Again, the Turing test is LOOONG since passed, so we're far past the point of being able to do any of this with confidence. It's not possible to gather meaningful statistics anymore, besides maybe identifying some tail/subpopulation created by incompetent devs. And even then, you'll start categorizing the college students paid for social media campaigns (you can find fliers of job postings at your local campus, you'll "have a blast!").

Unofficial Hacker News client; not affiliated with Y Combinator.