Blog: Stack Overflow

Most of these posts were originally posted somewhere else and link to the originals. While this blog is not set up for comments, the original locations generally are, and I welcome comments there. Sorry for the inconvenience.

Thoughts from a former community manager at Stack Overflow

I came back from Shabbat to a link to this interesting blog post by Jon Ericson. Jon and I haven't discussed this.

The original post contains links that I haven't reproduced in this excerpt:

After contemplating the situation for many years, I've come to the conclusion that Monica ran into a wall of injustice veiled in the language of progressivism. Applying Bari Weiss' framing, Monica was powerful within the community so her behavior was suspect by default. The factors I thought were to her favor by the new ideology didn't seem to matter:

  1. She has vision problems which puts her at a disadvantage in the age of screens.
  2. She's a woman in technology which means she's in the minority.
  3. She's Jewish which puts her in a minority that's been discriminated against so often there is a common word for it in English.

The analysis I should have understood was:

  1. It's possible the people deciding her fate didn't know about her vision. In any case, vision is a problem that can be corrected with technology and money.
  2. In the calculus of intersectionality transgender people are more marginalized than straight women.
  3. What I thought were strong arguments that removing a Jewish moderator on the Shabbat before Rosh Hashanah was a bad look, turned out to not matter. I can't prove it, but I suspect it's the result of subtle antisemitism that comes from observing that Jews tend to be successful in certain fields. Jew might be a minority, but they aren't under-represented so paradoxically that must mean they are among the powerful.

I'm not an expert on these things and so I operated under the naive assumption that progressive ideology was working toward the goal of treating people as if we were all created equal. But the standard tools of the new morality are ineffective. Instead, the logical conclusion of the new ideology appears to require mistreating people who don't conform to its evolving standards.

Stack Overflow: still digging itself into a hole

Granted that I'm biased, but if you're still using Stack Overflow or Stack Exchange, either the free sites or the paid service, it's probably time to reconsider. Squandering community trust was already a core business practice, and now it seems like they're having trouble keeping the lights on despite massive cash infusions.

2023 has not been a good year for them. In May they laid off 10% of the company including 30% (!) of engineering, and diverted 10% of those who remained to chasing the AI hype train. Then they barred moderators from acting against ChatGPT-generated nonsense while lying about that policy to the larger community, causing an unprecedented nine-week moderation strike. Early in the strike, it came out that the CEO had personally ordered that the regular data dumps be secretly shut down. (They apparently did not secure the silence of the people they fired.) Those dumps were, from the beginning, a company commitment to the users as an insurance policy against the company turning evil -- you could always take the data and go elsewhere. Except now you couldn't. So that was kind of a big deal, and restoring the data dumps got added to the strike demands.

The strike eventually limped to a settlement, with the ChatGPT policy mostly rescinded, the dumps restored, and a company promise to communicate better. Many remained skeptical; company claims of caring about the community have not stood up to scrutiny in the past, and the current CEO seems especially disdainful. I guess people decide when they've hit the trust thermocline at different times; for some of us it came in 2019, some earlier, and some over the intervening years, and some haven't hit it yet. (This is why it's so hard for communities to migrate. Communities don't move; they fragment.)

But while they've been mistreating their communities, it looks like they've also been having trouble with their paying customers. Cory Doctorow's essay on enshittification) comes to mind.

On Monday they laid off another 28% of the company. The layoffs included another two community managers who had advocated for the community, reminding me of when they purged people who had pushed back against toxic company actions in 2019. Questioning the executive team is dangerous to one's career. People are asking some rather pointed questions about the latest action, not that we should expect any meaningful answers. I think the VP who opened that discussion did it to try to channel the venting, not because anybody in company leadership cares.

In the past, the tension at Stack Overflow was between investing in the business to make money and investing in the community whose content enabled a lot of the business. There were trade-offs -- can we make more money from ads without pissing off users, can we neglect maintenance the communities depend on to invest in the SaaS product, can we lower our quality standards to draw more beginner "engagement", etc. "Trade-off" implies that you're giving up something to get something else, but what they're currently doing seems to be bad all around -- they're failing to make money from their paid products and also failing their communities. Prosus, who bought Stack Overflow in 2021 for a jaw-dropping $1.8 billion, must be feeling like chumps right about now. The cost-cutting feels like leadup to a sale, presumably at a large loss, to stop Prosus's bleeding. I wonder how that will go. I'm so glad I don't have to care.

