Up until now, I was thinking to be fairly sure what an automated edit is, and what constitutes mass editing or however you wanna call it.
Through recent discussion, my image of what needs to be discussed with the community was proven to be seemingly wrong, as I was told that e.g. every MapRoulette Challenge constitutes something that needs to be discussed with the community prior to going live.
I was always thinking that any mechanism that edits without a human reviewing every change that is about to be made is considered automated editing. Which I completely understand because the OSM Database is so complex that it is nearly impossible to automatically do any edit without side effects. That’s why a human needs to review every change that should be made and decide for every object individually if the proposed changes make sense.
But now I have heard about a new understanding of this rule. It was unfortunately not possible to find a clear definition, but the definition I destilled for myself was
“Everything that significantly impacts the mapping of at least one feature as a whole is considered automated editing and needs to be discussed before doing.”
But that is still not clear cut enough for me and I would apprechiate your words on what I should discuss with the community before doing.
Through recent discussion, my image of what needs to be discussed with
the community was proven to be seemingly wrong, as I was told that e.g.
every MapRoulette Challenge constitutes something that needs to be
discussed with the community prior to going live.
I’d say there is considerable fuzz in that definition. But from past
experience, MapRoulette participants often blindly trust the challenge
creator, so the challenge creator bears a lot of responsibility. Adding
a challenge to MapRoulette can come dangerously close to “making an edit
without humans reviewing every change”, and there has been demonstrable
abuse of MapRoulette to precisely that “mechanical edit cloaking” effect.
“Everything that significantly impacts the mapping of at least one
feature as a whole is considered automated editing and needs to be
discussed before doing.”
I’d perhaps remove the “is considered automated editing” from that, but
otherwise it sounds like a very good rule of thumb. If there are 30
objects of some class in OSM and you make a MapRoulette task to add 300
more (or a MapRoulette task to change those 30 from foo=bar to bar=baz)
then it is definitely a good idea to discuss this somewhere first.
Consider this a responsibility not only towards the community as a whole
but especially towards those who participate in your MR challenge. If
you lead them to innocently do something that the community then finds
issues with, you could have soured all of OSM for them with your
reckless enthusiasm.
The answer is a bit unsatisfying, but the reality is “it depends”.
More specifically, edits which are likely to be controversial, scrutinized, or disputed must be discussed with the community. The problem is that there is never a surefire way to know which types of edits are likely to fall into that category. That’s why you must get to know the community you’re editing in.
Edits which would be absolutely no problem in the United States can draw intense scrutiny in Germany, and a whole range of different reactions in different places around the world. You have to learn the zeitgeist of a community to understand it. There is a reason why tools that involve automation are embraced in some places and disliked in others.
This can be frustrating for newer users that are looking for strict and clearly spelled-out community rules, but we’re all humans – and humans and communities are all different.
If a company or non-profit sets up the MapRoulette campaign, they would need to follow the Organised Editing Guidelines and discuss about it beforehand.
I don’t know if there’ll be a 100% clear definition. This might be a “I know it when I see it” situation.
I would suggest that challenges that have the potential of ending up being bot editing should be consulted on.
Someone who edits OSM across tens of kilometres and the entirety of their edit is accepting iD’s suggestions is performing a bot edit IMO.
To make it a human edit, I expect human judgement. For example, in choosing which of iD’s suggestions to accept, or choosing what to map, or genuinely answering a question “is this A or B” that the tool is asking earnestly.
A bot edit is changing crossing_ref=zebra to crossing:markings=zebra in an entire town in a single changeset because a tool suggests it. A human edit is changing that tag in StreetComplete or Every Door when you are standing in front of the zebra crossing in question.
Bot edits can be OK if there is local community agreement that it’s a good change - for example I understand the change to crossing:markings is generally accepted in Canada, or to give a recent example, the deprecation of cycleway=opposite. But if all you’re doing is accepting what got put into the iD presets repo without any discussion with local community, then you’re going to cause problems sooner or later – as the other thread documented. This is the whole reason that “automated edits” policy was put in place.
To me, a bot edit is someone who’s never ridden a bicycle going into a MapRoulette challenge and pressing “accept” on all suggestions to upgrade cycleway=opposite. A human edit is someone who rides a bicycle going into a MapRoulette challenge for their local area, viewing each of the suggestions to see if the proposed new tagging matches aerial imagery, perhaps consulting OSM history to see if something not matching imagery is a new change on the ground that was surveyed in-person, and making those changes. Both could be from the same MapRoulette challenge. It is up to MapRoulette to make it as hard as possible to make that first edit.
It’s possible for the same change to be done in a bot manner or in a human manner. The outcome could be the same. But the process matters. Because the process helps us avoid problems. Pilots don’t go through their landing checklist every time because they expect the outcome to be different – they do it in case the outcome is different.