Research talk:Bot user
Global user group to categorize bots
Superm401, I very much like the idea of a dedicated global user group, created solely for analytics purposes and with no rights associated, to accurately categorize bots.
A few questions:
- is there any know precedent of descriptive user groups with no special rights?
- presumably, once assigned to such a group a user will indefinitely stay in that group. How would hybrid cases (human accounts occasionally behaving like bots) be handled?
- who would need to be involved in the creation and retrospective application of this new group cross-wiki? I am particularly worried about projects lacking BAGs where bot categorization might be problematic.
- how could we enforce the application of this group to newly created bots? Ideally, new bot categorization should happen before their first edit to avoid retroactive discounting of their activity. This may not be practical/feasible in a number of conditions.
- @DarTar: I don't know of any precedent about that, but I could be unaware of one.
- Presumably, "flood" (where a human temporarily gets the bot flag for verbose or semi-automated edits) would not get the no-rights bot group.
- I'm not sure. It would probably be good to talk to the stewards, among others.
- I'm not sure. It could be a global policy (shouldn't be overly controversial, since it doesn't actually grant any rights), but making it consistent would take work (and even then it won't reach 100% uptake). Superm401 | Talk 05:18, 24 December 2013 (UTC)
- Also, I think the link is http://lists.wikimedia.org/pipermail/wikitech-l/2013-November/073065.html . Gmane seems to be temporarily down. Superm401 | Talk 05:21, 24 December 2013 (UTC)
I don't understand the need for an actual user group. Erik's system works fairly well, it consider a user a bot if it's flagged as such, or flagged as bot on 10+ other projects, or (?) it contains the word "bot". Just publish the list of such usernames on Meta and let people edit it, if you want manual intervention. Relying on bureaucrats, stewards or Special:UserRights in general sounds an unnecessary complication, especially as userrights always require some process while editing is just editing. --Nemo 09:53, 18 January 2014 (UTC)
- @Nemo bis:, the goal here is two-fold: determine the cheapest and most accurate way to extract and store bot status for the purpose of research/analytics. Superm401's proposal to piggyback user groups primarily answers the latter, i.e. not how bot status is determined, but how and where it should be stored. Aside from determining the cheapest/most robust/most accurate way to identify a bot, we need to figure out how to make this information easily and consistently available to a variety of systems the WMF analytics dev and platform teams are building (such as Wikimetrics or the Editor Engagement Vital Signs dashboard) as well as tools designed by community members and researchers. The fact that internal and external research, feature analytics, wikistats, and 3rd party tools relying on the RC bot flag all rely on different ways of counting/discounting bots, makes it really hard to produce data of a usable quality. Whatever solution we come up with, it will need to be machine readable and not arbitrarily editable by anyone as it will be used to generate official stats. Hope this clarifies the scope of the discussion. --DarTar (talk) 20:04, 21 January 2014 (UTC)
Cleaning up this page
@Halfak (WMF):: what do you think, I feel this page should just describe the concept of a bot user and we should move all the identification discussion to Research:Identifying_bot_accounts. Any objection? --Dario (WMF) (talk) 21:53, 11 September 2014 (UTC)