User talk:WereSpielChequers/Smart Blocking

Socking

Some of this (users moving to new IP's once blocked) seems realted to socking, correct? Slatersteven (talk) 13:06, 2 November 2025 (UTC)[reply]

Hi Slatersteven, presumably yes, but sockhunting is not an area where I can claim much expertise (on wiki or IRL, I used to have a whole drawer of single socks before I got married, my wife threw them out and then mismatched socks became fashionable). Ϣere SpielChequers 13:27, 4 November 2025 (UTC)[reply]

Changing data retention

It will also be necessary to change our data retention policy so as to retain some checkuser data for longer. --> I think this is going to be a sticking point. I think, from a legal and user-protection POV, the fact that our data retention is severely limited is probably deliberate, aligning with Wikipedia's general techno-freedom (there's an actual word for that and I'm blanking on what it is right now) ethos - if we don't have data on people, it can't leak, etc. So I would expect pushback, from community or Foundation or both, on this point if you pursue this proposal. And I think your overall idea could probably be implemented, at least initially, without it (even if that would make the functionality not-as-perfect); maybe smart blocks can only last 90 days or whatever, and then if they prove a winner as a block method, we can revisit data retention with that as an argument for changing it. Note: I work for the Foundation, but I speak here only as a volunteer with privacy experience as an Oversighter and blocking experience as an admin. Please don't take what I say here as a indication of the Foundation's Official POV; it's not. Fluffernutter (talk) 15:59, 2 November 2025 (UTC)[reply]

Hi Fluffernutter, I'm doing a fundamental rewrite now I've learned that these autoblocks are 24 hour not indef. I think I can bring this proposal into line with current C/U policy. Agree that we should respect privacy rights even of vandals and other miscreants. What I don't fully understand is the trigger that generates an autoblock. I get that someone active on another wiki would trigger an autoblock if they tried to save an edit on EN wikipedia. But what if they viewed their watchlist, did a search or read an article on EN wiki? Maybe what I need from the devs is a bit more detail as to what would trigger an autoblock of an IP address from an account that is active and unblocked on Wikidata but was blocked over a year ago on this wiki. Ϣere SpielChequers 13:38, 4 November 2025 (UTC)[reply]

Adaptive blocking

I've called for something I call adaptive blocking. A block that changes the level of blocking and type of actions being blocked, depending on signals. The nice thing is that it could also be used the other way around, so that when reputation becomes good, or if there is a subblock within a range that is simply never used for abuse, then these will not be affected by annoying blocks for users. It could even auto surface these and suggest them for unblocking by admins, so that we keep ranges narrow. I recently brought it up (at PTAC) as a way to deal much better with the limited admin resources we have, while also dealing with the IP (anon-user) blocking problem. I don't think we need to extend checkuser for this. We'd just separate some data out into some sort of reputation score table (much like the ipinfo data we get from external sources as well) and then have that be combined with some recent checkuser signals. By making the limits dependent on that score and on recent signals from that ip range, I think we can deal with most of it without even having to expand the time we keep checkuser data for. This would be a major engineering effort, but I think we need to look at these kinds of things. —TheDJ (talk • contribs) 12:04, 4 November 2025 (UTC)[reply]

Hi theDJ, thanks for your interest. I'm not sure I follow how your adaptive blocking would work. I've had it pointed out to me that the autoblocks are only for 24 hours, not indefinite as I'd thought. However they can be triggered a year or more later in a way I don't quite understand. So I've started rewriting this proposal because I think my core idea of reducing collateral damage by looking at O/S is still sound, but I'm not quite sure how the autoblock works and whether it makes sense. Perhaps what we need is a way to stop autoblocking accounts that are in good standing elsewhere in the movement but who occasionally read EN Wikipedia. Ϣere SpielChequers 13:24, 4 November 2025 (UTC)[reply]