Evolving comments: An experiment to encourage engagement and follow-up questions

Changing the commenting policy

The new comment policies look like you want to encourage tangential noise on posts. As a post author, I generally have little interest in helpdesk'ing for individuals (policy 2), hearing arbitrary stories (policy 3) or even an entirely new discussion (policy 4).

Given the desired increase of comments that have little to do with the original author, do post authors get an option to opt-out of comment notifications?

What will happen to the content once the experiment is over?

At first glance this looks a bit like you want to turn a Q&A site into a forum (or Reddit). Which is kinda the opposite of the original idea for Stack Overflow. Doesn't mean this is automatically a bad idea, but I'd cautiously evaluate the effects this might have on how people might use the site.

To me Q&A isn't necessarily the only option, but it is the most valuable one for creating and curating content for later visitors. So even if we don't start with a Q&A pair directly and a longer discussion arises, I think in the end the result should be a question and an answer, to ensure a high signal to noise ratio.

I see a danger here that the better the discussion features are, the less likely people will be to post actual answers. I think one idea worth experimenting with would be a "promote comment to answer" action, to push people towards distilling comment threads into real answers. This is much more viable in your experiment as comments have actual formatting options.

It might even be worth it to experiment with allowing other users to promote comments to answers even if they didn't write them themselves. This adds a lot of potential problems so it would need much more care, but I'm imagining something more like an automatic Community wiki answer with a non-removable attribution to the comment it was created from. But this is a bit more of a crazy idea, not sure how people would react to this.

support question about what appears to be a massive regression in functionality, tantamount to introducing a breaking bug:

Additionally, we will modify the flagging modal to disable the “no longer needed” flag type for the duration of the experiment. If users attempt to use it, they will be directed to the help center article on commenting, which will have an updated copy for the duration of the experiment.

Why? How are we supposed to flag comments that are no longer needed, then? This experiment will in no way eliminate the phenomenon of comments not being needed any longer. I say this as someone who has cast nearly 30,000 helpful NLN flags on Stack Overflow alone.

we found it wasn't the ideal solution for these specific, contextual questions. This was partly due to lower visibility compared to the original Q&A page

I'm not a fan of the idea of making comments more visible than they already are. Giving them more visibility is going to distract from the Q&A. The example you showed takes up a huge amount of space and I find it very distracting. Even if the default is apparently to have them collapsed, I assume you won't be changing that a handful of highly-scored comments remain inline automatically, and the new form factor is much bigger than it is currently. Stack Exchange is supposed to be a place people can get answers to questions without the noise that is common in other places on the internet.

we will modify the flagging modal to disable the “no longer needed” flag type for the duration of the experiment [...]

I don't see why disabling NLN flags is necessary. Why would you do this? Unless your rule updates remove existing reasons not to comment (do they?). Please, don't do this. How are we going to flag the vast amounts of NLN comments that are created every day during this experiment?

Our primary metric will be an increase in constructive engagement within comments. We will also closely monitor comment flags and deletion rates to understand the impact on content quality and moderation load.

What about increase in unconstructive engagement? You seem to have a filter on your glasses to only look for success. Especially given that you're removing the NLN flag for the experiment.

Users also expressed confusion about the distinct purposes of comments, Discussions, and Q&A.

After reading the motivation for your upcoming experiment, count me in. Now I'm also confused. To me it reads like you want to have everything in one page and everything can be everything. Code in comments, discussions in comments, follow-up questions in comments. What can possibly go wrong? And will the confusion be reduced or reach new heights? The experiment will hopefully show.

But let me emphasize one more aspect.

At the very least, show all the extended, discussionary, code containing comments collapsed initially. From a perspective of knowledge consumer, already now the pages are unfortunately low concentration of knowledge (much space in between question and answers) and I can imagine this becoming even much worse, next to unusable.

Our primary metric will be an increase in constructive engagement within comments.

That seems positive, but ... engagement is not the intended purpose of comments. They're primarily a collaborative way to help improve questions / answers. I.e. if I post as a comment (for example) "What version of python are you using?" the asker can either reply in another comment or edit the answer into their question. The former is higher engagement but the latter is what we actually want them to do.

Comments are deliberately less fully-featured than questions/answers. Not because nobody bothered to add proper markdown and threading support, but rather to encourage people to use questions and answers instead.

