Topic on Talk:Growth/Personalized first day/Newcomer homepage

Hi Growth Team! Having glanced at the impact module a bunch since activating the homepage, I just wanted to offer some thoughts on it.

First, if new editors feel at all similarly to how I do, this module definitely seems to be succeeding at the goal of helping editors feel satisfaction about their impact. The main way you normally find out that others have seen your edits is when they get reverted, and this is a lot more pleasant than that haha.

Second, this is probably something that's a lot more acute for me as an experienced editor who edits tons of different pages than it would be for a newcomer who hasn't edited that many, but the selection of pages still seems off. It feels very heavily weighted toward pages I've edited recently (which have less views since I edited them by virtue of being recent), and sometimes pages I've only made quite small edits to show up. Adjusting the algorithm that picks which pages to show to favor ones where I've made large or many edits and place less weight on ones I've edited most recently might help.

Third, I think a likely impact of this module to consider is that it pushes editors toward editing more popular pages. When I see that 100 people have viewed my contributions to one page and 5000 people have viewed my edit contributions to another, that's a strong push toward editing the second page. If that's the case, the upside is that it'll allow their contributions to benefit more readers, and that it'll increase their odds of coming into contact with experienced editors who can help them if needed. However, the downside is that many more popular pages are already in much better shape, so they may have a harder time finding positive contributions to make, and the editors they encounter may be as likely to revert or bite them as to help them out.

I also had similar concerns about pushing new editors towards popular articles. The upside of that is that there would be more oversight of these new edits. If the algorithm were to be slewed so that it reflected the ‘’proportion’’ of the article that was altered, then a change to a stub article would look better than a similar edit to a large, popular one.

Hi @Sdkb and @Nick Moyes -- I'm sorry I missed this thread when you created it, but I'm glad you posted your thoughts. It's true that we the rules for displaying the articles in the impact module are geared toward brand-new users seeing the most impactful of their recent edits. We wanted them to have the moment of, "Wow, people are actually seeing what I've done!" If you're interested, you can see the actual rules it's using in the original task (which also contains a brainstorm of possible improvements).

"Among the set of 10 most recent pages that the user has edited, choose the 5 pages that have the most pageviews since the user first edited them (or in the last 60 days, if their first edit was more than 60 days ago -- because of constraints of the API). If a user has edited a page multiple times, their most recent edit of that page is what is counted for deciding whether it is part of the 10 most recent pages. For this sort, "null' sorts below 0 pageviews. If the user has edited fewer than 5 pages, show them all. Display those 5 pages in descending order of pageviews."

We have improvements planned for the module this year, and this is our nascent project page related to the work. I think an approach that would get at what you're saying is to add options and filters to the module, e.g. letting users select "most views" or "most recent" or "largest changes by bytes" or "exclude minor edits", and let those selections be sticky. What do you think of that idea?

Ah, thanks for the explanation of the algorithm! I think the thing that's making the selection feel off for more experienced users is the "set of 10 pages" part. I'm not sure how costly this would be in terms of API, but an improvement would be to look at the set of 50-100 most recent pages to which the user has made non-minor edits.