Wikipedia:Bots/Requests for approval/BLPWatchBot

The following discussion is an archived debate. Please do not modify it. Subsequent comments should be made in a new section.

BLPWatchBot

• (BRFA
• Approved BRFAs
• talk
• contribs
• count
• SUL
• logs
• page moves
• block log
• rights log
• flag)

Operator: --uǝʌǝsʎʇɹoɟʇs(st47)

Automatic or Manually Assisted: All edits are manually assisted.

Programming Language(s): Perl, using, primarily, my Perlwikipedia module and Poe::Component::IRC

Function Summary: A bot to help watch activity on at-risk BLPs that have previously been mis-edited, or are anticipated to be.

Edit period(s) (e.g. Continuous, daily, one time run): continuous

Edit rate requested: 5 epm, maximum

Already has a bot flag (Y/N): N, flag not requested.

Function Details: This bot maintains a list of pages (usually BLPs) tagged by volunteers as requiring monitoring, and transmits all edits of them to an IRC channel. Users in the IRC channel can instruct the bot to add or remove the tag from articles, revert certain edits, or change parameters to the template. The template is {{Blpwatch}}. Users who are using the revert functionality will be individually approved. The bot also automatically removes the template after three months monitoring, though any user may add or extend monitoring for a three month period at any time, if necessary. De-monitoring of pages is also checked for and reported to the channel.

Discussion

This requests partially dupicates the functionality of #wikimedia-otrs-watch channel. It, however, offers some new features. Here are a few questions:

• What will happen if a troll or vandal removes {{blpwatch}} from an article? Will the bot notice that?
• What do you mean by "revert certain edits"? Edits matching some pattern? Or simply respond to an IRC command?
• On which channel will it live? Will this channel open to everyone? MaxSem^{(Han shot first!)} 19:35, 25 March 2008 (UTC)

Well, most editors aren't on OTRS.

The bot reports to the channel for every removal.

Reversions are done by command by a user to revert a certain edit on a certain article. The user who requested it is linked in the summary.

It lives on #wikipedia-en-blpwatch. The channel is currently open to anyone. --uǝʌǝsʎʇɹoɟʇs(st47) 19:38, 25 March 2008 (UTC)

1. The aim is to provide a quick informal way that articles at risk can be watched. A typical context is some less active biographical article that has had some vandalism, and having reverted we want to ensure the vandalism isn't put back 2 weeks later. So we monitor its edits for a 3 month period "to play safe". if all's good then it automatically gets delisted at that point; if there is vandalism then the bot can be told to update the end-time of the monitoring for 3 months from that date. As such it's not the same as OTRS, because any user can tag, any user can help watch, and it is intended to be used on any page with a blp concern, not just the ones specifically brought to OTRS attention.
2. As well as watching for adds and removes of the tag in the rc feed, on startup it compares the list of articles it expects to see tagged, with those that are, and reports any that are detagged to the channel too. (That's to prevent a vandal lurking in the channel, and waiting for a bot restart to click "save" on the tag deletion.)
3. To do so, it keeps a simple plain-text event log file on the server that records adds/removes, bot start/stops, and pages noticed to have been detagged while stopped, with date/time. The event log has read-only access to the world, to allow anyone who wishes to review such events (and check for anomalies, adds, removes, etc) to do this via their browser at any time, and the log URL is given at request within the !info command.
4. The way reverts will work is that when edits are listed on the channel, an edit # (internal to the bot) will be given for it (eg: "#1234 <page> <user> <date/time> <edit summary>..." etc). A user who is approved for reverts can type !revert 1234, and if that is the most current edit, it will be reverted with their name in the edit summary. That's because if a vandalistic edits is identifiable by edit summary, then simplicity is useful. It does not excuse a user from responsibility for the revert, but if a user is responsible and experienced, then the ability to revert a bad edit quickly will allow them to help watch BLP risks more efficiently.