It seems pretty obvious to me that making comments more visible and fully-featured and making it easier to comment will result in more engagement with comments. It's less obvious that this is an improvement for the site as a whole. It's probably worth thinking about it you can find any way to measure that, e.g. if less askers are editing their questions that's a bad thing.

You got my attention; this is an interesting set of ideas. However, in light of very real problems we've seen with our current system, I must ask: How does a new user tell the difference between Question Response Input Box #1 and Question Response Input Box #2?

One of these allows pretty much any content you can write with no consequences, but posting the wrong thing in the other one leads to negative comments, negative votes, and sometimes it can even lead to the semi-automatic suspension of one or more Responding-to-Questions privileges. Of course, if you're familiar enough with SE (and this new design) to know that #1 is properly called a "comment" and #2 is an "answer", you can probably avoid such a travesty (and maybe even exploit it a little on purpose—leaving all your potentially controversial responses as comments). But how does a new user manage? Alternatively, how can we, the people who moderate the site, deal with this in a way that doesn't suck for everyone involved? We have an answer-to-comment mod tool, but it's always seemed quite lacking to me given that it's only available to mods.

In the same way, as a reader looking for an answer, my eye is drawn to the answer above "Your answer". But this isn't an answer at all but a comment. Comments are just so more visible in this mock up that I wouldn't be able to find the answers proper.

This doesn't even touch the fact that there's yet another Input Box: Questions. There's an undeniable appeal to using comments as a substitute (no downvotes plus you can notify an expert directly), but questions really should be asked as questions. Do we let a question-in-comments run its course and then wait for someone to repost it properly? This has its own cost in that information that's germane to the original question gets more buried and hard to find.

Measuring success. Our primary metric will be an increase in constructive engagement within comments. We will also closely monitor comment flags and deletion rates to understand the impact on content quality and moderation load.

This feels like a nice Italian restaurant saying “We’re trialling some fast food items. Our metrics will be the sales of beer, burgers, and chips; we’ll also monitor customer feedback on those.” You need to also monitor how it affects sales of other dishes — do existing customers switch to the new cheaper items, or get put off entirely by the change of vibe? Or does the fast food bring in new customers, who then trade up to the upscale menu? Does the change support your existing core strength, or undermine it?

A more flexible comment section might be great, but not if it draws attention away from the main Q&A. (That was the stated goal of keeping comments minimal in the first place.) Please monitor the effect on overall site quality, not just on comments themselves!

I see the upside of what you're trying to embrace, but I also see the downside and wonder if it's really worth it

I'm concerned that this attempt to become something more like Reddit will undermine the intrinsic value of the service. The Tour tells us,

The guidance on all our Q&A sites explicitly disallows meta-discussion, to reduce clutter and noise. There's no chit-chat; just questions and answers.

...

This site is all about getting answers. It's not a discussion forum. There's no chit-chat.¹

And yet your prototype presentation appears to step more than halfway to dispensing with those expectations. Why not, then, simply do away with the Q&A format altogether and shift to a fully discussion-based solution as is used on Reddit? Because that's what it appears we're becoming. (Assuming that AI-generated searches don't make SE obsolete anyway. But we'll set that aside.)

What Stack Exchange has been in the past is a place where a specific and reasonably objective question can be asked with a best-practice or most-acceptable answer rising to the top. Secondary questions were asked anew with a link pointing to the answer that spawned it, preserving the basic integrity of the service and its information structure.

That will no longer be true.

And while the honest at heart will find it easier to explore clarifications and variations, too many people (IMO) will use it for disagreement, tangents, even misinformation — which can only be minimized through an increased moderator burden, unless you allow users to down-vote comments that the community believes are unhelpful, unrelated or misinformed.

Finally, I believe this will eventually create a powerfully negative experience for future users who can no longer find tightly focused questions and solutions relating to their problem but must now navigate pages of comment chains just to determine if a variation exists that's closer to the exact problem the future user is trying to solve. It's already difficult to convince users to edit their questions and answers with clarifications rather than leaving them in comments where they're difficult to find or easily overlooked. You're about to motivate them, possibly even reward them, for doing exactly that.

