We all have our pet projects - things we wish SE would do but which seem to be forever stuck in the backlog pile. Other times, something that's well received gets forgotten in the shuffle.
Inspired by the post here and my own complaints about backlogged features, what would the community, both on Meta and network-wide benefit from SE paying attention to?
I want to see new user onboarding more seriously- particularly helping new users understand what the platform is, and how to use it properly (which in turn contributes to having a better experience, or at least avoiding unnecessary bad experiences). I'd like to see significant and intentional promotion of the Help Center in the site UI, and changes made to provide better JIT (just-in-time) guidance.
I think this is particularly relevant with traffic dropping as people start going to ChatGPT instead of SO to ask questions. See also Mithical's answer to What about the community is "toxic" to new users?, and BJ Myers' answer to "Come Take a Look at our New Contributor Indicator!".
Update: https://meta.stackexchange.com/a/398739/997587
status-completed(?): The return of Staging Ground to Stack Overflow
I'm not saying it's a perfect idea/solution, but there were some good ideas being explored, discussed, and actively worked on there, and it's sad to see that it got sidelined to jump on the AI hype train- especially when one of the main things I have seen in comments on reddit and hackernews as driving people away from Stack Overflow and toward ChatGPT is perceived toxicity from downvotes and closure, which supposedly Staging Ground would have helped to tackle.
At least on the bright side I suppose, work is being done on search, which could also help: if you find the answer without asking (a dup), then you can't get downvoted or closed :P
In 2020 we were promised that there would be quarterly roadmaps in order to rebuild trust with the community.
That lasted all of 8 quarters (2 years) and silently ended in 2022.
It would be good for the company to go back to this.
Started 2.5 years ago. Essentially died off with almost1 no consequences.
1Accepted answers were unpinned from the top which I admit is something. However, it did not really address outdatedness all that much.
There was also trending sort which probably helped a bit with the outdatedness. But I am not sure how much. Or how much is it even used.
A comment from V2Blast:
I think it would be helpful to identify what exactly you'd want to see worked on in this regard. (I'd also note that the Content Discovery initiative is sort of the successor to that project, in that it builds on learnings from that project.)
I never saw the Content Discovery as anything even remotely like Outdated Answers. These are two entirely orthogonal concerns to me:
Content Discovery tackles "How can I find information which is probably buried".
Example: some user trying to find a solution and the solution existing but not being able to be found.
Outdated Answers "How do we identify and deal with (either highlight or maybe remove) information which is visible but potentially wrong nowadays."
Example: a user easily finds a solution from 2010 and that is no longer a recommended path to take. It might have been wrong a decade ago as well but just the least worst. Or perhaps in the intervening years, new approaches that solve many more potential issues showed up. Might even be that problems with the old approach were found and it is dangerous to use it.
Part of the Oudated Answers is to also promote the recommended approaches one should take instead. However, toppling the historical inertial of old dangerous answers is a concern.
We solved this:
Only for accepted answers with low score. The same issue of reducing code quality around the world is still not solved where instead of a checkmark, the answers have accrued a score higher than all good or at least more correct answers combined. By virtue that the old answer was...older. Rather than better.
things we wish SE would do
Rebuild/Rework Area 51!
The platform is buggy, hard to navigate, and somewhat outdated. Is it really the best way to build new sites? (Plus, the friendly robots are a lie.)
I don't have any specific suggestions at the moment, but there are tons of feature-requests out there that could be reviewed and implemented.
The implementation of persistent notifications is something I did want for a long time. And since December 2022, I wanted the company to do the remaining work on this feature.
There are various issues:
These are just off the top of my head.
There was a podcast episode which I have not listened to. The episode notes read:
The inbox improvements were Radek’s graduation project. Not bad for a newbie.
Not everyone likes change, and the inbox change was no exception. So we looked into fixing that.
Yet, I am not sure anything was fixed1.
1Apparently the grouping of messages on own posts is fixed.
I would like to see more (public, community aided) research on ways we can improve the general perception of Stack Overflow Q&A and reduce friction while not negatively impacting our ability to curate the content it contains. It shouldn't be a constant conflict between driving people off and indicating content isn't useful or won't serve as a useful entry to the long-term system.
I'm all for people receiving the help they need even on posts that aren't a useful entry into the long-term system or are duplicates. However, the current way the system is structured works against that; it requires far too much effort from SME's to clean up this content once it has been answered, and the only tools we have to do so are often viewed as toxic due to the side effects they cause.
