On 2/27/07, Rich Holton richholton@gmail.com wrote:
FWIW, I think that decreasing the average workload of the "hyperactives" should be a major priority. The quality of their work will improve, the number of fatigue-related problems will decrease, and there may be less hesitation in de-sysopping one of them.
I wholeheartedly agree. One thing that jumps out from the list of deleters that geni posted (http://en.wikipedia.org/wiki/User:Dragons_flight/deleterlist if you missed it) is that about a third of the 840 admins listed have not made a single deletion in the last 500,000. Hardly an example of leveraging of the long tail!
This is not to say that these admins aren't doing their jobs. Indeed, Charles Matthews and Mindspillage are both in the first screenful of 0.00% admins, and both are quite busy enough with more important things than deletions. :-) However, there might be plenty of admins (and non-admins, for that matter) who wouldn't mind doing a few deletion-related tasks a week, even if (especially if?) they don't want to focus on deletion full time.
What might work is to have a process similar to jury duty: have a registry of volunteer admins who are assigned a very small, manageable number of tasks randomly by bot once per week. This would clear backlogs, and take a lot of the pressure off the current top workers to clear backlogs. Anyone who screws up, of course, would be stricken from the list of volunteers in minor cases, or could face desysopping in cases of flagrant error.
The coding of the assignment bot is certainly beyond any of my skills, but there are other other processes that currently rely on an overworked core of regulars who tend to burn out, which could be improved by introducing capable but non-daily-regular editor input. These are (I believe not coincidentally) the same processes that are most frequently described as "broken": !votes in RFA, AFD, and to a lesser extent the other xFDs.
By explicitly balancing more of the editors' load, we can prevent various bad things like burnout and criteria drift that result from overwork at the core.
Implementation of any of this would, of course, require some (mostly social) shifts in the affected process(es), but as this is not a finished proposal, I will assume that the community is smarter than I am and is capable of ironing out most bugs beforehand.
--Michael Noda
Michael Noda wrote:
On 2/27/07, Rich Holton richholton@gmail.com wrote:
FWIW, I think that decreasing the average workload of the "hyperactives" should be a major priority. The quality of their work will improve, the number of fatigue-related problems will decrease, and there may be less hesitation in de-sysopping one of them.
I wholeheartedly agree. One thing that jumps out from the list of deleters that geni posted (http://en.wikipedia.org/wiki/User:Dragons_flight/deleterlist if you missed it) is that about a third of the 840 admins listed have not made a single deletion in the last 500,000. Hardly an example of leveraging of the long tail!
Part of the problem from my perspective is figuring out what to do in some of these tasks. I typically do this "janitorial" work in bursts---I'll do a bunch of recent-changes patrolling for three weeks, because I'm in the mood for it, or clear out copyright violations for a day, then I won't do anything for two months. When I come back to one of these tasks after an absence, it takes concerted effort to navigate the web of rules and templates and informal policies. Even simple stuff like, "so what's the current policy on warning/blocking vandals, and which of the 100+ user warning templates should I use?". It tends to lead to me doing a small subset of things I feel confident in my familiarity with (new-article and recent-changes patrolling, mainly, and some prod-clearing), and avoiding anything else. To be fair, some parts of our "process" do have concise documentation on the relevant page with nicely laid out bold text saying something like "If you're just wondering how to do [x], do these steps: 1, 2, 3, 4", and in those cases it's easier to get into them.
On the topic of bots but with a slightly different intent, some automation of drudgery would help make this work more appealing. Nobody likes to do work that is super-repetitive and requires no human judgment. Clearing out the batches of images tagged "orphan fair use" is annoying and I usually avoid doing it, because at least 50% of them are trivially mistagged due to not being orphaned (perhaps they were when tagged, but they aren't now), and so I end up spending most of my time just deleting the template and pasting in a "not orphaned" edit summary. A bot could do that for me, letting me as the human look only at the *actually* orphaned fair-use images to decide whether to delete them or not. There are a bunch of other examples. Incidentally, two nice examples of automation are the bots clearing WP:AIV, and the bot that goes around adding Template:SharedIPEDU to talk pages of IPs from universities and school districts.
