User:STBotI/MessageCollaboration

For all warnings, on the third warning, the fifth, and every fifth thereafter, the warning to the user has this appended to it: Also, on the fifth warning and every fifth thereafter, a notice is sent to users of the IRC channel #wikipedia-en-admins advising them of the user. The message is worded as follows: On the tenth warning (second message to IRC), that message is followed by: On the fifth warning, the user receives.
 * If there is no tag whatsoever and the image contains the words self, my, mine, or I, the image is tagged with nld.
 * If the user is not opted out, they get tagged with.
 * If there is no tag whatsoever and the image does not contain the words self, my, mine, or I, the image is tagged with nld/
 * If the user is not opted out, they get tagged with }.
 * If the image lacks a rationale, the image is tagged with nrd
 * If the user is not opted out, they get tagged with.
 * If the image lacks a valid backlink, the image is tagged with
 * If the user is not opted out, they get tagged with.
 * Is that really the message you use? It looks exactly like the one you de-approved for BetacommandBot. It's written entirely in Wiki-jargon, and it makes it sound like "The rationale is presented in clear, plain language" is part of the problem. Also, I maintain that the "no backlink" message should not be a deletion template, because in many cases the situation will not warrant deletion.  r speer  / ɹəəds ɹ  19:29, 2 June 2008 (UTC)
 * I believe it is. One of the things I'm working on (read: thinking about and procrastinating on) is to allow a reference to the name of the article even if it's not linked. The code exists, if we changed it to only require a reference to any one article name, would that be acceptable? (Requiring a reference to all article names would require a further BRFA) I've also rewritten the deletion message. --uǝʌǝsʎʇɹoɟʇs (st47) 22:28, 2 June 2008 (UTC)
 * Any one article name? That would allow basically everything with words in it, so I doubt that's what you really want. Anyway, none of this addresses the problem that the guideline for how to write rationales is at odds with common sense. Most images are used on one article. Users follow the directions and write one rationale, thus establishing a rationale for each use. It never even occurs to them that they need to explicitly say that the one rationale applies to the one article, because it's obvious. Now, I recognize that it's important to have some bookkeeping, or else there could be some uncertainty to which article the rationale applies to if it gets added to another article. So I think cases like this should be put in a category so they can be checked or improved, but the user certainly shouldn't get a "rawr, you didn't write a rationale, BALEETED" message.  r speer  / ɹəəds ɹ  09:26, 3 June 2008 (UTC)
 * Any article that contains the image, of course. We aren't here to rewrite policy. --uǝʌǝsʎʇɹoɟʇs (st47) 10:34, 3 June 2008 (UTC)
 * The actions of bots and policy aren't equivalent. Think of all the subsections of NFCC that are impossible to enforce with bots for an example. So don't accuse people who are trying to give you suggestions to improve your bot of "rewriting policy", because you're not going to get an acceptable bot that way. I am trying to convey the idea that no bot can be relied on to determine whether the user has written a rationale for each article. Unless the policy changes to require the use of a particular template (an idea that was already rejected, and contradicts the idea that rationales are "plain English"), this requires human-level understanding of all the text on the image description page.  r speer  / ɹəəds ɹ  08:09, 5 June 2008 (UTC)
 * Back when ImageTaggingBot checked for fair-use rationales, it was able to identify images without one with a false-positive rate of less than 1% -- by the time I removed the check, the bot had gone over six months without a false positive. Of course, it also had a false-negative rate upwards of 99%, but that doesn't mean that it's impossible for a bot to identify an image without a fair-use rationale.  It's all a matter of what false-positive and false-negative rates you're willing to accept. --Carnildo (talk) 19:42, 5 June 2008 (UTC)
 * I'd prefer to minimize both, of course. --uǝʌǝsʎʇɹoɟʇs (st47) 23:09, 5 June 2008 (UTC)
 * It's a trade-off. You can't minimize both at the same time, for any reasonable sense of "minimize". I see two reasonable options here: you can avoid false positives like the plague, or you can rewrite the messages so that they aren't deletion templates and admit the possibility that your bot is in error.  r speer  / ɹəəds ɹ  10:22, 8 June 2008 (UTC)
 * How does the bot handle backlinks? FairuseBot is capable of tracking down an article through any combination of redirects and disambiguation pages.  This is less of a problem for new uploads, but handling redirects to take care of cases where people are lazy about diacritics and other special characters is essential. --Carnildo (talk) 09:03, 3 June 2008 (UTC)
 * It doesn't, that's what I'm trying to fix. I would, right now, require any link to the main namespace (including potential redirects) or the name of any article that uses the image. --uǝʌǝsʎʇɹoɟʇs (st47) 10:34, 3 June 2008 (UTC)
 * "Additionally, if you continue uploading bad images, you may be blocked from uploading."
 * "!imageabuse $user has been warned many times for uploading bad images, please check his uploads. (Report number $num)"
 * "User HAS been warned with blocking"