12

Today, I shared an update on progress from last quarter and what we’re up to this quarter on the blog. Community Products development teams are focused on delivering against outcomes, which are organized around strategic initiatives. With this approach, teams may test many different solutions, with the goal of driving toward their desired outcome. Below, I’ve outlined the strategic initiatives, goals (outcomes), and some example solutions you may see the teams testing. We’ll be linking to future posts from this list as they kick off, so you can find everything we’re working on in one place. My intent is to create more visibility into how we’re approaching product strategy and prioritization of work on the platform.

Strategic focus > Desired Outcomes > Example Solutions:

  • Growth through knowledge creation

    • Outcome: Increase content & contributions that foster a sense of community, promote collaboration and problem solving, or promote high quality, trusted content. Examples (but not limited to):

  • Supporting moderation at scale

    • Outcomes: Reduce the burden of moderation across all users and content types

    • Examples:

      • Leverage machine learning to reduce spam and streamline existing processes/systems

      • Improve existing moderator tools such as better anti-vandalism rate limiting and prevention tools.

  • Supporting community engagement

    • Outcomes: Increase activated and engaged users

    • Examples:

      • Improving and/or introducing new game mechanics

      • Ways to help new users get started, such as an updated tour, updated new user emails, etc.

      • Community Asks Sprint to address more community requests

  • Personalization and discovery

    • Outcomes: Increase engagement through personalized & relevant content recommendations

    • Examples:

      • Recommendations that match users to the right content
      • Improved content discovery through better search and summarization

We are looking forward to your feedback here and/or on the various posts that will go out over the next few months as we make progress against the goals above. If you’d like to participate in user research (interview, unmoderated tests, surveys, etc), we invite you to opt into our user research list by toggling on Research Invitations in your settings. We'll be keeping an eye on this post through May 6th and will do our best to respond, where appropriate, during that time.

Improve this question
8
  • 4
    Thanks for sharing this update, Des! Is there a specific date that you'll be tracking feedback on this post until, like with past updates of this kind?
    – V2Blast
    CommentedApr 23 at 18:29
  • 2
    @V2Blast We'll be actively monitoring the post for two weeks as we typically do. I've updated the post copy to reflect that.
    – RosieStaffMod
    CommentedApr 23 at 19:07
  • Is a functioning staging ground a desirable outcome? Because lately I heard there is a leakage somewhere and the supply ran dry but no plumber in sight who could connect everything again.
    – NoDataDumpNoContribution
    CommentedApr 23 at 20:01
  • 3
    @Rosie Out of curiosity, why do you only monitor the post for 2 weeks. If someone comes a bit later than that with useful feedback, don’t you want to know about that?
    – Starship Remembers Shadow
    CommentedApr 23 at 23:49
  • 2
    @StarshipRemembersShadow That's worth a question on its own I guess - but Its possible its an arbitrary cutoff that makes sense in the context of SE's software development lifecycle
    – Journeyman Geek
    CommentedApr 24 at 8:22
  • 3
    @StarshipRemembersShadow We still get notifications when people provide more feedback after the two weeks, and it is taken into consideration. We do often respond to later comments and questions, too. But with most posts, we're "actively" monitoring and responding this first week or two because that is when we receive the most feedback.
    – RosieStaffMod
    CommentedApr 24 at 13:04
  • "Outcomes: Increase engagement through personalized & relevant content recommendations" This seems like putting the cart before the horse. The desired outcome is simply "increased engagement." Personalization and more relevant content recommendations are two possible approaches to increasing engagement. (Personally, I'm adamantly opposed to more personalization. It's creepy, off-putting, and never really good enough to be helpful. Relevancy is far more dependent on context than it is on the individual.)
    – Adrian McCarthy
    CommentedApr 24 at 18:23
  • 1
    The fact that you said "Leverage machine learning to reduce spam and streamline existing processes/systems" instead of "Leveraging AI ..." tells me that someone with a reasonable amount of experience actually put thought into that. I'm interested in hearing more.
    – dan1st
    Commented2 days ago

6 Answers 6

Reset to default
14

I think it's great to see a vision communicated ("Increase content & contributions that foster a sense of community, promote collaboration and problem solving, or promote high quality, trusted content.") Finally. I've been wishing for the company to articulate its vision for the site for a while. Great to see it!

Those are ambitious goals. It could be exciting if you can deliver on them.

I think the implementation efforts have been poorly matched to the vision. Up until now, Stack Exchange's format has been designed to "promote high quality, trusted content". It is not designed to promote community or open-ended collaboration (that is an explicit anti-goal of the existing Q&A design). It's going to take a lot to build a site that can provide value in "fostering a sense of community" or on "collabration and problem solving", as those are deeply different from what we've got.