We should be able to indicate to the system that a post isn't a pearl without preventing the user from receiving help. The system should be able to get rid of or deemphasize these posts regardless of whether someone answered them. The only reason there's ever any real rush to shut down a post as soon as possible is because waiting makes it feel impossible to do so.
And, no, while I think AI aided search can certainly help people avoid asking things they should have asked Google, it isn't a solution to any of the above.
3 years old. Still not available even on Meta.Stackoverflow. Much less on any other site.
Over a 1.5 years old.
There are instances of things having ridiculously low to non-existent contrast in high contrast mode. And other instances of things just not looking great. I personally use high contrast + dark mode and it feels like that almost any UI update is basically not tested in that combination. A lot of the UI updates introduce more weirdness.
The whole initiative started a year ago. Few changes were made, some thing were made worse. For example, one of the areas that suffered was high contrast dark mode.
And yes, I am aware there has been more on the topic a month ago. As far as I know, no part of it has been released. Not even the better highlighting of questions with watched tags seen here. Which might (I hope) finally fix the basically non-existent highlighting in high contrast dark mode which has been reported since day 1.
There is not one or even a small number of things to point out here. Keyboard shortcuts are immensely useful for me, as it means I do not have to use the mouse. The problem is that there have been no updates to them. Not to my knowledge. There have been new features on the sites - like following or saving, yet those cannot be accessed via the shortcuts. Other old features like editing duplicates list or expanding comments without also trying to add a comment, or trying to edit your own post if you do not have full edit privileges, are not available, either. Almost nothing on the user profile is reachable via shortcuts, either - for example trying to go to questions, or answers, or all actions. And so on and so forth.
The keyboard shortcuts are not extensible programmatically, or I would have made a userscript for this.
I do not know when was the last time keyboard shortcuts were updated but it feels like there is no interest from the company to do anything with the project.
Since this is in the post concerning accessibility, also I already complained several times about high contrast dark mode: the highlighting of a selected post does not work in high contrast dark mode. Ever since one of the accessibility changes. For context: the highlighting is important in order to know which post are you currently working with. Without the highlighting it is exceptionally hard to know
In the old days, the links to meta and chat were front and center in the top bar. They're hidden under the SE/hamburger menu and its hugely unintuitive. This practically results in underutilisation of tools helpful for a healthy community - our second space (meta) and third space (chat) could use more love.
Especially considering the whole intent of the sidebar initially was to inline teams into regular sites, and teams essentially is its own thing now, it seems a better use of that space
Continuing to improve the close/flag dialogs. The text of the standard close reasons has been actively worsened by successive updates to the dialog, and they no longer communicate what they are actually meant to. For example, the "too broad" reason only mentions multiple questions, not questions that are too broad to handle in one Q&A.
Unfortunately it seems like almost all the feedback from April this year went entirely unacknowledged.
And in particular, Machavity's excellent proposal should be implemented:
I think flagging in general suffers from a poorly worded set of options. I wanted to try and address this so I mocked up what I think it should look like
Key changes
Red flags make deletion obvious
A common problem I run into flagging spam is that people don't always understand why we red flag. I had someone make this quote on Politics under a post that was spam
Mine was not a flag, but rather a vote to delete. Just being "nice"
A red flag is a vote to delete (not in the same way as a Trusted User, but the same net effect). I also slightly expanded the verbiage for spam to include promoting a website (a ton of spam is of the "visit my website" variety). I also expanded rude/abusive to include flagging unintelligible content (a reason few people know about).
Closed and Duplicate
Let's not mix terminology. The blue box says closed. The privilege says closed. "needs improvement" is highly confusing because it dumps you into the close dialogue box. Not every close reason is reopenable, either. Let's tell people we want it closed, and make it clear that closure is not necessarily a death sentence for a question or mean.
Mod flags
We need people to understand how and why you should use these flags. I've had several mods tell me that they prefer detailed mod flags when there's any question about a post.
NAA/LQP
This has been endlessly debated, but I think we can make some minor improvements. Most people do not know when to use Low Quality Post flags. Some verbiage cleanup would help and I try to denote that LQP flags are marked helpful if the post is edited (something that is not obvious).
There's third-party (community-maintained) libraries that work with the SE websockets that can be used to send chat messages automatically. The third-party library I'm using is super userful, but I'd love to see an official and documented API for sending chat messages and monitoring a room for new chat messages.
