User:Sj/mattersofpolicy

Graded permissions system

[3/2x 21:01] <SethIlys> And if Jimbo seriously wants a metric system, we should give ideas serious deliberation.

Concepts

leverage

The leverage L of the system is the ratio of t/v-hours to editor-hours in rv'ing mischief. We should aim for $L>>1$. (It should take much longer for a t/v to cause harm than it does for others to undo it.)

metric

A simple linear metric useful for making certain programmatic, unambiguous gradations in divisions. Inherently gameable but neutral.

Metric

Sj’s silly scoring system: give each account/IP a naive 'trust rating', T. (In the following descriptions, time (6 hrs / 3 days / 3 months) refers to time since an account's first edit, not since registration)

 Trust pts:  
   +0.01 if 'non-bot' flag has been set by another user, or for
         ~6 hrs w/ 3 edits [for screening out bots];
   +0.1  for 3 days w/ 5 article edits, [for increased troll leverage];
   +0.9  for 3 months w/ 100 article edits;  
   +1 for 1000 edits,  +1 for 10000 edits
   -1 for each ban in past 3 months (multiple bans w/in the same
         7 days count as a single ban),
   hold at 0 if a 'warning' flag has been set by a sysop 
         (IP shared by vandals and others, IP with only one or two
          trivial edits that may be a bot gaming the system, etc.)

Uses

• Need T > 0 to post normally (i.e., changes automatically update page viewed by anons) to article/User: pages
• Need T > 0.1 to post normally to meta/WP:/Talk: pages (voting, not being subject to immediate rv'sion at the whim of others), to set 'non-bot' flag for other users
• Need T > 1 to be nominated for adminship;
• Need T > 2 for certain actions, like QP creation, being on binding committees (eg: arb), etc.

Use cases:

1. A spambot, knowing these rules, comes along and tries to maliciously update a few thousand pages at once. This should not make the WP unusable (as a reference) for even a minute. [for instance, each bot account would have to wait six hours after its first edit before it could go on a visible rampage; if the bot tried to do this with hundreds of anon IPs at once, this would be obvious; they could all be preemptively warned.]
2. A vandal, planning to insert insidious false statements into many articles, joins and creates ten sockpuppets for this purpose. It eventually takes ~two editor-hours per sockpuppet to deal with the attack; it should take longer for the vandal to prepare and impelement it.
3. An eager user comes by and wants to start making changes to music articles right away. These changes should be allowed, encouraged, and thanked, without violating the above policies. Requiring that a user with T > 0.1 edit a page after such a newbie, before changes go live, is largely inoffensive. The new user should get polite messages upon hitting "Edit" explaining how and when their edits will show up, how they will be validated for instant gratification by the time they return the next day, and how they can speed the validation process (find any other user to set their non-bot flag, post to a given Wikipedia:ValidateMe! page, etc).

Implementation

1. Add a T field (for storing this value) to the users table.
2. Add a 'non-bot' flag and a 'warning' flag to the users table.
3. Add a 'user-validation' audit table to log when a user validates a new user (by setting their non-bot flag), or toggles a user's warning flag.
4. Run appropriate set of SQL queries (for users with T=0, run the first query once every 6 hours; other queries can be run once a day or once a week) to udpate the T field.
5. When a page is edited, in the same query that checks for banned status, extract the T field for the editing user. Then display different edit pages based on T (for newbies, show special edit page).
6. When a user's 'warning' or 'non-bot' flags are set or unset, actively update their T.

---

Notes on DB structure

• what's the timeline for implementing the 12/03 schema update?
  • see disc w/BV on 3/16... piece by piece.

• Needs some kind of 'many lang_ids per art_id' scheme.
  • disc w/BVibber on 3/16; negative, too hard, useless.
  • see meta:db schema

Notes on WP structure

• A beautiful terse article on WP Power structure.
  • Dev access disc by TStar

Policy projects

• Repair of Speedy deletion (via post-moderation sol'n for mostly-obvious cases, with strong consequences for being wrong)

• Articles on JW and LS which, unlike all previous instances, don't suck (so there's something tangible to agree on). Bounds on inclusion disputes: sufficient standard for fame/inclusion and necessary standard for exclusion.
  • LMS: see his VfD'ed article (also below, in case of deletion).
  • JW: 3apes and wikia; old objectivist supporter, m-list moderator. cf his user-page / personal site for 'all-around frood' cred.
  • public presentations? W/Michael Hart (when?), at WOS 2004, etc.

• Basic understanding re: lists as interface aid (cf. discussion of MW:Religions, MW:Religousfigures, MW:Cults). Sidebars? footers? headers? organizing schema for entire article? "series" designations? overall (in-db?) cats, etc.