_{¹ If you do go through with this, I hope you take the time to update the Tour and Help Center to reflect the changes. Your track record in that regard has been poor and that causes its own problems as people refer to outdated material in support of actions no longer considered appropriate.}

UPDATE: Stack Exchange has decided to not remove the "No Longer Needed" flag and instead rename it to "Not Needed" (which might be a good idea to keep as the name even outside this experiment). My original answer is preserved below.

Please don't remove the "No Longer Needed" flag. Plenty of comments will still have no place in the comment section, even with this experiment. "Thank you so much" is still NLN. As is a comment asking for a change to a post which has now been made.

I like this, but I personally feel it is highly situational, that is, it's only particularly useful on certain Stack Exchange network sites which include topics that are inherently a bit looser and more conversational in nature.

I'd strongly suggest viewing this as a feature you:

1. Only turn on for specific Stack Exchange sites where that commenting format is a better fit.

2. Have strict configurable limits on the feature (say, only ONE level of comment reply nesting, there's a few other limits I can think of that'd make sense.) Let the site's moderators and the site's meta community decide how best to use the feature on THEIR site.

The experiment is focused on the interaction needs of askers, not on the needs of readers coming to this site half a year later with a similar problem.

The key value proposition of SO is that when I have a technical problem, I can find an existing question about it with multiple answers that are maintained over time, with updates and so on. If instead I get my problem in 4 different comment threads that are unmaintained, this site becomes much less useful in solving the problem for me that other people already had. Follow-up questions are still questions and deserve to be treated as questions, not comments. Follow-up questions need to be visible to all as unanswered, need to be tagged, need a title, linked questions...

So this experiment ought to measure new flags on comments like "this should be a top-level question/answer".

feature-request for the UI/Design

I don't see a flag icon in the example layout image above, so I can only assume it is hidden behind the ellipsis. Please don't add extra clicks for the number 1 (or maybe number 2) most used action on a comment. There is plenty of space as shown below for a flag icon:

Unless clicking the ellipsis directly/immediately expands to show the current flag options (including the textbox for custom/mod flags), this seems like a clear (and completely unnecessary) regression in UI/UX.

There's a lot of things wrong with this proposal, as detailled by many of the other answers, but I'll focus on one part which was already a problem with various past experiments:

Our primary metric will be an increase in constructive engagement within comments

That sounds highly problematic. You do not define "constructive" engagement, which means this is open to interpretation, and of course as SE is the one proposing this experiment, SE will be biased towards seeing it succeed. Given past experiences, I can already see you counting all non-flagged comments as "constructive" even if they might be utter garbage and nobody got around to flagging it, and/or interpreting an increase in overall comments as a good thing even though that might just represent an increase in noise.

One thing I am certain of is that SE will not commit the manpower to manually inspect all comments or even a significant percentage of comments created during the experiment, which means it will be impossible for you to analyse the results in-depth to see if they actually represent positive, valuable engagement, or if they are noise, useless, or harmful (and you might be thinking of using AI to perform a classification, which would make this worse because now you're adding a BS generator into the mix). There are not many SE employees who are active SO contributers either, so your internal feedback from people who are directly affected by the change will be limited as well. All of which means that my trust in your analysis is effectively zero.

Of course, given the volume of comments, it's not realistic to expect you to be able to analyze all of them in-depth. But that doesn't make it better to use statistics that will at best be flawed to decide the success of the experiment. Instead, why not run the experiment, and at the end of it share as much data about it as possible and gather another round of feedback from the community to gauge which parts of the experiment are well received and which parts are problematic, and make permanent changes (or can them) based on that feedback? That would be a significantly broader evaluation of the changes and their impact, give the community a meaningful way to participate, and reduce animosity against unilateral decisions by SE.

So – this feels a little like trying to shoehorn aspects of discussions into comments. Some of these feel like old feature requests – threading and formatting might actually be useful.

On the other hand – The UI at present feels a little too much like a forum, especially with the comments uncollapsed. I wonder if some visual/site distinction would be useful. I also feel like practically this would be more useful in the process of refining a question, than an answer.

This experiment represents a significant shift from the long-standing philosophy that comments are ephemeral and intended solely for post improvement or clarification. We recognize that this change may be unpopular to some and welcomed by others.

