<Note this reply is written with my enwiki community member hat on, and in
no way represents anything official>
On Fri, Mar 6, 2015 at 5:19 AM, Ricordisamoa <ricordisamoa(a)openmailbox.org>
wrote:
It is complex and bureaucratic on the English
Wikipedia, i.e., less than
1/890 of the projects.
I note that enwiki's process for receiving the bot flag and rules around
bot editing are "complex and bureaucratic" in large part because what one
person thinks is an obvious fix that no one could object to (e.g.
"==Section==" versus "== Section ==") turns out result in a huge
outcry
when a bot is doing it all over the place.
The idea is that the review process (which is basically just having one of
a list of experienced bot operators look over the proposal for problems,
then review some sample edits) will hopefully catch problems before they
become a big deal, and the rules make it easier to stop for (hopefully)
calm discussion rather than arguing while perceived disruption continues.
Instead, I think bots are easily tricked by edge cases, whereas human
intervention usually decreases the chance of
mistakes.
On the other hand, a tool may be more aggressive with proposing changes
that would be fooled by edge cases while relying on the human to fix it
before submitting. Even if the tool is not being more aggressive, the human
is vulnerable to missing an error through inattention or through
misunderstanding their responsibility and blindly clicking "approve".