Introduction: Community Manager

General
1

I have recently been appointed as the Discourse Community Manager. Here I want to share my plans, thoughts and visions with the position and the community. I am looking forward to feedback from users, developers, educators and everyone else engaged in the Stan community.

It turns out brevity is not my strength, so this post is divided into four parts for clarity:

  • Process - what this text means and what happens next
  • What I think the ideal community is.
  • What I think is the state of the community.
  • What I think we can do to improve.

The structure is here to make it clear where any disagreements lie - it makes little sense to discuss low-level interventions if we donā€™t agree about the overall goals.

Process

In its current form, this is my personal statement and does not reflect the view of the full community - everything should be prefixed with a big ā€œI thinkā€¦ā€. However, provided there are no strong disagreements and after incorporating feedback from the community, an updated version could become an official medium-term roadmap for the Community Manager position.

There also seems to be disagreement about whether the Community Manager position should exist in its current (paid) form and whether the provisional SGB that appointed me has full legitimacy to make this kind of decisions. I hope the community will find the ideas below and the ensuing discussion fruitful regardless of whether the position will continue to exist and I believe little of this postā€™s content depends on my formal status.

What I think the ideal community is

Discourse is a nice place to be in. People contributing to the Discourse feel good about their contributions and their role in the community. People feel welcome and the barrier for constructive engagement in the community is low. Reading the forums is pleasant and motivates readers to consider becoming active in the community.

Discourse is a place to seek help. Everybody asking for help in good faith gets support. ā€œSupportā€ might not necessarily mean ā€œanswerā€ or ā€œsolutionā€ as some problems are just hard and some questions do provide either too much or too little information to allow us to give a good answer. However, the asker should always learn why the community thinks the problem is hard or how the question could be improved. Nobody in good faith should avoid asking questions because they feel afraid or intimidated by the community. Here good faith is a problematic quality that is hard to judge so we should err on the side of assuming good faith. I believe losing a potentially great member of the community is a higher price to pay than reacting constructively to some lazy questions. Also assuming good faith is IMHO the morally right thing to do.

Discourse is a place to discuss development and future directions for Stan. Discussion even about contentious topics should be civil and constructive. Everybody who is connected to Stan in any way should feel OK about expressing their opinions about topics they find important. The discussions are transparent and it is reasonably easy to stay updated on what is being discussed and what the proposals/conclusions/ā€¦ are.

Discourse is a place to share news/accomplishments/job offers/ā€¦ related to Stan. It should be easy to determine what people can (and should) post and what is off-topic.

There are clear boundaries of what Stan Discourse is not. This one I am less sure about, but I fear that if Stan succeeds, we may face the fate of Cross Validated: very broad community having way too many questions and too few people who can answer them competently. I think statistics communities face greater risks in this regard than software or other communities as the amount of people that need to use statistics is huge while the state of statistical education is sad across the world. We can partially mitigate the risks by high-quality community that helps people quickly become competent answerers, but I fear this approach might not scale. I therefore believe discussion about boundaries will have to be held at some point.

Where are we now

Discourse is a nice place to be in.

Mixed feelings. On one hand, Stan Discourse is IMHO way better than Cross Validated and many other online communities. My personal experience has been mostly very pleasant. On the other hand, there is a tendency for very compressed style of responses which (regardless of intent) may be interpreted as either ā€œno BS, straight to the pointā€ or ā€œstern, discouraging, almost aggressiveā€. I believe notable fraction of readers might read the posts with the latter interpretation and be turned away from involvement in the community. There is also a big difference between responses to user questions (which tend to be nice) and discussion on development topics (which tend to be way too harsh for my taste).

Discourse is a place to seek help.

I believe we are doing very well in this regard. Our response rate is very good. This does not necessarily mean people get the answers they need but it is likely they donā€™t feel neglected. Some more nuanced data could be probably gathered at low cost (e. g., a short survey sent out to all first-time askers from past few months). It is not clear to what extent people fear posting their questions and it would be hard to get good data on this (but a survey of first time askers could still shed some light). Also a few good questions still fall through the cracks each month and receive no attention.

Discourse is a place to discuss development and future directions for Stan.

While most of the relevant discussions do happen at the forums, the style of the discussion is IMHO far from ideal. There seems to be quite a bit of history between the more vocal members of the community which I cannot parse but which obviously influences the present discussions. Some posts can be read as openly hostile and it seems some contributors feel like they are not treated fairly.