Which leads to a very important question – one of the goals of SE's design was to lower noise. The core information should be in a question or answer. I'm not entirely sure how changing the ephemerality of comments helps with creating a clear, referable knowledge base. I get the purpose and friendliness of a more forum like metaphor. I'm not sure it plays to our strengths.

Need for Follow-Up Questions

A key takeaway from our previous experiment was that users want to ask specific, technical follow-up questions related to existing questions and answers. Our analysis showed common needs such as asking direct follow-ups, clarifying answers, sharing variations, or explaining why an answer didn't work for them (and why). These attempts often included code blocks or images, which current comments do not support.

I think its worth distinguishing these follow ups in terms of 'immediacy' – I'm quite happy to work with another poster on an answer I'm writing on now. If you asked me about some obscure answer I posted 15 years ago… I'm not going to remember. A useful goal would be to optimise for people working through the answer, but not delayed/unrelated follow ups.

Additionally, we will modify the flagging modal to disable the “no longer needed” flag type for the duration of the experiment. If users attempt to use it, they will be directed to the help center article on commenting, which will have an updated copy for the duration of the experiment.

So, I'm not sure about SO, but I have a handful of folks who flag very old comments we have missed as NLN – they are NLN and best gone. SO might handle things differently but would this be for 'new' test posts or all posts.

And on that note how are pre-test comments handled, and how would test period comments be handled should the old behaviour be reverted?

Do you have a plan yet for the situation where a comment is deleted that has replies?

Currently in discussions... this results in orphaned replies. I think in the case of threaded comments there's no reason not to just take the extra step and also automatically hide/delete all of the replies to the deleted comment.

The first thing that kinda looks odd to me is the upvote button. Sure, comments today have upvotes, but in this larger, more spacious format it almost looks out of place, or lacking a clear opposite action.

I'm not suggesting we should have downvotes for comments, rather, maybe we should use an icon that more accurately fits what upvotes on comments are, since upvoting on answers and questions are expressively different from upvoting comments.

feature-request Allow users to access their deleted comments, even more so since in your proposal, comments can be even more substantial.

My prediction is that this change will increase "engagement", just not from people you'd necessarily wish to engage with. Eventually the advertisers will wisen up to this.

The whole point, the entire differentiator of the Stack Exchange network is to offer on-topic Q&A content. Adding what amounts to a Reddit or Slashdot thread to every question and answer seems to be counter-productive.

Follow-up questions should be asked separately, as their own question. That's the case everywhere on Stack Exchange. This experiment attempts to change it - maybe? You're not clear on that. Does "follow-up" mean questions by the original questioner, or questions related to the question that ask for clarification? This is a technical discussion of a feature, so unambiguous language does matter!

On SE, an "ideal" question and answer should have zero comments. The purpose of the comments is, in essence, to point out various shortcomings of the content being commented on. Those shortcomings may be related to clarity, on-topic-ness, correctness of answers, etc. The ideal lifetime of these comments should tend to zero.

Since this cardinal premise will be broken, What will be the differentiator between an answer and a comment, or between a comment and a separate question? Do we need "me too" comments, "thanks", and such? What comments do you really wish to see? "Engaging comments" is meaningless in this context. That's a wish, not a solution, and you seem to offer a solution for evaluation - where, seemingly, there is not much to evaluate.

I would expect, at minimum, not to see "Lorem Ipsum" in your comment examples. You should give several examples of desired, "real life" comment chains that you'd like to see under questions and under answers. Cover as many use cases as you have.

The use of "Lorem Ipsum" tells me that you have no use cases, and the whole idea is vacuous.

Gavin's answer raises a great point:

Comments are deliberately less fully-featured than questions/answers. Not because nobody bothered to add proper markdown and threading support, but rather to encourage people to use questions and answers instead.

as well as this ensuing comment reply by tkruse:

The negative impact on Answers should be measured. If a comment can replace an answer because it can have code, answers may degrade in quantity. – tkruse Commented 2025-04-30 05:09:30Z

If you are adding code block and other formatting support to comments... you probably need to consider giving moderators the ability to convert comments into answers, and/or give certain users the ability to (auto) 'flag' or vote for a comment to be converted into an answer¹, at least for the duration of the experiment.

