I just had what may or may not be a silly idea about a way to deal with trolls and extremism in forums like Facebook. It doesn't deal with the issue of detecting bad content, it just deals with the content when detected.
It's similar to greylisting, a technique used against email spam. The idea is to delay posts identified as "bad" (by whatever criteria) or from a poster identified as "bad" (by whatever criteria). Posts are flagged as "delayed because of content" or "delayed because of source" accordingly. The messages are not published immediately. To be published, the exact same message must be posted again, no earlier than some specified later time. The exact same message posted before then will simply be discarded. If it's not the exact same message, it gets treated as a new message and if detected as bad will get its own delay applied. If a message is never reposted, it is deleted (say) 24 hours after the initial delay expired. Only if the message is reposted exactly as it was, after the delay has expired, will it be published. Messages exactly matching a message "delayed because of content", but from different sources, will be delayed. Identifying a post as bad simultaneously identifies the poster as bad. All messages from that poster are delayed, and go into the "more careful checks" queue. The more problematic posts the same poster makes, the longer the delay becomes. I'm thinking an initial two hour delay, then going up in twelve-hour increments. Or start smaller but be exponential. If the detection is up to it, start with larger delays for worse transgressions. The delay decays at a fixed rate - say two hours per day, so rehabilitation is automatic. Delays could/should be imposed on all responses made TO people with known (bad) provocative effect. Any response to e.g. Trump (retweets, reposts, replies etc) would automatically get the delay applied, regardless of the respondent or the content of the response, and regardless of whether any particular message from him was delayed. Delays could be applied to messages/posters that use particular hashtags or similar identifiers. And finally, any response to a post that was delayed because of content, gets delayed. All this is on top of other avenues against serious transgressors, such as banning them outright, deleting seriously bad material and so on. In my head this should have several good effects: - it does NOT stop people saying what they want to say, if they really want to say it - it gives a publisher such as Facebook more time to deal with a problem poster or a problem post (e.g apply a ban or decide to not accept the repost) - it sorts pending responses neatly by seriousness (if the detection process is good) - it reduces the explosive effect of instant responses around events or provocative messages from others - it will give people time to consider that a) the platform finds their post objectionable and thus b) do they really want to say it? - only the most persistent will actually repost their messages - a delay can be applied immediately with little harm being done to freedom of speech; automation of this measure is less dangerous - even if someone writes software to automatically repost delayed messages, the delay is still in there. Bad effects would be: - it imposes delays on those seeking to debunk or otherwise defend against a bad post. Given that such messaging generally just feeds the trolls, I'm not sure it's a big problem. And good posters, if they are keen, can repost just like the bad ones. - someone with a delay imposed who seeks to make an emergency communication via the platform would see that communication delayed too. - I can see a form of DDoS if someone posts a great many almost identical messages seeking to overwhelm the badness detectors, but I think that measures already exist to cope with it. If a single source does this, it is relatively easy to detect and to block or rate-limit. - the whole thing does need a badness detector for the platform. So - is it a silly idea? Regards, k. -- ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Karl Auer ([email protected]) http://www.biplane.com.au/kauer GPG fingerprint: 61A0 99A9 8823 3A75 871E 5D90 BADB B237 260C 9C58 Old fingerprint: 2561 E9EC D868 E73C 8AF1 49CF EE50 4B1D CCA1 5170 _______________________________________________ Link mailing list [email protected] https://mailman.anu.edu.au/mailman/listinfo/link
