Thanks to everyone for the constructive feedback on Monday’s microconferences post. I wanted to take a beat to engage with two salient themes in the replies: runbooks and homophily. I’ll start with runbooks today, and tackle homophily next week.
No two microconferences are alike, and we shouldn’t impose hard-and-fast rules on their structure. One of the fun things about small conferences is you can tailor them to particular goals and dreams. And there are so many ways to do this well.
However, I think it will be useful to compile a runbook of agreements, strategies, and rules that you can use to help modularly assemble your ideal microconference. Today, I’ll run down a bunch of disorganized examples. You tell me your favorite ideas in the comments. I’ll assemble these more coherently into a document that I’ll widely share.
We can learn a lot from existing institutions, and many people spoke fondly of places I should have shouted out in the first post, e.g., BIRS, Oberwolfach, Dagstuhl.1 I also adore the quirkiness of the American Institute of Mathematics, which has a very particular and very fun way to run a microconference, disallowing canned and prepared talks.2 I am inspired by unconferences that bring together dozens of people to spontaneously generate many microconferences. All of their best practices should be part of the runbook.
I’ve found an easy model for microconferences is a bundle of short, 5-10 minute talks with adjoined group discussions led by the speakers. The past four microconferences I’ve attended have run this way. Short talks are fun because they force people to think hard about messaging and concision, and lead to a lot of interesting back-and-forth with the right serendipitous assignments. I was somewhat randomly assigned to a panel with Dan Wang and Dan Davies on cybernetics, and it was probably the most fun and rewarding 90 minutes I’ve ever had at a conference. That single brief session reshaped the narrative arc of The Irrational Decision.
Johan Ugander raised several good ideas in his comment. About archiving and proceedings, he wrote: “The non-proceedings nature of workshops is key to drawing in a diverse set of people.” I agree. We should think broadly about what should count as “proceedings” or “archiving.” I liked Johan’s suggestion of simply inviting participants to submit to a special issue. You could consider this series of blogs ([1], [2], [3], [4]) and this youtube video the “proceedings” of the Cultural AI workshop organized by Leif Weatherby and Tyler Shoemaker this spring. I just want to suggest that we explore creative ways to archive, evaluate, and credit microconferences in the broader academic ecosystem. I’d add to my list of core microconference values that “archiving should encourage, not discourage dissemination and broader engagement.”
As Johan wrote:
“ACM EC has a forward-to-journal mechanism, which gets Computer Science, Operations Research, and Economics folks together for a coherent conference but lets them still go harvest their respective tokens. The EC reviews get passed to the journal. And EC also organizes a ‘Highlights beyond EC’ session, which lets people request to present work recently/already published “elsewhere” but relevant to the community. Both of these mechanisms help keep the ‘conference’ and ‘publication’ goals separate.”
Endorsed!
Anna Gilbert raised the idea of “bump sessions”:
“The Dagstuhl and Oberwolfach type conferences in TCS used to also have “bump sessions” (maybe they were called rump!) where people proposed open problems, noodled over difficulties, etc. Sometimes these micro workshops even “published” open problems from these bump sessions. They were quite useful and engaging!”
Anna also raised one of my favorite ideas, which I’m looking for an opportunity to try:
“Another model I’d like to advocate for is to have members of a PC present the papers they selected (see my rant about paper reviewing and big conferences) and then an audience discussion. Like what the statisticians do but in person and with a publication resulting for the authors.”
I call this the “Not-so-royal Society.” The conference or session would center on a single paper written before the conference is organized. The organizers invite discussants to compose a response/review of this paper. The meeting could start with a presentation of the paper by the author. Many papers have multiple authors, and the presentation can be as collaborative as the writing. The author’s presentation is followed by commentary from the discussants. The remainder of the session is a conversation that invites commentary from the rest of the workshop participants.
After the workshop, the author can revise the paper, the discussants can write formal commentaries, and the author can write a rejoinder if they choose. All of this writing could be posted to arXiv as refereed conference proceedings and linked from the microconference proceedings webpage. This format might be the most legible in the current regime of bean counting and could perhaps serve as a “gentle introduction” to the microconference format. If you have ideas of papers we should use as testbeds for this format, reach out! I’d love to help set something like this up.
Finally, I want to give a shout out to Henry Farrell and Cosma Shalizi, who have been experimenting with clever microconference formats for years. Cosma wrote up the idea of having people present others’ work. Everyone writes a talk or slides, but then the presentation is assigned to someone else. They found that doing this at the start of the workshop creates a shared comprehension, allowing for a lot of multidisciplinary crosstalk. And they also found that speakers had to work extra hard to make their points comprehensible.
This is by no means an exhaustive list of ideas. And it shouldn’t be. My next major to-do is posting a working document of this runbook somewhere. But before I jump in and commit myself to a design, I’m looking for pointers to good online institutional memories that allow for edits and revisions. I love the minimalism of PMLR and bactra.org, but I also want to host a living mission statement and runbook on the same site. Let me know what platforms I should look into. Tech tips would be most appreciated!
You can get a sense of the skew of my readership by the institutions they love.
I also love that it used to be in the backrooms of a Fry’s Electronics store
The lead time for workshops at places like IPAM, ICERM, etc. used to be *years* but they (well, ICERM) instituted some "hot topics" workshops that had shorter turn around times. And, the NSF used to occasionally include funding for local workshops on larger grants (like the CCF Medium). I have no idea if any of that stuff still exists.
What I really want is an all bot portion of one of these big conferences like ICML or NeurIPS (bot A writes paper, bot B referees, bots C - G form program committee, bot H delivers talk/video/poster) and an all human artisanal part in which humans write, review, and present (but the PC has to do the paper presentation). Everyone gets a line on their CV but the humans have fun.
I've been fairly critical of workshops at robotics conferences moving towards miniconference formats where many participants are focused more on self-promotion than on having an open, collaborative format to really push forward the science.
Here's some of my critiques and previously proposed solutions in the context of your proposal.
1) The biggest issue has been popular, "famous" speakers attending multiple workshops and not sticking around to actually engage with other talks and student work. RSS has decreased this issue by limiting folks to a maximum of 2 workshops, but 1 workshop still typically loses out under this rule.
Having these microconferences be essentially standalone would help fix this. If several related microconferences were co-located, some strong encouragement to engage with the entire meeting / strict rule to not present at two of them would still need to be enforced to address this.
2) The friend / cronyism / popularity contest issue is still a big issue.. Attempts to ensure geographic, gender, and career-seniority diversity have worked somewhat, but still it seems like the same folks get asked a lot and others doing good work are ignored. Given the self-promotion angle that many people take to these meetings, we suggested a few years ago that workshop organizers ask the senior person (faculty, industry researcher) to nominate students or postdocs to present in their stead or jointly with them to help these folks get more attention and credit for the work they're doing. I'm happy to see this suggestion has persisted somewhat in subsequent years and I've attended several joint faculty-student talks in the past couple years.
3) Along these lines, I would advocate that for many microconferences / workshops the topic call should be given before announcing / confirming any speakers (half the speakers?) and let those interested apply and be given equal time to invited talks, instead of the current standard of submitted short papers that get relatively short presentations relative to the big invited names.
4) The most batshit insane thing that's been taking over recently is people with papers in the main conference submitting essentially the exact same work to be presented at the workshop in order to get more attention, especially if the main conference presentation is just a poster or lightning talk. The most upsetting version of this was someone not even bothering to show up for their short workshop presentation and letting a video play even when they had presented at the main conference. I hear this is also very prominent at CVPR. Here again, not collocating with bigger conferences would help.
5) One big benefit of collocation is of course broader participation, especially by students.