Currently moderators can convert answers into comments, but not vice versa. Yet, we have a real problem (and have for many years) of people answering in comments (something they shouldn't be doing and something comments are not meant to do). If users can now give all the same formatting capabilities to their comments as they can to their answers, it stands to reason that they're going to be posting answers as comments more often, not less.

The consequence of this is that, while you may see more comments with this experiment/new feature set, you'll likely also be shooting yourself in the foot regarding the answer rate on the network, which has been declining for several years, if I recall correctly.

I'm not sure whether it would make sense to have this restricted to just comments under an answer. Any such conversion would also necessarily raise a notification to the comment/answer author.

^{1: Similarly to how a certain number of No Longer Needed flags from users will automatically delete a comment, if a comment gets a "this looks like an attempt to answer" flags, it should be handled by a moderator. Alternatively, if some number of such flags are raised on a comment (say, 5), then it should automatically convert to an answer, and all child comment replies under it should be converted to comments under the answer.}

   ^{To avoid abuse, this flag could be restricted to the first 90 days (or 365 days or whatever) after a comment is posted, just like how we can't flag/vote to migrate a question after 60 days). You can restrict it further if desired by exposing the flag only to users with a certain reputation threshold (say, >=1,000).}

It's problematic that the "Add a comment" box is at the very top, above the other comment threads (pictured below). This is encouraging people to talk without reading what has already been said, which dramatically increases the number of duplicate comments and the amount of noise.

To write this answer, I scrolled down all the way to the bottom of the page, checking all the other answers (but not their associated comments!) on the way, and then hit "OK" on a button reminding me that with >30 answers already posted here, I should make sure that I read them before adding mine. If you want to encourage follow-up questions in comments (which my other answer, in addition to several other answers here argue is not a good idea), then they should be held to the same basic standard as other posts (both questions and answers): listen to what other people have said before you speak.

This concern also applies to the "Reply" button, though an improved UI there is somewhat harder to conceive.

_{I understand how ironic it would be if some other answer here already mentions this, considering how obvious a flaw it is, but I doubt that such an answer could have escaped my notice.}

Additionally, we will modify the flagging modal to disable the “no longer needed” flag type for the duration of the experiment.

Then for the duration of the experiment, I will be raising "Unfriendly or Unkind" flags on any comment that should be removed like begging for acceptance/votes, discussing votes, "thanks" comments, "I have this problem too" comments, gibberish, anything that was already addressed as an edit to the post, etc.

I can only hope I will not be flag banned for this behaviour.

^{_{Ironic Note: I had to leave this answer (feedback) for a few hours, and come back to finish it. Rather than posting a comment on someone else's answer and exploring how my ideas are a little different I've done what we're supposed to do; write my own answer. Not a duplicate answer, but one bearing some similarities to a few others.}}

I disagree with this part of the announcement, it lists rules to create more noise on our completed work, long finished and forgotten. It invites "me too" comments with a provisio that entitles them to pollute your comment free post that has long satisfied almost everyone for a long time, except them, now.

"What are the new acceptable commenting rules?"

"To better support an environment where discussions can take place on Q&A pairs, we are experimenting with these updated commenting guidelines on Stack Overflow for the duration of the experiment:

• Asking specific follow-up questions about the post.
• Seeking clarification on how an answer works or why it might not work for you.
• Sharing variations or related experiences pertinent to the Q&A.
• Engaging in constructive, technical discussion sparked by a question or answer, even if it explores associated concepts.

These guidelines will allow for more interaction than the traditional methodology of only using comments to improve a post or request a clarification of its intent. We want to build a space for technical exchanges.".

They should ask a new question, they can provide an attribution link to the original, and quote the specific point(s) they want help with; I'd be willing to have a setting to disable notifications when that occurs; leaving my post followup free, and excessive notification reduced.

The above reason: "Sharing variations or related experiences pertinent to the Q&A." could be OK (even useful to many) if the person simply said: "There is a similar and not duplicate question here: [provides link]".

Benefits of not making our answers (and unanswered questions) a discussion forum:

• More Q&As, which many sites need; instead of a comment hidden away somewhere.
• Rep for the asker, instead of being a commenter and earning nothing for a valid point.
• Maybe a bit of pension rep for the OP, instead of a series of pings as a discussion opens up.
• A tab for the "Newest Comments" (à la the "Newest" question tab) would be needed (and extremely active, unusable) if everyone is to be aware of the new comment; otherwise if the OP simply ignores it and then it likely gets forgotten and remains undiscovered (not replied to by others).