Three weeks into the Stack Overflow strike

I still don't have time for deep commentary (just got back from Origins; post about games to come), but there have been some developments since the Stack Overflow moderation strike began on June 5:

Data dumps

From very early on, Stack Overflow Inc. has published a quarterly data dump of all of the content (with attributions etc) from all network sites. This was the explicit insurance in case Stack Overflow turned evil in the future, like Experts Exchange, the company that led to SO being created, did. That stuff all uses the Creative Commons license and is meant to remain available.

Someone noticed that the June dump had not been posted on schedule, and asked a question about it. One of the people who was part of the 10% layoff in April replied, saying that the dumps had been disabled at the end of March with an annotation that they were only to be restored at the direction of the "Senior Leadership Team" (this usually means C-level executives). That drew some attention.

The company spent several days ignoring, then brushing off, then making excuses for this unannounced change. Nothing they said was credible. The strikers added "restore the data dumps" to their list of demands. After almost a week, the June dump was posted. No public promises have been made about the future yet as far as I know (though, see "was away for several days" above).

Spam overflow

With about 1500 curators (including about a quarter of moderators network-wide) on strike, and more importantly with the volunteer-run anti-spam automation turned off, the junk's been piling up. Reportedly, employees are now spending time handling spam, cutting into their day jobs.

While we're told that discussions are happening between representatives from the moderators and the company, they don't seem to have made much progress. A moderator told me that the company committed to keeping the data dumps coming, but it sounded like it was specific employees making the commitment, so the promise might not outlast their employment.

Rules for thee but not for me

In addition to violating the moderator agreement in a few ways (leading to the strike), the gen-AI-hype-chasing company recently announced that they are going to launch a site for "prompt design" (I am not making this up), but they're not going to use their existing process for creating communities because it doesn't work well, so instead they're looking for people to be part of a behind-closed-doors steering committee or some such, with the goal of launching the site by July 26.

The CEO is giving a talk about gen-AI hype at some conference on July 27.

Meanwhile, people who are trying to launch communities using the current process would like a word.

Meanwhile, over at Codidact...

Stack Overflow Inc. has given us a gift. We have lots of new participants and new activity, and some active efforts to build new communities here. Nice! We've gotten some questions about differences and was starting to think that we need an "immigration guide" and then someone reminded me of this early post asking about differences -- with a new answer from one of our new users. Nice.

It sounds like we might also attract some contributors on GitHub, which would be great. We have many things we want to do and not very many people.

Stack Overflow is alienating its community again

I don't have time for a full writeup of this right now, but here are the "highlights" of Stack Overflow Inc.'s latest community-affecting actions.

The CEO has recently gone all-in on generative AI and LLMs, the technology used by ChatGPT. He allocated 10% of the company to work on unspecified ways to use LLMs in their platform, and he's made some incoherent blog posts that scream "chasing the hype train". He also laid off 10% of the company including 30% of engineering and two community managers.

Stack Overflow the site does not allow answers written by ChatGPT. They worked together with community managers to develop that policy. Their moderators are seeing an increased workload because there's so much machine-generated crap showing up now, but the moderation tools and processes in place are handling it.

Or were. On Monday the company announced a policy that basically bars moderators from moderating this content. For further complication, the public announcement does not match what moderators say they were told privately -- they were actually told to start enforcing a strict hands-off policy without letting users know.

(The public post kind of back-handedly called moderators bigots, too. I guess at least this time they didn't smear anyone by name. But still... ick.)

People are, naturally, upset, both by a policy that invites non-vetted machine-generated "answers", and by the way it was done. Moderators' attempts to discuss these issues with the company have been rebuffed. One popular theory is that the CEO, having gone publicly all-in on LLMs, was embarrassed to find out that his flagship site deletes that stuff.

So there's going to be a strike. More than half of the Stack Overflow mods, many other mods across the network, non-moderator users who do the important curation tasks, and the user-run tools that detect spam and other problems across the network -- all shutting down. These people are all unpaid volunteers who are realizing that the company that relies on their free labor doesn't actually care about them.