It also looks like that being stern in technical discussions is sometimes equated with being rigorous and being nice with being lax or lowering standards. This is IMHO unfortunate as I believe one can easily be both nice and rigorous and keep high technical standards.

Keeping track of what is being discussed also feels challenging since topics frequently meander quite a bit from the first post.

Discourse is a place to share news/accomplishments/job offers/ā€¦ related to Stan. This seems to be happening to an extent. I donā€™t believe this is a high-priority role of the Discourse forum and so will not discuss it further.

What can be done

Small changes to etiquette for a kinder community:

I believe there are a few small changes that we could encourage across the community which are both ā€œcheapā€ to do while possibly having noticeable impact.

  • Let readers avoid the ā€œsternā€ interpretation of posts by adding small pleasantries (ā€œHope that helpsā€, ā€œGood luck with your modellingā€, ā€œSorry to hear you are having troubleā€, ā€¦). This is some extra effort, but this is exactly what common courtesy is: giving away extra effort signals you care.
  • If a question (especially by a first-time poster) looks reasonable, like it. This is a clear signal to the asker that the question is welcome here.
  • Communicate emotions. If a topic frustrates you, say it and explain why.

It doesnā€™t really make sense to enforce this, but maybe just agreeing that this is what we as a community want might help?

Code of conduct & Discourse Guidelines

The SGB is planning to have a discussion on adopting CoC for all Stan-related spaces. I plan to help with that. Further, some Discourse specific guidelines might need to be developed, possibly derived from the parts of current guidelines (https://discourse.mc-stan.org/faq) that are too specific to be included in a Stan-wide CoC. The guidelines might involve some of the suggestions from the previous section. Once CoC is settled I plan to draft the guidelines and post them here for further discussion.

Inspiration from other communities

Iā€™ve heard very good recommendations about the way RLadies and The Carpentries handle their communities and plan to investigate what we can learn. If anyone has other positive examples or links to comprehensive resources I could use, it would be welcome.

Making sure people get support

I generally think we should reply something to basically all questions, but I am not sure about the response time we should ā€œguaranteeā€. I feel that too short ā€œguaranteed response timeā€ might be problematic as we lose the opportunity to signal that a question is not really well written or has other issues (and hence takes longer to receive support). This might be important if we start to get more bad-faith askers (most notably the ā€œdo my homework, pleaseā€ kind, which we fortunately almost never get). Longer period also gives more opportunity for people that are not usually active to spot a question they can answer, increasing their engagement in the community. Several workdays ā€œguaranteed response timeā€ might be a reasonable goal.

  • I am monitoring single user topics and try to provide support where I can or @ someone who can answer.
  • Set up a list of people and their expertises to @ in topics that donā€™t get replies quickly. This list probably should be public, so that all active users can use it to help others get support, but I would not go as far as advertising it directly on the forums.
  • Invest in connections with active participants, especially those outside the Stan dev team - both informally (direct communication) and formally (e.g. a special badge on forums, reduced/no fees for attending StanCon, ā€¦). Make sure they feel part of the community and support beginner answerers.
  • Be generous with likes to answers.

ā€œNoviceā€ Category

A special category for questions of novice users. The goal is threefold:

  • Lower barrier for entry for people having problems with Stan, who are intimidated by the high technical level of the general discussion
  • People not interested in helping beginners can mute the category
  • People who donā€™t consider themselves proficient but want to help can focus on those questions

Not sure how this should interact with the interfaces subcategories (especially ā€œbrmsā€), which get a notable share of novice questions these days.

Nicer discussions among developers

This seems very hard to influence from my position. Maybe I can at least sometimes express when a discussions seems to have become too aggressive.

27 Likes
2

Hi @martinmodrak!

Congratulations on the position! :) I donā€™t have much to contribute to what youā€™ve said - it all sounds pretty reasonable to me. :) I love the idea of a spreadsheet where people can sign up and put their area of expertise for tagging - thatā€™s a great idea.

On the code of conduct - @andrewgelman asked me to think about this and write a draft. Perhaps we can collaborate? Iā€™d love to hear your ideas.

Weā€™ve also done some work developing items for a survey of Stan users, which might be helpful for your new users survey.

