An idea about fixing the bot

How about rejections by the bot go to voting so actual humans look at the submission in question and confirm/train it? All that money won’t go to waste by shutting it off and people get to actually vote things out rather than a machine making mistakes.

Because if you have reviewed at times when “the bot” is down you would understand why it is required.

The vast majority of the Rejections are correct, unfortunately there will be “false negatives” but this is better than the absolute dross that would come through without it.

What use would your idea serve because what you are saying is that all nominations should be human reviewed so we quickly go back to nominations taking months or even years to be decided on.

Might as well just switch the ML tool off entirely, then.

…and go back to 50 bazillion coal subs clogging the reviews and subs taking over a year to get through, just like the good old days.

I have reviewed just fine while it’s down not to mention it would help more people get the wayfarer badge in game. I did not say “all reviews should be human” as my post clearly calls for just the ones disapproved by the bot to go into voting.

If a submission is really that bad then people will vote it down, things taking forever to be decided on sounds more like a lack of people voting rather than the bot speeding things up, I guess I’ll spell it out; I’m ok with it passing things but I have noticed submissions denied by the bot get passed by the community if they do go into voting so that’s actually also causing more time for new POIs to get approved while showing that it’s not doing the intended job it was designed for.

It works opposite from what you propose. The AI approves nothing, just rejects if it believes the nomination is ineligible.

Humans double check what the AI did not reject. Staff will check the appeals for the nominations (rejected either by AI or the community)

See also:

How about this as a workaround:

If ML refuses a sub, the submitter can use an appeal and choose whether to force it into normal review, or to go to the appeals staff.

I’ve just had a Lord Provost lamp post blocked by ML. I’ve appealled, but frankly I don’t trust that a non-Scottish staff member will understand what this lamp post is, and why it isn’t generic. I would much rather have it reviewed by locals. :person_shrugging:

The problem with sending AI Rejections to reviewers is that you are opening all the real Coal to the “Accept Anything as Any Stop is Better Than No Stop” brigade :frowning: Great in principle but IMO would cause more issues in practice.

This just proves that you do not understand the process. The AI bot only “Rejects”. Apart from a short time when it was introduced it does not accept. So “All AI Rejects” + “Those sent to Reviewers” are the vast majority with only the minute amount that are sent for “Internal Review” that would not be sent to Reviewers which takes it back to the days of Nominations having birthdays.

How about if the auto system declines a submission you have the choice to appeal or return it to draft status.
This would give the submitter a chance to reflect and get advice on the submission and make changes including photos and text.
Instills good practice of reflect and change
It may save an appeal being needed ( which saves on human costs)
Hopefully a better submission passes the auto system and a suitable submission gets assessed by reviewers - it can still be rejected of course but at least you have had a second chance.
We didn’t have draft status previously perhaps we should use it more creatively.

Maybe. But sometimes it seems impossible to get certain things past ML, no matter what description, title or photo angle.

In my case I’m going to keep trying (assuming the appeal isn’t successful), but I bet it will keep seeing a generic lamp post.

I have sometimes had to try multiple times different photos different text…..and I think the UK postbox topic demonstrates how problematic it can be.
But if the appeals are not working either then maybe it is fundamentally problematic?

I just think that when you combine the existence of ML with the staff inability / unwillingness to spend a little extra time to really consider the local aspect, it becomes problematic, and certain things that are absolutely eligible become impossible to get through because the community never sees them.

But maybe I’m being too cynical. My appeal hasn’t been rejected yet :joy:

I’ll start a new thread if / when that happens…

I remind myself at those times that it may not be as wonderful as I think it is.

Winnie The Pooh Thinking Deeply

Totally agree that there are issues but companies has to decide what is acceptable.

  • If AI gets 99% right and spending £1m to improve it to 99.1% this is not acceptable.
  • If staff get 99% right and by taking twice as long on each nomination increases that to 99.1% it is not acceptable.

I was not here before eMiLy but from what I have heard I will take the occasional hassle with an appeal and even the odd “not accepted when it should be” rather than going back to wading through the coal.

I literally just had it autoaprove at least 3 nominations in the past 6 months.

The machine learning model cannot approve Wayspot nominations.

Wow took a while to find the link where they announced that rejections had been turned on but not accepts. And then when I posted, the link went on the wrong topic :woman_facepalming:

When auto-accepts were enabled, the emails said “our team” had made the decision. I haven’t seen any announcements or reports that these have been turned back on.

The auto system can currently only reject nominations.
It can approve photo and text edits.

IMO They should drop the idea of bringing Auto Accepts back.

An incorrect Rejection can be Appealed and / or re-submitted.

An incorrect Accept means it is in the game and it would takes someone that is not a “i’ll leave that as it’s handy for me” to report it.

Also hoping they have discarded this idea. I did love auto accepts, but requires all submitters to be acting in good faith and within the criteria guidelines.