People on Stack Overflow might like this idea, and appreciate the opportunity to engage in this experience (I don't go there enough to speak for that site), but I don't know anywhere else that would like this; except possibly our Photography site, where after the discussion (sometimes very lengthy) everyone comes back and deletes their own comments (leaving a polished post without a lot of comments between every valuable answer).

Lowering the reputation limit

You mentioned that this will lead to more spam and I think this will also increase the amount of other comments that should be removed for whatever reason. How are you considering the moderation aspect of that? Are there plans to add comment mods (which got 204 upvotes and 6 downvotes), show flagged comments to 10k users (+73/-2) or something similar?

Please at least think about ways to handle the increase of unnecessary comments (including the gap left by disabling NLN but there are already sufficiently many answers on that) and tell us what you are doing.

Should there be an explicit message to posters and commenters that the platform is still intended to be a Q&A -- not a Q&A&C?

Since I assume curators won't be able to edit other users' comment threads (only flagging), then there should be explicit guidance that ideally the commentary should be used to improve the original post (question or answer) AND the comments should be deleted by the commenters themselves after the post has been improved.

We want all of the "signal" in one coherent place with an absolute minimum of "noise". Scrolling through and expanding rambling threads is not the goal of Stack Overflow.

Tangentially, I would like it if "comments that are on-topic, professional, and beneficial to users" were used to shield curators from -1 rep penalties when downvoting posts. I would be doing a lot more vote-based curating if I wasn't penalized for it. Currently I just comment saying how the answer doesn't attempt to answer the question or provide an online demo of how the answer is provably wrong for the asked question, but don't bother to downvote because the post from 2013 already has a score of +43. Food for thought.

This should include improving the notifications, the follow feature, and the global inbox.

Below is a non-exhaustive list of options

• Ignore all comment notifications, but keep edit notifications and on your own questions about getting new answers.
• Ignore comments notifications on your specific question / answer.
• Ignore comments notifications on a specific question
• Ignore comments notifications on a specific answer
• Ignore comments notifications from new users
• Ignore comments notifications from a specific user

If, in the early stages of this experiment, notifications are not part of the heavy work, consider including at least an option to receive all or none of the notifications triggered by the new content types.

Related answers about notifications

• MisterMiyagi
• Rob

feature-request for the UI/Design

Please add some kind of border or styling to the comment "vote"/"reaction" and Reply buttons. It's difficult to tell now which one the button is replying to, even with the Reddit-style indentation lines. In fact it looks like this comment design is basically just pulled from the modern Reddit design, which is a very bad thing:

Reddit's old comment styles were much cleaner, and far better:

GitHub's design is also much better than the sample that is shown in the announcement above, and somewhere in the middle (a happy medium, one might say) between the sample above/Reddit's modern layout and SO's current comments/Reddit's old layout:

While GitHub's comments aren't threaded, there is a border around the entire comment and any reactions are shown within that border. SE comments being threaded here shouldn't impact that design much/at all.

Please also make the reaction/reply buttons smaller. Why are they equal to the height of two comment lines? They should be the height of one line of text or maybe a bit more.

Another answer makes an important schematic of how much unimportant information spreads across the webpage and hides the other direct answers to the original question.

I wish to add that if there were indeed important information in the now-encouraged follow-up questions in the comments, that would be just as bad. I land at Stack Exchange pages when searching Google for information. If the answer to my Google'd question were in the comments, I would not find it, since I would not dig through those conversations. That's why I never open Reddit links that show up on Google searches.

If our goal here is to build a library of information that is helpful not just to the person who asked it, but to people of the future, then this change is counterproductive. The present policy of encouraging users to go ask a new question is still the best way to go.

I think this experiment would be useful; but I think it should be a separate kind of thing (e.g. associated discussion – or, as Rubén points out, chat). Comments are still needed for their original purpose, as temporary post-it notes.

Removing the "no longer needed" flag suggests you're trying to introduce a radical new thing, which is fine, but beware Chesterton's fence.