4 ms·

> Short answer: yes. > Longer answer: Basic data analysis can pick up most if not all the shenanigans. Doesn't matter how much they "age" the accounts or hide

> Short answer: yes. > Longer answer: Basic data analysis can pick up most if not all the shenanigans. Doesn't matter how much they "age" the accounts or hide their tracks, eg IP address etc. Don't even necessarily need high powered hardware or even machine learning. They do help in particular ways if you choose to use them. > Fake vote detection is now a choice not to rather than a technical blockage. I don't believe this is correct. Bots can create simple sentences that very much appear as a human. Spread these out across enough accounts, and filtering noise becomes very difficult. There are also apps that activists use and I'm certain PR firms as well which alert humans to go and vote in certain ways. These and other technologies combined with IPv6 make it exceptionally difficult for site admins and moderators to discover manipulators. Plus, many sites have few resources to put into the task anyway. HN has at least put one step against it by not having downvotes on article submissions, though that doesn't help against upvoting manipulators.

rs23296008n1 7y ago
You are operating on the assumption that analysis of the text is required. Nothing I'm admittedly vaguely mentioning requires analysis of any bulk amounts of text or requiring massive manpower or anything that ends in a hopeless sense of "it can't be done reasonably". It most certainly can. Distributed mobs of clever humans or bots can't actually hide. One other thing: I'm not expressing a theory up for debate. I'm only mentioning that there are already alternatives of detecting downvote mobs and similar interference. This is already in production. The old saying "If you believe its impossible then you're right". As for HN, many downvotes are (ab)used for disagreement rather than the comment being wrong or factually incorrect. I've seen this on HN plenty of times. At minimum, all downvotes should require an associated comment or upvoted reply so the offending commenter knows exactly what happened.

rs23296008n1 7y ago

You are operating on the assumption that analysis of the text is required. Nothing I'm admittedly vaguely mentioning requires analysis of any bulk amounts of text or requiring massive manpower or anything that ends in a hopeless sense of "it can't be done reasonably". It most certainly can. Distributed mobs of clever humans or bots can't actually hide. One other thing: I'm not expressing a theory up for debate. I'm only mentioning that there are already alternatives of detecting downvote mobs and similar interference. This is already in production. The old saying "If you believe its impossible then you're right". As for HN, many downvotes are (ab)used for disagreement rather than the comment being wrong or factually incorrect. I've seen this on HN plenty of times. At minimum, all downvotes should require an associated comment or upvoted reply so the offending commenter knows exactly what happened.