Per T107829, let's identify the top 100 bots (within the last 30 days) on each of the following projects and then randomly choose 25 to send survey invitations to:
(If some projects don't have 25 active bots, we'll have to revise this.)
| Status | Subtype | Assigned | Task | ||
|---|---|---|---|---|---|
| Resolved | kaldari | T107983 Administer community tech support satisfaction survey | |||
| Resolved | Niharika | T113404 Create a randomized list of bot accounts for 10 projects |
Here's the output: https://raw.githubusercontent.com/wikimedia/community-tech-tools/master/survey_bots.txt
Format:
<Wiki code> <Number of bots in sample> <Bot name> <No. of edits in last month> ... ... ...
And repeat.
SomeThoughts:
Hmm, looks like Arabic Wikipedia and Chinese Wikipedia have less than 25 active bots total. That would potentially skew the results from those wikis. We could limit it to 15 bots per wiki instead of 25, but considering that only 20% of survey invitations are typically successful (and probably less for bot accounts), I'm wondering if it's going to be worth the effort for getting 2 or 3 extra responses per wiki. The effort includes writing a custom survey invitation for bots (addressed to the bot operators), translating it into 9 languages, and creating separate MassMessage batches on the 10 wikis we are targeting. Personally, I'm skeptical, especially since a lot of bot operators are already going to be eligible to receive survey invitations by virtue of being active on those projects themselves.
@kaldari, I'm wondering how different the invitation would be, and if at all we can keep the invitation same and just add one/two more lines to it indicating that it's meant for a bot operator.
@NiharikaKohli: Adding one or two more lines to the invitation is about the same. Either way we have to spend a couple weeks prodding people to translate it (unless we hire a translation service to do it). I was skeptical of doing it for only 25 bot accounts, but if we have to do even fewer, I'm not sure it's worth the trouble. We also have to make sure that whatever number we choose will still work a year from now, which I'm not sure will even be true for 15 (Arabic wiki only had 19 bots running this month). 10 would be a safer bet. It just seems like it's going to greatly complicate the process for little or no gain.
Woah. Phabricator randomly chopped off my thoughts. Anyway, my suggestion was also to pick ten bots only. If you think there is nothing to be gained here, let's drop it.
As an aside, it would be nice to investigate some translation services for our future surveys.
I think it's not worth the trouble if we can't get a larger number of users to invite. It's still useful for us to have the script though since we might do some kind of bot-specific workshop in the future (or have other surveys with less strict restraints).