I'm concerned that, when it comes to actually doing something, the steps that are being taken are to make minor tweaks to the existing platform (e.g., minor tweaks to Discussions, or to how comments work). That strikes me as a deep, deep mismatch between the breadth and scale of the ambition ("foster a sense of community" or "promote collaboration") and the degree of commitment and investment to build such a thing. Small tweaks to Discussion or to comments are not going to achieve those ambitious goals.

There are many competitors that already serve some of those other goals, and have put in hundreds of engineering person-years into building websites to support those goals. Facebook, Discourse, Twitter, and Reddit already have a lot that's aimed at fostering a sense of community and promoting collaboration. To make a difference in that area, you're going to need something that can go beyond what's already available. I don't think Discussions or comments are going to get you there or are the right starting point. I encourage you to re-think from the start how to approach those ambitious goals (what are the problems with existing sites? what's the unique new value that Stack Exchange can add?), and come up with something fundamentally new.

I also encourage you to avoid under estimating the level of investment needed to achieve that vision. So far, it seems like Stack Exchange has very limited resources in time from software developers to work on the public platform. We see occasional sprints that do a bunch of tiny changes. That's not going to get you to achieving that vision. If you want to have an ambitious vision, you need significant software development resources. I encourage you to take a serious look at whether you truly have the software development resources available to you that will be needed to execute on those goals.

Improve this answer
6
  • 1
    analogous(?): langdev.stackexchange.com/q/656/251
    – starball
    CommentedApr 24 at 5:51
  • 3
    Thank you for the thoughtful response. I am curious—how do you feel that Facebook, Discourse, Twitter, and Reddit are fostering community and promoting collaboration?
    – PiperStaff
    CommentedApr 24 at 14:22
  • 2
    @Piper their design, technical infrastructure, contribution rules, and social mass / momentum. the last one is significant. the same thing happens with software technology choices. even if there's new, better technology out there (or just alternatives with general parity and different tradeoffs), the benefit of choosing the well-established things is the network effect (Ex. CMake, or frontend JS libraries/frameworks). same goes with social platforms.
    – starball
    CommentedApr 24 at 18:47
  • 1
    @Piper, Thank you for listening! I have no insights about what they do well towards that mission. Sorry. I suspect you are far more knowledgeable about that subject than I am.
    – D.W.
    CommentedApr 24 at 18:57
  • @DW I’m also excited and inspired by this vision. Primarily because I believe it’s a shared vision between community members and the company. To say these goals are ambitious is an understatement. They are, but they are also achievable. Will we execute on everything this quarter? No, of course not, but we are hitting the ground running. Will we pivot on some of the implementation based on feedback in the discovery and experiment phases? Of course, we will. But that’s all part of good product development. (1/2)
    – RosieStaffMod
    CommentedApr 24 at 21:11
  • 2
    I hope people take advantage of Des’s call to opt into Research Invitations. Building this together is what will make this vision successful. (2/2)
    – RosieStaffMod
    CommentedApr 24 at 21:12
10

Chat: investing in discoverability and usability

WOO!

Expanding comments to support technical discussions

That... is kind of a departure from what comments on Q&A are currently canonically for. The ideal is that all the good content is in the post. The whole "Q&A without the noise and chit-chat" thing. I'm curious what this will look like, and how consistent it is with Prashanth's direction of keeping Q&A the good thing that it is and doing new things through other "lanes" (Recapping Stack’s first community-wide AMA (Ask Me Anything)).

Stack snippets beyond frontend tags and as a stand alone library

Interested to hear more about this and how you plan to implement and maintain this. There's a lot of good stuff out there like godbolt.com and the TypeScript playground. If there are ways to partner and get guarantees about permanence of content, or somehow own data but use their frontend, or maybe self-host those, I think that would be great. Lots of work, but there'd probably be more work to do it from scratch.

Ex. Leverage machine learning to reduce spam and streamline existing processes/systems

See also How could AI be used to augment curation tooling?

Improving and/or introducing new game mechanics

Please don't forget consider with some level of seriousness existing feature requests for badges on the various network meta sites. Ex. on MSE and on MSO.

Ways to help new users get started, such as an updated tour, updated new user emails, etc.

Good idea (at least in principle. I'll hold my tongue for when design and implementation comes out). I'm sure the community has ideas and experience to share about how they got started and what was helpful, and would love to share. Ex. how to find questions to answer, how to get involved in curation, and how to learn site mechanics. Ex. How did you develop your understanding of the Stack Exchange network?

Recommendations that match users to the right content

Curious what "right" means. Defining things when it comes to connecting users and content doesn't seem to be super strong historically... See For the purposes of the "Related" section of the Q&A UI, what definition of "related" aligns closely to what users of Stack Overflow find value in?, These are not interesting posts for me. I'll throw in a link to a suggestion I had before that I still think is good: https://meta.stackoverflow.com/a/430810/11107541.