Noticed in passing: there are a bunch of userscripts that power users use to make the site easier to maintain. These scripts are very popular. One of them now adds a banner to the top of the site that says:

We are calling for solidarity against actions taken by Stack Overflow Inc, which is posing a major threat to the integrity and trustworthiness of the platform and its content.

Clever.

For more detailed background and why this matters so much to the people involved, I recommend this post from a former community manager.

--

Update, 2023-06-05: From Meta.SE: Moderation Strike: Stack Overflow, Inc. cannot consistently ignore, mistreat, and malign its volunteers (includes demands), mirrored on Stack Overflow Meta.

As 5782 draws to a close...

There years ago there was a pile of bad behavior at Stack Overflow Inc. This week, one of the people involved, who no longer works for them, posted Reflections on years of guilt, through the lens of Teshuva. How unexpected and refreshing! Some excerpts:

To Monica, who I hope still thinks of me as a friend: I failed you because I couldn’t stop a horrible train of bad decisions without exposing things about myself that could have ended my family if they came out in the wrong way, and the health insurance I desperately needed. I was also worried that those who knew these things about me were increasingly strained in their restraint and that things coming out was a possibility; I had very real reason to believe more people would speak out. You did not deserve to be let go the way that you were and I’m sorry that I couldn’t stop it. You really didn’t understand what everyone was taking issue with, and I didn’t get you the space necessary for that to happen. Clarity now exists around this, but it came at your expense, and my failure to act enabled that. Monica Cellio isn’t a bigot, she’s a pillar that I stepped on instead of building up more.

To coworkers that I steered away from helping Monica: I had the most terrible of best intentions, keeping you out of harm’s way. I realize that I took away your choice to do something better than the person I was capable of being due to … constraints. While it was coming top-down, I should have refused to let it go any further. Resigning wasn’t an option I could take. I didn’t feel like I could even privately question anything anymore. What’s bad for a manager is twice as bad for those that report to them; I won’t make that mistake again. My piece in the puzzle should have broken by design.

To coworkers that were let go due to retrenching — I didn’t know it was coming, but I sure as hell didn’t fight the thing that was running you over once I saw it running you over. I’m not proud of my silence that day and you deserved something way kinder than what you got.

I had thought of Tim as a friend in the past. Then that happened and I didn't. I feel like we now have a basis for rebuilding.

Activity trends on Stack Exchange (mostly downward)

Stack Exchange published some year-end statistics, as they've done for the last few years. Their focus was on looking at how many questions get closed, per site (sometimes a few, sometimes more than 60%) and how many of them get edited and reopened (very few). I don't really care about that, at least at the level of detail they have in some wide tables, but the data includes the number of questions asked for the year, and that piqued my curiosity. My impression has been that activity in general and new questions are down, sometimes a lot, across the network, but I hadn't looked at all the data. Until now.

I downloaded the CSV files for 2018 through 2021. As with all such efforts, the task starts with data-cleaning -- some site names were not consistent across the four files, which messes up sorting and grouping. There are older files, but I got tired of hand-adjusting site names and filling in placeholders for sites that didn't exist that year. While, anecdotally, SE's been in decline for much longer, it feels like things accelerated in 2018 and 2019, so that's what I looked at.

I loaded everything into a spreadsheet for now, just to be able to eyeball it easily. (I might load it into a database later for actual queries, and would include the close-related data if I do that.) I added a "trend" column, up/down/stable, based on eyeballing the data. There were some edge cases that led to me adding an 'erratic" option too. And sometimes it's a judgement call; I didn't work out precise formulas, and some of my "stable"s could be "down"s or vice-versa. I can share the file I assembled, if you want to ignore my assessments and make your own.

The counts include all questions that were asked; we don't know how many were subsequently deleted.

There were a few sites with upward trends:

  • Biblical Hermeneutics
  • Chess
  • Engineering
  • History of Science and Math
  • Islam
  • MathOverflow
  • Quantum Computing (which has a corporate sponsor and is a hot research topic)
  • Christianity (recent rise)
  • Literature (recent rise)
  • Operations Research (recent rise, site created in 2020)
  • Retrocomputing (recent rise)

There are about 30 I judged to be steady; while there are fluctuations year to year, the activity is all in the same general ballpark.

The other ~140 sites are having more difficulties.