FT2 ^{(Talk | email)} 20:05, 25 March 2008 (UTC)

I don't like the idea of allowing users to revert edits without even looking at them, with a bot. — Werdna talk 03:58, 26 March 2008 (UTC)

Comments on this. First, we already have a facility with that exact action on-wiki. Any user with rollback right, which all admins have and indeed many non-admins too, has been given a tool which allows reversion of edits based upon edit summary alone, with a link to the diff but no obligation to view it.

Second, it is restricted; those who use it, if they do not exercise good judgement, will have it removed. Lastly, a wide range of vandalism can be very effectively identified by edit summary (otherwise rollback would be unhelpful). For example, a given BLP may be exclusively vandalized by a given IP range that for some reason is not blocked, and many times vandalism or pov editing/defamation will be accompanied by an edit summary "JOSH IS GAY!!!" or "NO THIS LEGAL MATERIAL IS IMPORTANT AND MUST BE KEPT!". There is no obligation to use revert, nor is it given to everyone. But undeniably there are cases a decision can be made quickly. Others, not so. The expectation is those given that access have a clue. FT2 ^{(Talk | email)} 06:35, 26 March 2008 (UTC)

I think it's very important to consider the purpose of the bot when talking about this functionality. My understanding is that this bot is not for combatting vandalism, per-se (otherwise it would be VandalWatchBot) — rather, it is for watching biographies on living people.

To respond to your first point, I want to go back to the purpose of the bot. Rollback links on user contributions and history pages are there for mass-reversion of vandalism and the like — not something that comes up in BLPs particularly.

To your second point, I would claim that, as a general principle any user who has seen only at the edit summary and (probably more likely) the username, should not revert it — unless the edit has been given an automatic edit summary, or it is blatant vandalism. This is part of giving a user's edit the fair consideration it deserves. The purpose of this bot does not seem to be to revert blatant vandalism — it seems to be to watch for the addition of controversial uncited claims on articles on which this has previously been a problem.

The other point I had, which I don't think I explained properly in my first message, is that the edits are done by a bot. For things like requests for adminship and arbitration, proxy editing such as this does not leave a paper trail under the user's contributions, and so it's harder to attribute that edit to that user reliably. — Werdna talk 12:35, 26 March 2008 (UTC)

You might need to wait for ST47 to comment; I can only comment from a limited perspective so to speak. The aim is to watch at-risk BLPs of all sorts. Some BLPs will be at risk from addition of controversial uncited claims. But others may be at risk from blatent vandalism, or slur insertion. Where a BLP is blatently vandalized, it makes sense to provide a quick way to handle it. The odds of an edit summary "JOSH IS GAY!!!" or "PEN1S VANDALISM ADDED BY USER ON WHEELS!!" (or whatever vandals and slur-artists write) being a bona fide edit are slim... especially if the first is on an article where the bot reports from the tag that the main risk is "homosexual slur vandalism". A user trusted with a quick revert option will also be expected not to use it except in clear cases. It would of course always also be attributed in the edit summary to a specific, named, on-wiki user. FT2 ^{(Talk | email)} 13:28, 26 March 2008 (UTC)

{{BAGAssistanceNeeded}} Issues have not yet been satisfactorily resolved. — Werdna talk 10:56, 31 March 2008 (UTC)

I've put up a link to this request on the Community bulletin board, let's see what others will say. MaxSem^{(Han shot first!)} 11:20, 31 March 2008 (UTC)

What form of access control does the bot use, to prevent possibly unwanted reverts? SQL^{Query me!} 07:57, 1 April 2008 (UTC)

Also, the bot has a small IRC Context leak. It'll take input from wherever, regardless of which IRC server it came from. See: here. SQL^{Query me!} 08:38, 1 April 2008 (UTC)

NM, in playing with the bot, it clearly appears to use hostname recognition. SQL^{Query me!} 09:09, 1 April 2008 (UTC)

( Fixed) SQL^{Query me!} 14:57, 1 April 2008 (UTC)

Werdna: It appears to me that FT2 has answered your question. If not, please restate the relevant parts so I can answer them. SQL: Approved users, based on irc cloak. The user must be logged in to IRC. The bot only accepts RC information where the user is 'rc!rc@*', which is the bot used on irc.wikimedia.org. --uǝʌǝsʎʇɹoɟʇs(st47) 10:28, 1 April 2008 (UTC)