Feel free to shout out or tag me if you need a hand with anything I can help with.

5 Likes
5

Great introduction @martinmodrak, I agree with what you write, and I believe we can improve even with with small actions when they are more organized actions.

very compressed style of responses

I know Iā€™m one of the very compressed responders.

  • Iā€™m slow writer even in my mother language, and here Iā€™m writing in foreign language which makes the mental effort even bigger.
  • There are cultural differences: Finns tend to have less small talk and go promptly to the topic. For a Finn an adding extra sentences is another mental effort and I often worry that these extra parts can actually look like sarcasm (e.g. Iā€™ve learned years ago that ā€œhope this helpsā€ definitely has a negative connotation).
  • Itā€™s often time and mental effort issue whether I write even very compressed answer or donā€™t answer at all. I often wait day or two and hope that someone else answers certain questions and if no-one else answers, I prefer to write even very compressed answer instead of not answering, because I care.
  • I have already been thinking that I should have a template for ā€œsmall pleasantriesā€ and just not think about it and risk that they look silly.

Yes, I want to be better and avoid ā€œthe ā€œsternā€ interpretation of postsā€, and will follow your advice, but I may also now more often decide that I donā€™t answer compressed and rely that Community Manager will contact me if there are unanswered questions in my domain of expertise :)

One addition to how we could better get right people to answer is if community manager and others could adjust category of topic if itā€™s wrong. Seems to happen often that people donā€™t realize to choose the category based on the interface or other Stan related package.

Writing this took about 37mins (which is unusually fast for this many words for me), and I didnā€™t need even think about any technical or modelling question.

Hope this helps (uh, now I donā€™t know whether this is pleasantry or negative thing as Iā€™ve learned more than 20 years ago)

39mins because I started to rethink that HTH thingā€¦

10 Likes
6

I didnā€™t know if anyone was on this already so I wrote a draft CoC and contributor guide for Stan math on Monday (which could probably be used project wide). The CoC is pretty general though if yinz want to write your own Iā€™m open to whatever!

3 Likes
7

Congrats on the post, @martinmodrak! Youā€™re certainly one of the most helpful people around (along with likes of @Guido_Biele, @stijn, @nhuurre, @mike-lawrence and @jjramsey, to cite only people I think are not developers).

Iā€™m happy to help with general modelling questions and more conceptual questions. Those seem to be @betanalphaā€™s favourites too, so I guess Iā€™ll scoop up the ones he misses. Not very comfortable answering brms-related questions (even if they have modelling questions underneath) as Iā€™m not a frequent brms user myself.

Since earning my ā€œRegularā€ badge or something like that Iā€™ve changed the topics of one or two posts, but in my experience most topics seem to be in the right category (i.e. have the right tag). Iā€™ll be more attentive from now on and try to help on that front.

Anyway, am excited about this new structure of the Stan community and happy to contribute a bit more, as time permits.

7 Likes
8

Iā€™ve moved this discussion to a separate thread, tagged all I know who are interested and want to be actively involved (please add anyone Iā€™ve missed) and added a link on the issue. :)

3 Likes
9

Congratulations, @martinmodrak. Youā€™re an excellent hire!

Thanks for the shoutout, @maxbiostat. Youā€™re generous with your praise and likes and, as Martin alluded to, that actually works motivating and welcoming. I am trying to emulate this more.

I am happy to be on some kind of list or spread sheet. I think I am pretty good with general modelling questions, linking domain knowledge to the stan models, interpretation of parameters, and the R part of Rstan. My main limitation is that my activity here waxes and wanes with my teaching load.

Would it also be helpful to have a ā€œdiscussionā€/ā€œmetaā€ category on discourse aimed at frequent contributors? It could be a place to gather (links to) responses to similar questions, links to non-discourse resources, or to share tips on how to answer nicely and to-the-point on particular topics. It might also be a way for Martin to light the Bat-Signal if there are questions that need more help.

7 Likes
10

Thanks everybody for the kind words! Some reactions follows.

First - I am glad you took the time and effort to write your honest and important reply. I certainly donā€™t want to force anybody to do something they feel uncomfortable with - there are many tools to help the Discourse feel as a nice place and not everybody has to use all of them all the time.