A few that jumped out at me and/or are of personal interest:

  • Mi Yodeya had about 4000 questions in 2018. In 2019 there was an organized activity that led to more questions, about 4500. It took a drop after that (~2900 and then ~2600).

  • Writing was at ~1800 in 2018. 2019 was the year of the question drive and a big push to finally be allowed to graduate. We succeeded, and then six weeks later Stack Overflow Inc. blew up our world. 2021 is down to ~1200, which is actually higher than I expected.

  • Some technical sites are down by about half, including Software Engineering and Quality Assurance. Hmm.

  • Cooking went from ~2700 questions in 2018 to ~2300 in each of the next two years and then to ~1500 in 2021. I would have expected that site to get a boost from the pandemic (more people are cooking at home). On the other hand, Travel dropped a ton from 2019 to 2020 (and stayed low for 2021); I think the reason is pretty obvious there.

  • English Language and Usage dropped by half, from ~22k in 2018. The Workplace is also down by half, from ~5500 to ~2500. Graphic Design is down even more, from ~6800 to ~2700.

  • Interpersonal Skills had ~2600 questions in 2018. In mid-October of that year, they were kicked off the hot network questions list amidst some controversy, reducing their advertising reach. That doesn't explain everything, but the next year was down by half, and in 2021 they were down to ~450, less than 20% of the number from 2018. Noteworthy: when SE decided to "graduate" almost all remaining beta sites, this one asked to stay in beta. So they seem to recognize that they have things to sort out.

  • Beer, Wine, and Spirits, which has struggled for years, is down from 151 questions in 2018 to 70 in 2021. Homebrewing is also down by two-thirds.

I couldn't figure out how to make Markdown tables without a ton of manual effort. Here are a few screenshots for the curious. Read more…

Licenses: only as strong as the will to enforce them

I realized, while doing some digital housekeeping, that Stack Overflow Inc. is using some of my writing, without attribution, in violation of the Creative Commons license under which they obtained it. To correct it, I would have to file a DMCA takedown demand, which would require me to be in contact with them and their lawyers, which is icky quite aside from any retaliation they might engage in. I don't currently have the will to deal with that, so they get away with violating my copyright and license.

I'd likely be ok with them using it, at least on the public sites, with the required attribution, and if they do so for other people's work they've taken, too. (I know I'm not the only one, because another help-center article that I wrote is in turn built on something by someone else.) I think they're banking on the fact that engaged users don't tend to mind and people they've driven out don't want to take on the burden of securing a correction. Such things do a lot more damage to the individual than to the corporate behemoth, after all. In their younger days they would have readily done the honest and ethical thing, but these days?

They can get away with it because they can make things unpleasant. But having noticed it, I can also note it here.

Tainted elections

Stack Overflow (Stack Exchange) has been faltering for a while for a variety of reasons that boil down to "still don't know how to work with rather than against their communities and power users". Even so, I'm surprised to see three corrupted moderator elections in a span of two weeks, one of them on the one site they actually kind of care about, Stack Overflow.

First up is a Stack Overflow election. I became aware of this incident when I noticed an extreme spike in view stats for Dear Stack Overflow, we need to talk on one day and looked around to see what might be causing it. During the voting stage of the election (the second week, after a week for nominations), the moderators and community managers (SO employees) jointly decided to remove a candidate. They did not suspend the user, so this is already on very shaky ground -- the community is supposed to choose its moderators from candidates who meet the eligibility requirements, which this candidate still did. Now, this candidate had done something problematic, and if they had suspended him for it then his candidacy would have been revoked legitimately, but they decided it wasn't bad enough to suspend over.

They didn't stop there, though. They announced on their meta site that the candidate had been removed, they talked about the allegations, and they did all of this before talking with the candidate. Their reasoning was that they had to make a prompt announcement so that people who had already voted would know to review their votes. Except, no -- they didn't need to do that. SO is fully capable of pausing an election; they only needed to announce a pause (without removing anyone), discuss it with the candidate, and reach a resolution -- like letting the candidate gracefully withdraw. SO recently restarted an election on another site, so there's already precedent for intervening in the timeline for extraordinary reasons.

I would think that "we'd like to avoid smearing a user in public" would count as extraordinary reasons, but apparently not. The candidate returned a day or two later, having suffered a local Internet outage in his part of the world. Imagining getting back online, going to a site you care enough about to want to lead, and seeing that. Cringe.

