74

My reading of thousands of posts here, from over most of the time period going back to shortly after SO started, indicates the company has often not done a particularly good job of communicating with the community. Nonetheless, it seems that, over the past few months, the company has been particularly bad, even rather unnecessarily secretive, regarding communicating about issues that are quite noticeable and/or important to the community. Two important examples are:

  1. Stopping the quarterly data dump. This was asked about at June 2023 Data Dump is missing a couple of days ago, with it first getting an answer by AMtwo, who was just recently let go by SE. This answer states that

    The job that uploads the data dump to Archive.org was disabled on 28 March, and marked to not be re-enabled without approval of senior leadership.

    Note this disabling occurred about 2 1/2 months ago. The question then got an answer from the CTO, i.e., Jody Bailey, with it saying

    Stack Overflow senior leadership is working on a strategy to protect Stack Overflow data from being misused by companies building LLMs. While working on this strategy, we decided to stop the dump until we could put guardrails in place.

    and

    We are looking for ways to gate access to the Dump, APIs, and SEDE, that will allow individuals access to the data while preventing misuse by organizations looking to profit from the work of our community. ... We will provide regular updates on our progress to this group.

    I find the last sentence, i.e., "We will provide regular updates on our progress to this group", rather ironic as they didn't even announce doing this until well after they decided to implement it, i.e., on or before March 28.

    About 4 days after Jody's answer, Philippe wrote What is the current (June 2023) status of the data dumps and the company’s commitment to them?, and an answer stating

    Our intention was never to stop posting the data dump, only to begin to collect more information on how it was being used and by whom - especially in light of the rise of LLMs and questions around how genAI models are handling attribution.

    with the emphasis being mine. As written by itself, this states there was never any intention to stop posting the data dump, but these posts were specifically stopped. Relatively shortly after, Philippe's question was merged with June 2023 Data Dump is missing, and his answer was deleted. Instead, his earlier, basically same, answer there was updated to now state

    Our intention was never to stop posting the data dump permanently, only to pause it while we begin to collect more information on how it was being used and by whom - especially in light of the rise of LLMs and questions around how genAI models are handling attribution.

    with the emphasis showing what was added. This, for me at least, resolves that particular discrepancy. Nonetheless, there are still several others ones (as well as other issues), but I won't address them here as that has already been quite ably done in the other answers there.

  2. The company basically unilaterally changed, without following proper procedures, the diamond moderator policies to largely prevent them from taking appropriate actions regarding users writing posts (and their associated posts) which basically came from AI (I assume mostly from ChatGPT). There was relatively little notice given, and parts of it were private, regarding this change in policy.

    This was discussed, such as at Stack Exchange is failing its community (although this also deals with several other, possibly related issues) before there was an official statement by Philippe at What is the network policy regarding AI Generated content? (with this then also being discussed later, e.g., at Why wasn't the guidance to moderators on how to moderate AI Generated content made publicly available?). However, this had an official post lock, and discussion wasn't supposed to occur (with answers deleted by a staff manager, but fortunately a regular member created Discussion: Network policy regarding AI Generated content for that purpose), even though it originally had the tag. Also, Philippe's answer, according to several diamond moderator statements, did not particularly accurately describe what the diamond moderators were told in private.

    In addition, the answer starts with "We recently performed a set of analyses", but it didn't give very many details about it. Instead, they weren't provided until a couple of days ago (after a moderator strike, e.g., as discussed at Moderation Strike: Stack Overflow, Inc. cannot consistently ignore, mistreat, and malign its volunteers, had already started) at GPT on the platform: Data, actions, and outcomes. It was relatively comprehensive but, as many of its answers indicate, there are quite a few weaknesses with the analysis. Regardless, this information should have been provided (at least to the diamond moderators, but preferably to the entire community) while the moderator change policy was being considered, and especially well before any such change was implemented.


I hope this apparent degradation in communications quality is temporary, with the company improving the timeliness, and level, of notifications with the SE community in the future. Since this post is primarily about the company, in particular the management, I would really appreciate them noticing this and responding, ideally to give some reassurance to us that the situation will improve.