Improve this answer
2
  • With regard to snippets: note that godbolt.com supports both short & full links. Short links use their own database, so there's probably a time limit, however full links simply provide the code (and settings?) as a HTTP parameter (a blob). For the purpose of SO, it seems full links would be better: SO would retain the snippet on their side, and simply encode it dynamically when generating the link. As a bonus, should the encoding change on godbolt.com, the method of encoding can be adapted on SO to match.
    – Matthieu M.
    CommentedApr 24 at 7:22
  • @Starball All great points and I’m glad that you are excited about a lot of these initiatives. In regards to Snippits, more is coming on that in the upcoming weeks, and we’ll link to that here when the team posts about it.
    – RosieStaffMod
    CommentedApr 24 at 21:06
8

Expanding comments to support technical discussions

This has me a little curious. We've always treated comments as temporary. What would these technical discussions look like, and what would be the proposed technical, design and cultural changes be?

Supporting moderation at scale

  • Leverage machine learning to reduce spam and streamline existing processes/systems

I suspect as one of the sites hit worse with spam, it'd be noticed and appreciated.

Outcomes: Increase activated and engaged users

While it's tempting to look primarily at gamification and design, it's worth looking at causes of friction with the community, and addressing those sources of unhappiness. These are social and maybe political issues as much as technical and design issues.

It's at a different level from what's normally looked at, but positive staff engagement - and nurturing and creating the next Yaakovs, Odeds, and Shogs, folks who're comfortable with the community and can both communicate and identify problems more naturally/organically would be nice. The planned community hires for the community team's a good idea, and there's a few current folks who're emerging as a result of current dev work on chat but accepting this at a formal level would be a good start towards this.

It's also a people/optics problem - so it's worth considering how best to rebuild trust, and banking goodwill with the community.

I'd think a 'great' end result would to bring folks who still care about the network, but are elsewhere 'home', but even as a resident optimist, it's going to be a long, hard, and frustrating road (even if it'd be worth it, if we do it). To that end - well, can we trust that the company's willing to take the time, effort, and concessions it'll take and keep this in focus over time?

Improve this answer
    7

    In your "Supporting community engagement" section, there's an example of

    Improving and/or introducing new game mechanics

    I believe the main current aspects of this site which may possibly be considered to be "game mechanics" are reputation, privileges and badges. Are these the sorts of things that you're referring to?

    I realize your post is just an overall summary, and many aspects are still only in the early stages of investigation and development, but I (and likely many others) would appreciate some more explanation about what sorts of things you're specifically considering possibly doing.

    Improve this answer
    6
    • 1
      In terms of possible improvements, perhaps it's time to stop tying everything with reputation. Reputation is a single score, roughly based on expertise in specific content. Tying editing rights or moderation rights to reputation has always been somewhat iffy...
      – Matthieu M.
      CommentedApr 24 at 7:24
    • 2
      @johnomielan, engagement and game mechanics does encompass some of the areas you mention, but we're also exploring possibilities beyond that too. On MSO for example, we just posted about some early discovery work on a potential gamified coding challenge and we'd love community feedback on it.meta.stackoverflow.com/questions/433667/…
      – RosieStaffMod
      CommentedApr 24 at 14:34
    • 1
      I know that was a bit vague - the surface area is broad and there are many possibilities. One area that is actively being looked at is how to help users demonstrate their work in a more impactful way. For example, improving profiles so that they can be shared externally. Another idea is how we might connect contributions more directly to job opportunities (posted today). Beyond this - we know the rep system might work for Q&A but doesn’t necessarily support other content types. We need to consider how rewards and recognition could/should differ for different modes of participation.
      – DesStaffMod
      CommentedApr 24 at 21:02
    • 1
      @Des what are you referring to when you say "posted today"?
      – starball
      CommentedApr 25 at 2:25
    • 1
      @starball Apologies. This one: meta.stackoverflow.com/questions/433667/…
      – DesStaffMod
      Commented2 days ago
    • @Des I read that post yesterday. ctrl+f for "career" and "job" in there doesn't turn up anything like you alluded to. ?
      – starball
      Commented2 days ago
    5

    In respect of Leverage machine learning to reduce spam and streamline existing processes/systems, the Ask Ubuntu moderator andrew.46 made the following suggestions in AI enhancement of Ask Ubuntu:

    I am convinced that AI has a place on Ask Ubuntu but not in content generation. Some uses that I see as a positive use behind the scenes for AI would be:

    • Generating a warning if the Code of Conduct is or has been violated
    • Following on from this: detection of abusive / racist language
    • Create a better search. I don't know about you but I use Google rather than the SE search
    • Moderator assistance with the detection of sock puppets who are violating the rules
    • Better detection of duplicate questions on the fly

    While the above points originated on the Ask Ubuntu meta site, I think they could be useful on all sites. Therefore, are posting here as Community wiki so the suggestions from andrew-46 may be voted on.