It was completely avoidable, had the community managers running things cared to avoid it.

Read more…

Followup to "Dear Stack Overflow, we need to talk"

In 2018 I posted Dear Stack Overflow, we need to talk after SO took some serious Twitter-driven steps against their own communities, moderators, and users. Three years later I got a reply from someone who'd struggled as a new SO user. This was my reply to that comment.

Thanks for sharing your journey. I, too, was an active contributor for many years (not so much on SO but on several of the other sites), and I had to learn the rules and the process and the culture like you did. Individual communities vary, but I had some great experiences.

In recent years the company has taken some (IMO) serious missteps, driving out a bunch of their (former) top contributors. It felt like the company, instead of working with the community in partnership, found the community to be a burden. They lost the community focus they once had, unfortunately.

Some of us decided to try again, putting the community first with an open-source platform and a non-profit organization. We're small but we've been able to build some exciting new tools on the Q&A foundation, and we're building a network of communities that are community-run. We have a Software Development community; it doesn't get 6000 questions per day like SO, but it gets questions and they get answers. Perhaps you'll add it to your rounds, popping in occasionally to see what interests you there too? You can learn more about the project at https://codidact.org.

New VP of community at Stack Exchange, prognosis unknown

Stack Exchange recently promoted someone to VP of Community, and he posted on Meta asking what to change and what is inviolate. It's too soon to tell if these are just empty words, as is the norm with Stack Exchange leaders in recent years, or if he intends to and will be allowed to work with the community. Someone pointed all this out to me, so I figured, hey, I'd log in for the first time in many months and accept his invitation (posted yesterday afternoon):


Today is Tisha b'Av, the date the ancient Jewish temple was destroyed. (I promise this is relevant.) According to our tradition, the second temple was destroyed because of baseless hatred, sinat chinam. Among all the problems of the time, one incident stood out as the precipitating event:

A wealthy man held a party and sent his servant to invite his friend Kamtza. The servant misunderstood and made the invitation to Bar Kamtza, whom the host hated. Bar Kamtza, thinking the man was offering an olive branch, attended. The host was furious and ordered him to leave. Bar Kamtza, trying to save face, repeatedly tried to make peace, offered to pay for his food, and even offered to pay for half the party. But the host expeled him in front of all his other guests, none of whom objected, setting in motion a chain of events that led to the destruction.

The host hated Bar Kamtza so much that he no longer saw him as a fellow human being deserving of basic decency and dignity. Presented with the results of a misunderstanding, the man in power escalated instead of de-escalating, harming everybody present (and, according to the account in the Talmud, the whole nation).

Philippe, your predecessors didn't destroy a whole people or a national treasure, but there has been a lot of baseless hatred and harm and pain to lots of people over the last few years. Some of that can never be repaired, but some still can be, even at this late date. What has been missing is not the ability to correct errors but the will.

What should you change as quickly as possible? This ongoing failure to make what amends and repairs can be made. It's the ethical thing to do, and -- to speak to the company's business-driven interests -- it would show the people who build Stack Overflow and the SE network that you're willing and able to correct mistakes. Everybody makes mistakes; we learn a lot about people and institutions by seeing how they handle their effects. Yes you have the power of the wealthy party host, but is that the kind of person you want to be?

What should you never touch? The community's goodwill. You have the potential for awesome partners in growth, people who still want to see Stack Overflow succeed despite it all, people who know a lot about how to do that on the community side. You've got lots of professional experience but you're new to SE and SE jettisoned decades of its CM expertise in January 2020. The previous people at upper levels not only didn't engage with the communities but shunned them. By coming to Meta and starting this conversation you've taken an important step. Keep that up and follow through: engage with the community, participate on some of the 170 communities, ask for feedback regularly, carefully listen to feedback (which is not the same as "do what we say"), don't spring disruptive changes on people -- treat the community as partners not enemies.

(I realize much of the previous paragraph belongs in the "what should I change" paragraph, because what needs to change is the corporate attitude, but the reason it needs to change is that somehow you still have a community here that cares, and you should work hard to maintain a good relationship with it.)


I was asked in a comment some time later if anybody from the company had contacted me. On 2021-09-27 I updated the post to say that I have not been contacted.