Werdna's not quite got an answer to his question. The issue stated is "bot edits will not reflect under a user's contributions for RFA and similar processes". My answer here is, the bot has very limited scope for editing. It cannot make "freeform" edits; it can merely revert a diff, or manage a specific hardcoded tag. A user who misuses this will not have their edits on their contribs, it's true. But

1. the scope for misuse is minimal,
2. there is a strong "trust and accountability" barrier to access (users must be trusted and individually given access; bot enforces accountability for each edit), and
3. a user who despite these did misuse those very limited functions (and really almost the worst they can do is revert a bot-specified RC edit) would be tracable/noticable in future, seen by many almost "on the spot", be questioned immediately and possibly also lose access very quickly... and that would be noticed at RFA and the like -- even assuming that somehow none of the many watching had posted a comment on their talk page at the time with a link to the diff asking them to explain it.

FT2 ^{(Talk | email)} 12:17, 1 April 2008 (UTC)

Well, we can make a public list of approved users if that would help give oversight into the users' edits using this bot. Also, I anticipate that a large number of those approved will already be administrators. --uǝʌǝsʎʇɹoɟʇs(st47) 21:38, 1 April 2008 (UTC)

Works for me, zero reason not to say who; never really thought about it being private. it's on your talk page anyway. FT2 ^{(Talk | email)} 23:06, 1 April 2008 (UTC)

Sounds like most of the bugs have been dealt with (ST47, if you can, I would really, really, strongly recommend that you seperate the IRC context's, or at least, check what channel the data is coming from, but, I can't really see how this could be abused. But, that's how problems start...) Anyhow, I'd like to see this in action. Whenever you're ready, please start a short trial. Approved for trial (5 days). Please provide a link to the relevant contributions and/or diffs when the trial is complete. SQL^{Query me!} 04:42, 2 April 2008 (UTC)

Also, given the task, could I ask something a little out-of-the-ordinary of you? Could you please link to this BRFA during the trial, in the edit summary, so long as the bot is operating until it is approved? (Request, not a requirement) SQL^{Query me!} 04:56, 2 April 2008 (UTC)

Also, I'd like to suggest that this trial be expanded up to 30 days, pending input on this task, if more feedback is needed, depending on how things work out SQL^{Query me!} 05:54, 2 April 2008 (UTC)

Having discussed, this is the idea that edits made by the bot during trial, should specifically note in its edit summary that the bot's on trial and that feedback is welcome (with a link). It sounds sensible. Also, I'm not sure that a 5 day trial will pick up many actual edits though so the second point makes sense. FT2 ^{(Talk | email)} 05:57, 2 April 2008 (UTC)

Comments related to bot trial -> below

Bot trial

This bot is now under trial. Comments and feedback on the bot and its usage are welcomed below. Specific bugs or unexpected responses (which may well arise at times) may be reported at User talk:BLPWatchBot for fixing and discussion.

Of especial interest is whether the bot is being appropriately and correctly used by those authorized to use it for reversion or tag modification, and any problems arising in usage. Note that bot usage is a different issue from loopholes in bot coding

• The bot will add any text to any page, from any user. See: [1]. I did that with !tag User_talk:SQL test2. Does not even have to be an article, or a template. very abusable IMO. I didn't realize there were commands that cause edits, with no authentication required. SQL^{Query me!} 22:02, 2 April 2008 (UTC)

Input to the !tag command will now be santized, requiring it to fit the format {{blpwatch|[^\}]+}} - essentially, it must all be the template. --uǝʌǝsʎʇɹoɟʇs(st47) 22:35, 2 April 2008 (UTC)

Small suggestion, could the edit summary be changed so that it doesn't include details of the user who initiated the revert. The ability to completely invisibly remove vandalism and other undesirable edits in a very small number of cases is most desirable. The details of who initiated the revert could instead be logged by the bot, available for inspection. Nick (talk) 23:01, 2 April 2008 (UTC)