This is the first time Iā€™ve heard about this, but I guess online communication can be that terrible :-(. I find it useful to approach this from a place of honesty: When I reply to a question, I genuinely hope I am helping/wish good luck/ā€¦ and I have every reason to believe this is the case in at least 99% of the posts here. So my suggestion is less like ā€œuse polite phrasesā€ and more like ā€œremember to make your intentions explicitā€. Does that make sense?

I also understand some people can have a negative motivation when engaging in a discussion (e.g., ā€œI canā€™t let this crap stay unansweredā€, ā€œHow can they be so dumb?ā€, ā€œSOMEONE has to take care of this messā€). While it is OK to feel like this occasionally, if anybody is driven by negative sentiment on a regular basis it is IMHO very unhealthy. Either there is something broken in the community or the person should rethink the way they interact here and try to find a way to contribute to Stan without incurring pain on themselves and possibly others. I am open to discussing both cases if a need arises. No snark or sarcasm here - I sincerely believe that anyone being frequently upset while interacting with the community is a lose-lose situation and we can do better. I also very much donā€™t think this actually occurs on the Discourse now (at least not to a noticeable extent, but let me know if you disagree), but it may happen and it felt like an important point to follow the previous paragraph.

That is both completely valid in my book :-) But sorry if you understood my post as saying you are harming the community - your contributions are very useful and I just wanted to encourage everybody to try to be even better.

Good point - I mostly brows all topics so I was not paying attention to their categories much. Will keep an eye on that.

@maxbiostat Thanks for highlighting people who do good work. And thanks to the whole Stan team, @Guido_Biele, @stijn, @nhuurre, @mcol, @mike-lawrence, @emiruz, @Max_Mantei, @stemangiola, @increasechief and @jjramsey and everybody else pouring their time and energy into this forum (sorry to those I missed). Only after sifting through site stats have I noticed how many people keep this forum running. I hope to be able to support you in helping others.

Thanks for the suggestion - it might be useful. We currently do have quite a bunch of categories, so maybe just having a tag under the ā€œGeneralā€ category could be sufficient (or using the ā€œdiscourseā€ tag we already have) unless it turns out we really have a lot of topics falling under this category. What do you think?

4 Likes
11

I noticed some active members of the community from whom I would really like to receive feedback, have not reacted to this topic. I would be very glad if you did! This includes @Bob_Carpenter, @mitzimorris, @wds15, @andrewgelman, @rok_cesnovar, @paul.buerkner, @betanalpha, @sakrejda, @jjramsey, @increasechief, @ariddell, @charlesm93, @anon75146577, sorry if I missed somebody. It is however completely OK if you arenā€™t interested in this topic or if you think this discussion is secondary to other stuff happening in the community and you donā€™t want to react until those other things are resolved.

I also think some of the people I tagged are less than enthusiastic about of some of the recent developments in governance and related stuff, and some have had interactions with me that could have left them feel less trust toward me (sorry if I misread anybody), so I feel it is especially important that I know whether they are in at least broad agreement with my plans .

2 Likes
12

I appreciate the unexpected acknowledgement @martinmodrak , and Iā€™m generally happy to help more if I can. I find trying to answer questions allows me to live vicariously through other peopleā€™s problems which in turn develops my own skills. I see it as a net gain.

Iā€™m also very grateful to warrant the attention of heavyweight statisticians and professors, some straight of the cover of my BDA3 book, who take their time to answer questions here. This is a phenomenal forum in my opinion, perhaps currently quite centered on Stan the tool, but could easily grow into being the epicenter of the Bayesian movement.

13

Thank you @martinmodrak for this draft and your willingness to work as a community manager! I have not much to add other than saying that I agree with what you say and try to follow it more in the future.

And thank you also for answering so many brms related questions in such detail! This has been a huge help for me, especially since I sometimes get annoyed by certain types of questions, and then my answers are often not very helpful and/or impolite. I will try to do better :-D

14

Its good to me glad to have you!

15

Donā€™t worry, I didnā€™t assume that. I maybe should have also mentioned explicitly that itā€™s good that native English speakers should assume that often reason for short or impolite sounding wording also for other Stan team members and novice question askers can be due to language difficulties. I see you write very good English (but it may also be just because Iā€™m not native speaker)

1 Like
16

I am not really sure what the best option is. It depends a bit on what we actually want to use the channel for. You could have a separate slack as well but then thatā€™s not necessarily open to everyone. Or a separate wiki with good answers but then a wiki is less open to discussion or short conversations. So I thought discourse was the best option but I wanted to make it easily ignorable for the devs (if they want to ignore it), hence a category. But I realise that it can still clutter the forums.

