thanks for walking me through that! signing off as it sounds like figuring out whether using Commons' imputed title should be a new request.

thank you! re: DWIM, I don't think we will have a semantic DWIM and I suspect apps didn't deal with this because they have been testing in typeahead search (happy to be wrong), so my question is: will it make sense for us to put a "Did you mean [potentially pretty low relevancy lexical suggestion]?" there? is that better than either (1) putting the DWIM directly above lexical or (2) no DWIM or (3) maybe another creative option I haven't thought of? lots of tradeoffs and not ideal, I know.

agreed the missing title issue is not a launch blocker. before I sign off is it possible that we're seeing that behavior because the Goya portrait seems to be used exclusively in infoboxes and galleries (at least for frwiki and enwiki)? or is it still unexpected that in-article position would affect the title display since that is metadata from Commons? trying to think about how we want to handle this known bug, i.e, do we know the pattern is likely to be pretty rare and/or hard to fix, etc.?

if I could add one more question: what would an affordance for qualitative feedback look like?

absolutely, I had in mind to update this ticket with Justin's updated designs but thank you for swapping in the better ticket!

I see the question about whether to include question marks as part of the semantic heuristic. thanks for raising! I prefer not to add logic about question marks for this test. since that wasn't part of the heuristic identified for English by Research and since question marks can also be used as wildcard characters, I think the lower risk choice here is to avoid changing the current behavior for now.

confirming both the image carousel retest and semantic search test coming up will impact only logged-out readers

tested the ones above
(1) https://commons.wikimedia.org/wiki/File:Adolph_Menzel_-_Fl%C3%B6tenkonzert_Friedrichs_des_Gro%C3%9Fen_in_Sanssouci_-_Google_Art_Project.jpg

copying from Discord chat for phab records:
@stjn agreed that the expanded sections was the driver behind the retention decrease in the ToC experiment (also why the results matched the prior Expanded Sections experiment so closely). that being said, another thing we found in follow-up qualitative study was that logged-out readers preferred using the section headers for that type of navigation because they just tend to gloss over buttons--they basically didn't notice either ToC treatment so I don't think having the ToC button + section headers is a winner either. we're working on getting those qual findings written up for the project page too.

We tested a ToC on Minerva and saw a statistically significant decrease in logged-out reader retention (we think this is a closer proxy for whether readers had a good experience than some of the other metrics critiqued above); based on these results, we decided not to roll the feature out. For more details on the test: https://www.mediawiki.org/wiki/Readers/Reader_Growth/Mobile_Table_of_Contents#Results

as arms B&C are going to open MMV from carousel tap, we should upgrade this task to launch blocking and we can discuss any unsolved edge cases again tomorrow after post-train testing in prod