With your figures, assuming 100k edits being created.
Cut the "bullshit" by half leads to 47.5k useful edits (95%/2; or at least ones without bullshit) plus your original 5k leads to 52.5k useful edits. Except we've only a third of the people - assuming equal distribution of edits - so 17.5k of edits are now useful. More than a 3 fold increase in useful edits in the same amount of time.
So in fact if you change the system to put off a lot of people in your hypothetical manner you can improve the useful edits.
This is akin to pricing out "toxic" customers. Or having complex methods that filter out the more intellectual as those with most power (arguably the problem being ascribed to Wikipedia in the article).
Assumptions here make the figures entirely unrealistic but I think the general thrust is nonetheless sound?
Cut the "bullshit" by half leads to 47.5k useful edits (95%/2; or at least ones without bullshit) plus your original 5k leads to 52.5k useful edits. Except we've only a third of the people - assuming equal distribution of edits - so 17.5k of edits are now useful. More than a 3 fold increase in useful edits in the same amount of time.
So in fact if you change the system to put off a lot of people in your hypothetical manner you can improve the useful edits.
This is akin to pricing out "toxic" customers. Or having complex methods that filter out the more intellectual as those with most power (arguably the problem being ascribed to Wikipedia in the article).
Assumptions here make the figures entirely unrealistic but I think the general thrust is nonetheless sound?