I've implemented this. The reverter is still logged privately, so anyone who abuses the bot can be found out. --uǝʌǝsʎʇɹoɟʇs(st47) 10:48, 3 April 2008 (UTC)

I'm not sure this is such a good idea. in what situation is there a need to revert anonymously? Everyone should be able to see who did a revert (or any kind of edit, for that matter). --Conti|✉ 21:56, 3 April 2008 (UTC)

Per Conti. FT2 ^{(Talk | email)} 22:14, 3 April 2008 (UTC)

Switched back off, but enabled for users who request it for anonymity. It is still logging all reverts. --uǝʌǝsʎʇɹoɟʇs(st47) 10:35, 4 April 2008 (UTC)

Thanks. I'm still rather uncomfortable that this is possible at all, tho. Anyone can simply create a new account (like User:RTFA did) if there are privacy problems. If an edit is made, there needs to be a place where this edit can be discussed with the person who made that edit. Without (public) attribution, this would basically become a role account. --Conti|✉ 14:38, 4 April 2008 (UTC)

I don't like that it doesn't show the user in the edit summary, either. -- Naerii 23:39, 3 April 2008 (UTC)

I would strongly suggest that the bot transparently say who is actually making each edit. I am fairly uncomfortable with it not doing so. SQL^{Query me!} 01:06, 4 April 2008 (UTC)

• Not sure if I'm allowed to post here, but I don't like this bot at all. Reverting edits without actually looking at the edit seems counterproductive. The idea that you can know that something is a BLP violation without actually looking at it seems ridiculous at best to me and the fact that I'm unable to actually tell who is making the edits is annoying to say the least. This is a horrible, horrible idea for a bot. BLP issues should be approached sensitively. -- Naerii 23:48, 3 April 2008 (UTC)

See above for full discussion. This is one of the areas trialled, to see how it works in practice. The idea is that an experienced editor can sometimes -- not always, but possibly quite often - tell what is a problematic edit for a BLP. Simple vandalism, edit summary slurs, and childish comments in edit summaries, often happen and an edit summary that says "JOHN SMITH IS A PEN1S ON WHEELS!" is unlikely to need looking at, to decide it's probably not a useful edit......... FT2 ^{(Talk | email)} 00:14, 4 April 2008 (UTC)

Also, note that the trusted users who have access to the bots revert command should not, and most probably are not, blindly reverting, no matter what the edit summary says. Yes, the bot does feed the edit summary into the channel, however, that is along with a link to the diff, for inspection, before any action is taken. - Rjd0060 (talk) 00:22, 4 April 2008 (UTC)

If that's the case, then they can click "rollback" on the diff link, and there's no need for the bot.

I must stress that it is absolutely beyond the terms of the bot's approval for trial to fail to identify the user who initiated a revert in the edit summary. Wikis work because we have accountability. This breaks that. — Werdna talk 09:36, 5 April 2008 (UTC)

I think it's safe to conclude that ST47 was seeking to meet what seemed to be a valid users' criticizm arising, and he has since noted that consensus is very heavily against that, has put it back as it was, and has noted that if it were desired the capability could be used (but isn't presently). That's my reading of it. On this one though I find myself with utterly the same view as Werdna and ceased doing any tests involving edits whatsoever, until it was switched back again. FT2 ^{(Talk | email)} 14:16, 5 April 2008 (UTC)

Edit summary now updated (example edit). With the exception of the slightly misleading word "automatic" (it only acts on any occasion upon a specific manual command by a recognized user) the bot now uses proper edit summaries, including the user's on-wiki name, a link to the bot and its trial, and the edit summary they provided. FT2 ^{(Talk | email)} 21:52, 6 April 2008 (UTC)