This is primarily for users developing chat bots to do various tasks on SE.
A while ago, I asked Let question-banned users know that they're banned even if their 6-month limit has opened up. The request received quite a few votes and was marked deferred as the responding employee was planning it for an upcoming project on new user onboarding, but the employee no longer works for Stack Exchange.
Essentially, when more than six months have passed since a user who's banned from posting questions has posted their last question, the system will allow them to post a new question to try and get out of their ban. However, users aren't told that they're still banned - it looks the same as if the user doesn't have any sort of ban. This creates a gap in communication:
Often, I've seen cases where users post a bunch of bad questions and get banned, and forget about the ban. Then, after a long period of time (in most cases, several years), they have another question they want to ask, so they go and post it. When they type up and post their question, at no point are they told about their prior history and that they were banned in the past due to bad questions. Then, when they try to ask another question, they all of a sudden find that they are banned without any advance warning.
In many of the "why am I banned" questions that come up here and on Meta Stack Overflow, this is the cause: they were banned due to bad participation years ago, and the previous ban was immediately re-imposed after they asked one question. As they'd forgotten about the previous ban, such users are left wondering why they were suddenly banned for just one question which may not have been negatively received.
I hope this didn't fall off the radar after the staff member ceased working for SE.
I suggest to bring back the Mentorship Project.
As its results clearly say, it was a huge success and improved the questions quality.
Near the end of the wrap up, it's said:
For larger communities like SO, we’re going to explore ideas for a scalable mentorship program
So please, go ahead. Explore those ideas.
Disclaimer: I was a mentor.
Probably only needed on Stack Overflow, maybe a few other network sites at most.
Put in the work to generate a list of 'master' or 'parent' tags for things that have all sorts of specific sub-tags associated with them, whether those sub-tags are about features or particular versions. These parent tags would not count toward the 5-tag limit on questions, and would be give pride of place in the tag list somehow; appearing first, but maybe also above or off to the left side of the rest of the tags, with special coloring, just like the required tags group on Meta (discussion, bug, feature-request, etc.). Sub tags could be associated with parent tags (similar to tag synonyms today) by vote or by moderator action; a parent tag of "Visual Studio" would prompt users for sub-tag choices of the various version tags, e.g. "visual-studio-2022". Conversely, using any sub-tag which has a parent tag would automatically apply that parent tag to the question.
Also create a process by which users can request and vote on a new master tag being created. Bonus points if it is an automated process (e.g. one that doesn't require posting a feature-request on Meta and having moderators handle it).
Create a system that lets answerers set version "labels" (or "tags") as part of the answer (rather than just relying on creative tag-formatting in the post body) whether that be specific versions or "this applies to all versions before/after version n". This would ultimately allow developers to provide sort and filter options based on what version labels are listed in given answers. Readers would be able to sort by version, or version + score, and/or set filters on answers "where version >= 3" (for example).
Way, way down the road, this could even allow for things like multiple accepted answers or showing certain answers side by side, so long as answers are labeled for different versions.
Just as with approved edits, rejected edits should cause a notification for the editor.
The request is (IMHO) wrongly tagged status-completed. Currently, there is a little info box telling you that the latest edit was rejected when doing another edit. Which is
The official argument for implementing this "solution":
Receiving an inbox notification saying you did something wrong at any given time (likely not while editing) would have a more negative impact on most users than reading a notice while they are actively in the flow of editing.
If that were truly the case, I would suggest removing all notifications for downvotes/closed question etc as well, instead showing them in the question editor when the user is actively in the flow of asking.
This is the only cost-effective solution to the obsolescence of information. Many people pick edge cases to have a total dismissal of the idea itself. SE should not make rules based on exceptional circumstances and shouldn't allow Hecklers veto to allow these users to ram against the most useful behavior that a wiki-like site has.
I'd love if SE would revisit the possibility to ignore users, see e.g. Add the ability to ignore users
I think the ability to ignore known trolls etc. would improve the mood of many users. When visiting a community, their actions won't be the first thing which greets you. For example here on meta.se, the first thing I saw today was a -64 question, which got bumped to the home page because the OP replaced an image with the same image. I'd rather not get greeted by such posts.
For the larger community, this will have the advantage, that users in good mood will be more patient when dealing with other things in their communities and overall make communities a nicer place to be.