1 Like
17

I hadnā€™t heard that either. I often say ā€œhope that helps!ā€ when I respond but I mean it sincerely! In fact I just used that here:

1 Like
18

@martinmodrak Iā€™m glad we appointed you to this position and thanks for the great introductory post! In general I agree with everything that you wrote, so Iā€™ll just add a few brief comments.

Agreed!

I agree. Like @avehtari said, in some cases this is probably due to English not being everyoneā€™s first language or different languages/cultures having different styles of communication (in Akiā€™s case I think those fully explain it). In other cases I think people are just short on time and trying to be helpful without it taking too much time, but I completely agree that this can come across to the reader as ā€œstern, discouraging, almost aggressiveā€ regardless of intent, so we should definitely encourage everyone to be mindful of this. In the minority of cases maybe someone is trying to be discouraging or aggressive and thatā€™s definitely a problem.

I agree. To give some background as someone who has been involved with Stan for many years now: this is a problem that predates the Discourse forum and in many ways has just carried over from a problematic way of communicating in person (or by video call) back when Stan was much smaller. Some of us have been trying to improve this for years, and with some success, but it has been difficult. To be fair, in some cases developers are used to interacting with each other like this and donā€™t take offense (and in fact are often friends!), but that isnā€™t a good justification to behave that way especially given that the discussions on Discourse are public and, as you say, itā€™s difficult for anyone who hasnā€™t been around for a long time to parse the history between longstanding developers.

Completely agree. This seems to be a problem in every technical community that Iā€™ve ever participated in and itā€™s very unfortunate.

I havenā€™t participated in either of those communities but Iā€™ve also heard good things!

This is a good question. brms is a bit of a tricky case since itā€™s easy enough to use that it will get a lot of users but allows for many quite advanced models (compared to rstanarm, for example), and that particular combination of ease of use but advanced models will inevitably lead to a lot of questions on the forum (as weā€™re already seeing). Some of those questions would definitely make sense to be in the novice category but others probably not. So Iā€™m not sure how to handle that but Iā€™m glad youā€™re thinking about it and soliciting feedback!

It probably is hard to influence but I think the rest of us can help out with that. For example, there are times Iā€™ve noticed an aggressive discussion but havenā€™t said anything. I can try to say something instead!

Anyway, thanks for the thoughtful post @martinmodrak. Looking forward to having you in this role!

2 Likes
19

Perhaps instead of specifying the length, we encourage users to think about tone and content? i.e. is it helpful, is it kind, is it necessary etc.

The Finnish way might be more direct than others, but in my experience the tone is always very kind.

Edit to add: ā€œgreat question/hope that helpsā€ suggestion may have filtered from a suggestion I made once - I used to use this a lot with undergrads. It is an easy trick that helps me with my tone and fits with my writing style, but I donā€™t think that it would suit everyone.

2 Likes
20

Hi, Martin. Thanks for taking on this bg job! Here are my quick thoughts, in no particular order:

  1. Thereā€™s so much to be done here that Iā€™m pretty sure that anything you (Martin) do will be helpful.

  2. It seems we have many proposed codes of conduct floating around. Thatā€™s fine. Good to have lots of ideas out there; then we can settle on something.

  3. A key problem weā€™ve been having with Discourse is that it serves so many different functions, including:
    (a) Help for people who canā€™t get Stan working,
    (b) Help for users at all levels regarding Bayesian modeling and Stan coding,
    Ā© Discussions of new research ideas,
    (d) Discussions of plans or proposals for Stan improvements,
    (e) Meta-discussions (such as this one),
    (f) Announcements of job postings, new releases, etc.

Regarding item 3, Discourse has many categories but they donā€™t always map cleanly into what people might post on. This is important because Discourse relies on the free labor of all of us who reply to peopleā€™s questions.

One thing weā€™re planning to add, separate from but related to Discourse, is an announcements page which will include job postings, new Stan releases, and other important news. This will have a single web location (I guess it will have the form of a commentless blog, with more recent results at the top) so that important new announcements will be accessible without sliding off the front page as they now do in Discourse.

Anyway, my main message is thanks and encouragement to Martin!

1 Like