Note in passing - updates to date during bot trial, A number of significant updates have been made during the bot trial. In summary the most important ones are: 1/ !reason command is now also restricted, 2/ bot reports all edits made by or through it, to the channel, as a safeguard and as confirmation, 3/ much "sanitizing" of input and fringe case fixing (some updates, then due for retesting), 4/ ambiguity fix to template ("date" might be start or end date), 5/ status and response messages to users much improved for clarity, 6/ context issue noted above by SQL is stated to be believed fixed, 7/ Bot now watches for "blpwatch" in edit summaries and checks + auto monitors the article (and reports in any event for clarity) if relevant, and so on. Bot has already been instruumental is spotting a subtle disruption editor on a BLP that was not otherwise noticed. Testing continues. FT2 ^{(Talk | email)} 15:11, 5 April 2008 (UTC)

I'm largely in support of this bot account. It seems that an alternative approach to BLP monitoring is necessary, and this new proposal seems to satisfy that requirement quite well. There will always be concern over a robot account reverting edits without supervision. There always has been: when the first antivandal bots were released, there was widespread concerns regarding them, that are somewhat identical to those being expressed regarding BLPbot (granted, of course, BLP monitoring is more complex than simply vandal reversions). Vandalism bots are now routine in the course of counter-vandalism. This bot is going to fill a useful role on-Wiki, and I support its introduction. Anthøny 12:26, 6 April 2008 (UTC)

• Question - could the bot provide rollback links to IRC so that users can click to revert rather than having the bot revert? --Random832 (contribs) 15:57, 8 April 2008 (UTC)
• Could the bot attempt to provide a "micro-diff" on IRC so that users can get a sense of the content of the edit? (like e.g. Popups) - I'm aware this would be challenging since everything has to fit in ~400 output bytes. --Random832 (contribs) 16:12, 8 April 2008 (UTC)

  Actually, I suppose it could put it in a TinyURL... SQL^{Query me!} 12:39, 9 April 2008 (UTC)

• Does this bot really go and reduce article content by 95% just like that, with no checking? see [2] --Enric Naval (talk) 16:32, 20 April 2008 (UTC)

Here we actually have a problem with somebody who gave the bot a command to do that (see the edit summary), rather than with the bot itself. People who have access to these restricted commands need to know when they should and should not use the bot. - Rjd0060 (talk) 16:45, 20 April 2008 (UTC)

Isn't the bot in this case doing exactly the same thing that an user can accomplish by using the "undo" button? Dunno why this command can be issued on the first place, since editors can just revert the article themselves, just like this user did twice before the bot and once after it. (I assume there are cases where the command is useful) --Enric Naval (talk) 16:57, 20 April 2008 (UTC)

I am rather worried by that edit. There was really no reason to use the bot to make that edit, and it seems to me that it was used solely to create a false sense of authority over the issue. That's surely not how the bot is supposed to be used, right? --Conti|✉ 17:30, 20 April 2008 (UTC)

I don't see a problem. When I'm on a better connection, I'll be looking for sceptre to make this bot's purpose more clear, however the user can be sanctioned for that edit just the same as any other edit. 12.185.224.87 (talk) 17:34, 20 April 2008 (UTC)

I'm st47 on airport wifi, by the way, can't be bothered to log in. 12.185.224.87 (talk) 17:35, 20 April 2008 (UTC)

That's true, of course, and it's certainly not the bot's fault when it's not being used properly. But a newbie might see an edit by "BLPWatchBot" and shy away from reverting. That's probably a minor problem, tho, and a short note on User:BLPWatchBot that there's nothing official about the bot (or that every editor is still responsible for every edit they make, or something) would probably suffice, IMHO. --Conti|✉ 18:00, 20 April 2008 (UTC)

There is definite (minor) concern here, and I see ST47 is aware, and will act accordingly (make the bots purpose clear with said editor). Thanks for bringing this to attention, Enric Naval. - Rjd0060 (talk) 20:02, 20 April 2008 (UTC)

{{BAGAssistanceNeeded}}

•  Approved. due to the anti-vandal like edits bot will run WITHOUT flag. β^{_{command 2}} 19:50, 11 May 2008 (UTC)

The above discussion is preserved as an archive of the debate. Please do not modify it. Subsequent comments should be made in a new section.