    Based upon manually flagging on some sites such as Super User which have been getting a lot of SPAM recently, it would be useful to automatically detect the following patterns used by spammers:

    1. They initially copy the title and (the initial part) of an existing valid question before subsequently editing their post to contain SPAM. This could be helped by the better detection of duplicate questions on the fly.
    2. They post gibberish before subsequently editing their post to contain SPAM. Not sure how easy it is for AI to detect gibberish.
    Improve this answer
      -4

      Growth through knowledge creation

      I find it interesting that, aside from an editor improvement (snippet), none of the examples given for growth through knowledge creation involve Q&A. It seems like you're moving away from Q&A because AI has become quite good at it, as evidenced by the tenfold drop in posts over the past few years, and instead shifting toward social (chat) and subjective (discussion) content to avoid that competition.

      Improve this answer
      14
      • 10
        Careful saying that AI is good at Q&A. What ChatGPT does and what SE does are very different to someone with even a marginal interest in a real answer. For the record, ChatGPT is incapable of even correctly counting how many characters are in a string.
        – controlgroup
        CommentedApr 23 at 21:35
      • @controlgroup they fixed the strawberry
        – Franck Dernoncourt
        CommentedApr 23 at 21:37
      • 9
        It fails on arbitrary strings, and can also be convinced of absurdities if you speak in the right tone of voice. It only sounds like an intelligent person because that’s what OpenAI trained it to do; there is no intelligence in a pile of linear algebra and a tokenizer.
        – controlgroup
        CommentedApr 23 at 21:38
      • 1
        @controlgroup But maybe there is also no intelligence in a pile of organic material like me. This is too philosophic. The point of this answer is that the company moves away from Q&A. Franck thinks it might have to do with increased competition. I think it's because Q&A is already well supported. There is not much to develop be there anymore. The new features are additional, not replacing the old ones. Any why not.
        – NoDataDumpNoContribution
        CommentedApr 23 at 21:44
      • 7
        I remain to be convinced that ChatGPT can be trusted. There was one example question I was involved in recently where circuit design errors were being pointed out, which when comments from OP revealed that ChatGPT had been used for the design the question was promptly closed (I voted to close). This comment summaries the issue with ChatGPT : It constantly gets things wrong, and often in subtle ways that are hard to notice until it just doesn't work.
        – Chester Gillon
        CommentedApr 23 at 21:46
      • 7
        ^^ This. If you say “couldn’t this be true?” in the right tone, it’ll say “yes it could!” and spout obviously-false nonsense. We deal with the consequences of that a lot on Physics SE and others.
        – controlgroup
        CommentedApr 23 at 21:48
      • 1
        @controlgroup AI definitely fail on some questions e.g. arxiv.org/pdf/2501.14249 but it is becoming increasingly good at answering the type of questions asked on SE, at least on many SE sites including SO.
        – Franck Dernoncourt
        CommentedApr 23 at 22:17
      • 1
        that is not the reason provided. meta.stackexchange.com/q/407025/997587
        – starball
        CommentedApr 23 at 22:20
      • 1
        Would you like to have a look at Physics SE?
        – Vincent Thacker
        CommentedApr 23 at 22:49
      • @VincentThacker sorry I'm unfamiliar with that SE. arxiv.org/pdf/2501.14249 has a benchmark on physics questions: i.sstatic.net/0Nj6KfCY.png
        – Franck Dernoncourt
        CommentedApr 24 at 0:21
      • 6
        @FranckDernoncourt, forget "Humanity's Last Exam". I asked it how to defeat the hotdog in Sandcastle Builder (the sort of question that might be asked on Arqade). It gave an answer that uses some of the game's idiosyncratic terminology, but is otherwise complete nonsense. (Correct answer: there is no hotdog; all references to it are an inside joke by the game's developer. I'd also accept either an "I don't know" or an answer that extends the joke.)
        – Mark
        CommentedApr 24 at 1:18
      • @Mark 84 github stars, sub with <500 followers on Reddit, no wikipedia page... not a well-known game. Most questions on SE aren't that arcane.
        – Franck Dernoncourt
        CommentedApr 24 at 1:22
      • @FranckDernoncourt, that doesn't change the fact that ChatGPT gave a very confident and very wrong answer to a question of the sort that would be asked on SE.
        – Mark
        CommentedApr 24 at 21:21
      • @Mark ChatGPT gave me a decent answer: i.sstatic.net/GPKXavJQ.png
        – Franck Dernoncourt
        CommentedApr 24 at 21:25

      You must log in to answer this question.

      Start asking to get answers

      Find the answer to your question by asking.

      Ask question

      Explore related questions

      See similar questions with these tags.