15
  • 24
    Tbf, re specifically (and only) #2, announcing it in advance would've brought a whole lot of attention to the vulnerability while it was still live. It's not optimal, but it's one of the extremely few circumstances where a warning simply didn't make sense
    – Zoe
    Commented Jun 9, 2023 at 22:12
  • 22
    While I generally agree with this post, the Gravatar change is a bad example. They note in the post you reference that the secrecy was due to the sensitive (PII) nature of the data involved. We can debate the efficacy of that given the existence of web caches, but it's reasonable that they at least wanted to stop distributing the sensitive data via their own website before announcing the vulnerability. Announcing vulnerabilities only after remediation (assuming remediation is reasonably fast, as was the case here) is a common and well-accepted practice.
    – Ryan M
    Commented Jun 9, 2023 at 22:13
  • 1
    @ZoestandswithUkraine Thanks for the feedback. However, if their main concern was about notification while the vulnerability was still live, I wonder why didn't wait even longer since, as I stated, at least most of my icons didn't change until several days after their official announcement (and, relatedly, why wasn't the process involved given a higher priority). Nonetheless, I believe you have raised a valid and important point, so I've added a link to your comment in my answer. Commented Jun 9, 2023 at 22:30
  • @RyanM-Regenerateresponse Thank you for your detailed feedback. As I stated in my answer and in the comment above, I find the company not waiting until even later to making the announcement to be somewhat inconsistent with it being a concern about the vulnerability possibly being discovered and exploited. Nonetheless, you have made several valid and useful points, so I've also added a link to your comment in my answer. Commented Jun 9, 2023 at 22:34
  • 6
    There's more: Why hasn't there been an official community announcement or blogpost about charging for usage of subscriber content in the training of LLMs? is still unanswered. and New blog post from our CEO Prashanth: Community is the future of AI was just a dump of a CEO blogpost drenched in market-speak with no adaptation for the MSE audience.
    – starball
    Commented Jun 9, 2023 at 23:23
  • 2
    Re: #2 i.e. Gravatar Identicon, I think it's similar to decommissioning mobile app infrastructure, where they initially didn't mention the critical reason and then updated the post after they had done decommissioning it. Commented Jun 10, 2023 at 2:01
  • oh, but my dear friend, there probably was an implicit announcement that the data could no longer be released for free for everyone to grab. This one
    – SPArcheon
    Commented Jun 10, 2023 at 15:10
  • 4
    I won't argue that the company has been ignoring some of the principles it used to claim to stand for. But the three paragraphs about the gravatar change are just totally wrong. This was explained quite well here and if you think the company was doing something nefarious to keep a secret from you, rather than try to protect the PII that was inadvertently made vulnerable due to the previous values stored, you should probably focus on something else. You can nit-pick all you like but they likely revealed details due to relentless pressure here. Commented Jun 11, 2023 at 17:08
  • 2
    If I remember right, the identicon changes were rolled out site by site. Probably a lot safer to succumb to the pressure here after SO, Meta, SF, etc. were handled, and a much smaller population were still left to fix. Like you said, there was no indication that any of the data had been exploited, and they didn't exactly give out the formula for how one could do so, so tin foil hats about timing are just that. "Inform us immediately! Don't inform us immediately!" You gotta pick one and maybe settle for something in the middle. Commented Jun 11, 2023 at 17:12
  • @Stuckat1337 Thanks for the feedback, but I believe you misinterpreted my intent, although that's likely because I wasn't particularly clear. I apologize for that lack of clarity. As such, I've now updated my answer to indicate I'm not considering any kind of conspiracy type situation but, instead, just that there could be some other reasons that weren't mentioned and, if so, they are likely quite innocuous. Commented Jun 12, 2023 at 1:56
  • 1
    This behavior of the company is characteristic of the last 5 years. E.g. they dropped collectives out of the blue sky on us or the infamous thank you button. The strategy seems to be that community communication comes last and only by the strictly necessary amount, maybe as a way to avoid friction. If you don't say much, people cannot protest much. And on the occasions that they say something, typically then the ideas aren't received well, but then they need to deal with it only once. It's been like this for years (with a few exceptions). Commented Jun 12, 2023 at 5:42
  • 2
    Still sounds very much like grasping for a conspiracy theory hoping to latch onto something. Why can't it just be that they handled the big sites and were coming clean to all the people complaining here? What else is even in the realm of possibility for why they would explain the situation at that time vs. any other? A day earlier, a day later, two weeks after the process was done? What would have been the perfect delivery time & day for you to suddenly not be suspicious? This incident seems very out of place to me given the remainder of your (quite legitimate) examples. Commented Jun 12, 2023 at 11:37
  • Note: Despite what several comments imply, I have never considered there to be anything inappropriate, or even nefarious, about how the gravatar identicons update was communicated, but just that more details about things like timing could have been provided, as I explained. Nonetheless, I've now removed that item (so item #2 is now the last one) since it is such a relatively minor issue compared to the other two. Also, I wish to have the answer focus more on them, including the last one as it directly applies the current ongoing moderators (and many curators) strike. Commented Jun 13, 2023 at 19:07
  • @JohnOmielan Re: your last edit: I don't see how Philippe implied that the dumps accidentally stopped. I understood the statement as saying that it was not their intention to stop doing periodic data dumps. And this specific dump was temporarily stopped so that they could get "more information on how it was being used and by whom". Commented Jun 14, 2023 at 0:50
  • @CaveJohnson I only mentioned about "the dumps accidentally stopped" as that was what I considered to be the literal meaning of the statement, considering the dumps were stopped on purpose based on the 2 answers to the other question, with this conflicting the statement there was no intent to do. As such, I've updated my answer to try to make this more clear. Nonetheless, I still think that the statement should be corrected appropriately, such as with something like my suggestion of adding "for very long" to it. Commented Jun 14, 2023 at 0:58

1 Answer 1

41

I hope this apparent degradation in communications quality is temporary, with the company improving the timeliness, and level, of notifications with the SE community in the future.

They won't. Don't hold your breath.

I say this as someone who's been a member for 10 years, 2 months. Actually, it's longer but I've had a few accounts over the years.

You joined: 3 years, 11 months ago

3 years, 10 months ago I posted on MSO: What's up with the breakdown in communication between Stack Overflow and the community?

Things have gotten worse. Much worse. Not better.

Infinitely chasing higher profits and more user activity has got us here and that mindset is not changing any time soon.

4
  • 2
    nit: the question asker's math.se and mathoverflow profiles were created on 2018-10-08- before their MSE profile.
    – starball
    Commented Jun 10, 2023 at 0:12
  • 1
    @starball good point. I forgot that you could join communities later.
    – Script47
    Commented Jun 10, 2023 at 0:38
  • 1
    @Script47 yeah, only CM's insta "join" all 180+ sites when assigned as a CM, with a diamond in each. :-) Commented Jun 10, 2023 at 13:14
  • Infinitely chasing higher profits and more user activity has got us here and that mindset is not changing any time soon. => and the only way to change this is to either: a) Start a competing site or b) Acquire enough resources to buy out SE, as was recently done by Elon Musk with Twitter. There is no option c) Commented Jun 14, 2023 at 10:43

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .