This is not within the scope of PageCuration, which is also only on English Wikipedia.

There are various places it could be put if we want to do this. There's a cost, though. The E3/Growth team spent a lot of time radically simplifying the create account page to increase signups. Anything added there could have the opposite effect.

I understand the goal here, but there's a trade-off, and various things to take into account (will people answer honestly, understand the question, etc.)

Also, the question does not actually match the text in the Terms of Use. For example, a user paid to edit for a charity is required to disclose, but would not have to answer yes unless they intended to promote the charity. (There are other mismatches as well.)

• Make the question direct and appropriate.
• Possibly include a hover-tip to the policy page and a short explanation.

I do not really see it affecting signups, and even if it does, policy clearly trumps that considering that the pillars to be upheld is a higher stake. There's no way to prevent new users from lying, so I do not see any reason to consider that at all. This is only meant for paid editors who are willing to edit constructively, for our sake and theirs.

I'm wondering: can we test the premise of this proposal quickly? For example, would it be useful if we randomly sample a subset of editors (the exact sampling method to be determined) and while they're logged in, show them a quicksurvey widget that asks them a similar question as to what's being proposed? (The exact wording to be determined). This way, we can learn what percentage of editors interact with such a question at all and what kind of self-reporting response distribution we receive.

I'm willing to do the community part of this program (community organizer) but I cannot take over as mentor, I'd guess the idea has a proposer, right?

In T157848#3060021, @leila wrote:

I'm wondering: can we test the premise of this proposal quickly? For example, would it be useful if we randomly sample a subset of editors (the exact sampling method to be determined) and while they're logged in, show them a quicksurvey widget that asks them a similar question as to what's being proposed? (The exact wording to be determined). This way, we can learn what percentage of editors interact with such a question at all and what kind of self-reporting response distribution we receive.

Although the idea is feasible, I do not know if it will be helpful. If we were to categorize on basis of paid editors, we'd have, 1) undisclosed paid editors unaware of policies (who might or might not respond) 2) disclosed paid editors (professionals who probably will) 3) undisclosed paid editors who are aware of policies (and are purposefully evading)

The clear-cut problem is the very first statement that you mentioned. When you throw the question at a subset, how do you know if your subset is accurate enough, also if you do have a target subset, what are you going to compare the result against?

Similar ideas have been mentioned in https://en.wikipedia.org/w/index.php?title=Wikipedia:Administrators%27_noticeboard/Incidents&oldid=770261399#General_thoughts_about_ToU_etc

@leila I'm in the process of selecting tasks from the community wishlist for the Wikimedia-Hackathon-2017 I'm wondering if you've some ideas for questions that @Ankit-Maity has raised? If any feasible/ helpful ideas come out of this discussion, we could turn them into research projects and address to participants during the hackathon, and we perhaps get some prototypes in return :)

@srishakatux to be super clear: I'm not the proposer of this task and not its owner. My comments on this task should not be considered as blockers for others to do something with it. :) At this time, I don't see a way for development to be done on this task given that there are still conversations about what should be asked, where it should be asked, and how we will measure its impact if a change is implemented.

@Ankit-Maity: Re your comments in T157848#3060021

• I suggested a cheap solution for us to get a sense of whether people interact with questions along this line at all. Depending on the question we ask, we should expect some level of interaction to be a minimum required to make this effort meaningful. For example, suppose we ask the users: In the past year, have you edited Wikipedia to promote content for a company while you were paid by the company to do so? (the exact question should be carefully designed). If we show this question to 20K logged in users and we get 100 responses, I'd say this way of data collection will most likely not work.
• Now suppose people interact with the question (with Yes/No responses), the question becomes what you raised above, that is: how do we make sure we are showing the question to the right people. Basically, I hypothesize that paid editing is a rare event and we need to carefully think about the way we sample to get some signal out of the data (for example, if we know paid editing is more prevalent in certain article topics, we can sample more from those topics.
• The last point you raised is also important: what are we going to compare it against? This is likely easiest for those who intentionally hide this information from their profiles (which is less interesting for this task). For that specific set, maybe we can compile a list of editors identified to be at very high risk for doing paid editing, ask them the question(s), and see if they disclose? We can then consider our assessment of very-high-risk as the ground truth or we can wait to see if there will be cases against them in the future, in which case we can have ground truth in the future.

btw, this topic is not my favorite topic. I'd much rather focus on improving the positive aspects and helping them become more magnified than playing a cat-and-mouse game. ;) but I understand that this is an important problem.