Worst-case quickpolls scenario

1. Marriah[M] has been an admin since Jan. M knows a lot about Murghian history and requested adminship to counteract a recent influx of users with radical Murghian views who were changing world history pages and lists all over WP to support their version of history (they all came over from the same online discussion board in Dec.).
2. Since becoming an admin, 25% of M's time on WP has been devoted to monitoring the edits of these 20 or so users. After 3 months of this vigilance, they have mainly retreated to their own WP.
3. M patrols vandalism with zeal. Every 20 vandals or so, M overreaches and deletes a valid contribution; some other longtime user notices this and issues a warning; M thinks the article in question did merit speedy deletion, and replies dismissively to the warning.
4. A couple days later, in response to a similar deletion, the same user starts a quickpoll to temp-desysop Marriah. This is the first time M discovers that a number of other admins also disagree with M's interpretation of 'speedy deletion' rules. Most other admins are ambivalent about this poll, but there has recently been a big debate about tightening up on vigilante admins, and 12 or 15 vote for it, to get M to stop deleting so actively. M apologizes the next day, but has already been temp-desysopped.
5. The next week, after that brief suspension, M gets into an edit war with two longstanding controversial contributors; making 3 rv's one day and another 2 the next. Both users put up separate quickpolls to desysop M again, noting that 4 of these rv's were made "within 24 hours".
6. Initially, only a handful of people vote in (both) polls, half of them opposed to it. One of the two polls is taken down. There are arguments as to whether the polls had a right to be started in the first place. Poll-confusion reigns.
7. Meanwhile, the users who put up the polls again revert M's rv's to their POV, with a few choice insults. M gets angry, and instead of apologizing for not noticing that the 4 rvs were really within 1 day, or finding another admin to help maintain NPOV, in an incautious moment, temp-bans one of the complaining users. Many users think this was a rotten thing to do.
8. A new poll is put up, in an attempt to clear the poll-confusion; of the first 10 votes, 8 are in favor; ten minutes after the poll opens, an eager bureaucrat unbans the user in question and (on the basis of the first 10 votes) desysops M. Subsequently, with 15 votes in, only 10 are in favor of the poll... 2 hrs later, M gets angry that he hasn't been resysopped, angering 2 other users who come to vote in favor of the poll; at 12/17, the vote is above 70% again, so nobody reverses the previous decision. M leaves in a huff; finally apologizing two days later, however, 48 hrs have passed and the 'successful' poll has already been removed as finalized. M has to wait the full week before being reinstated (since going through RfA takes time), despite the fact that most of the people who voted for the poll now wish they could change their vote.
9. M now holds a grudge against the bureaucrat who desysopped so hastily (since poll support immediately after dropped below 80%) and a few admins who supported that bureaucrat; they likewise hold a grudge against M, and rather wish that M weren't a sysop any more.
10. Meanwhile, the Murghians realize they can collaborate to get M desysopped 'permanently'. First they all make sure they have enough edits on the en: WP to vote, and a week later a few of them (with new user accounts) intentionally bait M. M complains loudly about their activity and advocates banning them, but the admins who dislike M speak out as devil's advocates in their favour, without knowing anything about the topics under dispute. Neutral admins, looking from afar, see M alone against 2 or 3 decent-looking new users and 2 or 3 admins, and figure M might be misbehaving.
11. M is careful at first about reacting calmly to the offensive edits, but they are careful to also contribute neutral content and to respond on relevant talk pages as though they were serious contributors (and anon users start reacting, in strong language, on M's user talk page, claiming he is fanatically supporting his fascist anti-Murghian minority POV, abusing his position [such as it is]).
12. The collaborators start making subtle offensive additions to minor pages that only M monitors; M discusses better alternatives while rewording and NPOV'ing their additions, which they ignore and rv using 3 different new accounts. They complain on the talk pages that M is 'reverting' their work; though M only uses admin-powers to rv once, they later claim M reverted over 5 times in one day -- twice "as a 'minor' change without edit summary".
13. Finally, one of the Murghians, with an old account, starts a quick-poll to desysop M again. M responds angrily that this is obvious nonsense; the two regular admins who dislike M quickly support the poll, as do a few of the Murghians. Over the next day, the other 20 Murghians chime in, in staggered fashion, each claiming similar treatment to his or her work (possibly true, since some of this happened before even the first complaint). Most admins, seeing this outpouring of opposition to M (and not knowing the relevant history) join them or abstain.
14. Less than two weeks after the first quickpoll, and without ever having knowingly violated WP policy, M is desysopped; the case goes to arbitration, but too late to remedy this slap in the face.
15. Chaos breaks out on mailing lists and talk pages, and most people say "let's wait for the arbitration committee to decide". Over the next few days, while this is progressing, the Murghians return to their efforts to subtly alter articles on history and Murghianomics, claiming the high ground in any edit wars, twisting the knife.
16. Long before arbitration is finished with the case, Marriah and a few fellow editors leave Wikipedia in disgust, and M turns the same energy to an anti-Wikipedia PR campaign among fellow academic historians. Many worthy expert contributors are lost.