-Mark
On 27/02/07, Delirium delirium@hackish.org wrote:
When I come back to one of these tasks after an absence, it takes concerted effort to navigate the web of rules and templates and informal policies. Even simple stuff like, "so what's the current policy on warning/blocking vandals, and which of the 100+ user warning templates should I use?". It tends to lead to me doing a small subset of things I feel confident in my familiarity with (new-article and recent-changes patrolling, mainly, and some prod-clearing), and avoiding anything else.
Ah, yep.
To be fair, some parts of our "process" do have concise documentation on the relevant page with nicely laid out bold text saying something like "If you're just wondering how to do [x], do these steps: 1, 2, 3, 4", and in those cases it's easier to get into them.
Rewrites for clarity are always a good idea, if you can form the patchwork of special cases into something coherent without losing hard-fought detail.
- d.
On 2/27/07, Delirium delirium@hackish.org wrote:
Part of the problem from my perspective is figuring out what to do in some of these tasks. I typically do this "janitorial" work in bursts---I'll do a bunch of recent-changes patrolling for three weeks, because I'm in the mood for it, or clear out copyright violations for a day, then I won't do anything for two months. When I come back to one of these tasks after an absence, it takes concerted effort to navigate the web of rules and templates and informal policies. Even simple stuff like, "so what's the current policy on warning/blocking vandals, and which of the 100+ user warning templates should I use?".
Test1-7. The rest are for people who enjoy categorising too much.
On the topic of bots but with a slightly different intent, some automation of drudgery would help make this work more appealing. Nobody likes to do work that is super-repetitive and requires no human judgment. Clearing out the batches of images tagged "orphan fair use" is annoying and I usually avoid doing it, because at least 50% of them are trivially mistagged due to not being orphaned (perhaps they were when tagged, but they aren't now), and so I end up spending most of my time just deleting the template and pasting in a "not orphaned" edit summary. A bot could do that for me, letting me as the human look only at the *actually* orphaned fair-use images to decide whether to delete them or not.
Already exists. Except that it also delete the orphans.
Delirium wrote:
Part of the problem from my perspective is figuring out what to do in some of these tasks. I typically do this "janitorial" work in bursts---I'll do a bunch of recent-changes patrolling for three weeks, because I'm in the mood for it, or clear out copyright violations for a day, then I won't do anything for two months. When I come back to one of these tasks after an absence, it takes concerted effort to navigate the web of rules and templates and informal policies. Even simple stuff like, "so what's the current policy on warning/blocking vandals, and which of the 100+ user warning templates should I use?".
It would seem that such a multiplicity of warning templates alone would turn people off the task. Are they all regularly used? Could some be combined?
Ec
Delirium wrote:
Part of the problem from my perspective is figuring out what to do in some of these tasks. [...] When I come back to one of these tasks after an absence, it takes concerted effort to navigate the web of rules and templates and informal policies. [...] On the topic of bots but with a slightly different intent, some automation of drudgery would help make this work more appealing.
Is it possible to encode more of this knowledge in tools, not just bots? I just recently started using [[WP:TWINKLE]], and having drop-down menus of the various warnings and templates has definitely helped me be more precise and situation-appropriate. Are there similar things available for admins?
William
Michael Noda wrote:
This is not to say that these admins aren't doing their jobs. Indeed, Charles Matthews and Mindspillage are both in the first screenful of 0.00% admins, and both are quite busy enough with more important things than deletions. :-) However, there might be plenty of admins (and non-admins, for that matter) who wouldn't mind doing a few deletion-related tasks a week, even if (especially if?) they don't want to focus on deletion full time.
What might work is to have a process similar to jury duty: have a registry of volunteer admins who are assigned a very small, manageable number of tasks randomly by bot once per week. This would clear backlogs, and take a lot of the pressure off the current top workers to clear backlogs. Anyone who screws up, of course, would be stricken from the list of volunteers in minor cases, or could face desysopping in cases of flagrant error.
I don't even think that the random assignments. A lot of the people who might do a few without becoming hyper look at the RfA questions, and quickly decide that it's not worth it to put themselves through that process. The other problem with random assignments is that unlike the hypers many of these people will refuse to work on subjects that they do not understand.
Ec