ICER attendees introduce themselves in #general (names blurred).

ACM ICER 2020 trip report: virtual serendipity

This post summarizes my personal experience helping to plan, run, and participate in the ICER 2020 conference. In case it’s not clear form context, this is a personal trip report and does not reflect the views of the other organizers or the SIGCSE board. But I’m sure they’d be happy to share their experiences too!

I’ve been attending the ACM International Computing Education Research conference since 2013, when it was in San Diego. I’d published there before, but San Diego was the first time I could carve out a week in August to network with computing education researchers. While the conference was small—perhaps just 70 or 80 people—it was riveting. Small group conversations were the norm, critical debates were constructive, and there was a strong sense that everyone in the room was oriented toward the same central goal: effective learning for every person learning about computing. I’ve gone ever since (and since I starting writing trip reports, blogging about my experiences: 2016, 2017, 2018, 2019).

This year was different for multiple reasons. First, I was the “junior” program co-chair, which meant that I was helping to run our peer review process, under the steady and thoughtful leadership of Anthony Robins, and junior and senior site chairs, Renée McCauley and Adon Moskal. The four of us started meeting shortly after ICER 2019, jumping right in to finalizing the organizing and program committees, finalizing contracts and budgets for our site in New Zealand, and bracing ourselves for the inevitable growth in submissions that the conference has seen for years. I also came with a whole list of changes I wanted to the see to the conference, including more transparency in communication, forming a steering committee, broadening PC expertise, expanding page length limits, requiring reviewer training, creating reporting standards, adopting a new web platform, and having more neurodiverse inclusive session chairing practices. I was so excited that the team was open to all of them, and that we successfully implemented all of them.

And then, there was a pandemic. Some things didn’t need to change at all. Our reviewer process was already distributed across the world, though reviewers had less time to give to reviewing. But everything else — venue, budget, communication, sessions, presentations — had to change. We set out to read about other’s online experiences, participated in online conferences ourselves, and brainstorm conference designs that might express the values of the ICER conference, including rich engagement about papers, and community building. We eventually specified a virtual conference plan, circulated it online for feedback, and started implementing.

Our design had several key principles:

• Prioritize community building over consumption. We started from the long held value at ICER that interaction, and relationship building are far more important than consuming presentations. This meant promoting discourse and engaging groups in conversation about the published work, and deemphasizing presentations.
• Prevent fatigue. Pretty quickly after the lockdown in March, it was clear that few people can sustain more than 40 minutes of a video chat without needing a break. That meant that talks needed to be short, breaks needed to be abundant, and there needed to be many ways to escape, while still staying engaged.
• Support serendipity. We knew from several years of conferencing that what makes academic conferences special is unexpected encounters. We wanted to maximize this, and that meant minimizing scripted activities and creating abundant opportunities to mix random groups of people together, much like the round tables used at all previous ICERs.
• Don’t trust the internet. We knew from months of working remotely that it would not be reasonable to expect live talks to work consistently. Speakers’ connections would be unreliable, audio would be misconfigured, speakers would go over time. It must be possible to fall back to asynchronous chat, and pre-recorded content.

The above principles led us to choose a four day, 3 hours per day model that would allow most of our attendees, who span North America, Europe, and Australasia, to participate between 7 am and midnight in their time zones. Each day was four 40 minute sessions and three 20 minute breaks. In those sessions we placed keynotes, single track paper sessions, parallel track paper discussions, lightning talk sessions, and poster sessions. Importantly, all of the talks were pre-recorded and 12 minutes, and all talks had abundant backchannels for discussing the presentation. This made the conference more like a movie screening than a performance. Additionally, we wanted to have abundant places for topical discussion and just in time birds of a feather sessions, with ample opportunity for emergent groups. To fit these new models, we chose Zoom webinars for a few single track sessions that needed to accomodate 300+ people (keynotes, award sessions, and some papers talks), and importantly, Discord for everything else, especially for its support for seamlessly hopping from one video chat to another and for seeing who is in each chat before joining.

We had no idea how well this would work. Would Discord scale? Would people engage in meaningful conversations in small groups? Would people actually attend? And would they want to attend again (which was especially important to me, as the ICER 2021 senior program chair)? This week we found out.

I reunited from friends in Breakroom G (faces blurred for privacy).

Day 1: Keynote, breaks, and sessions

When I first woke up on Day 1, about 7 hours before the conference started, activity on Discord was already starting. Attendees were chiming in on the #general channel, introducing themselves, and other attendees were reciprocating with waves, flags, and other emoji reactions. I watched presenters go into presentation rooms and test their A/V. I saw colleagues meeting in break rooms, catching up. Others were messaging in #help-desk, trying to join the newcomer welcome last minute. I nagged a few people to set their nicknames to their real names, and assigned presenter roles to people who’d only just joined the Discord server. In other words, it was like any first day of conference: printing last minute badges, welcoming people at the registration desk, and activity throughout the conference center, just in Discord instead of a hotel. By the beginning of the conference at 1 pm Pacific, we were at 317 registered attendees and 290 on the Discord server. Community took the reigns, creating all kinds of channels to support the topics they wanted to discuss.

The day started off with the newcomer welcome. When participants registered, they were able to say whether they identified as a regular and were willing to meet newcomers, and whether they identified as a newcomer and wanted to meet regulars. Based on this, we made a spreadsheet, matching newcomers and regulars, and assigned them Discord breakrooms to meet in 30 minutes before the conference opening. I didn’t join, since I was busy with conference logistics, but I saw more than a hundred meeting in Discord.

Our opening session set up the dynamics for the conference, and then we dove in to a short break before our keynote. I had just enough time to jump into a break room and catch up with some old friends and meet some newcomers. We did introductions, met someone’s cat, and then jumped over to the keynote.

Tim giving his keynote, outside, in New Zealand, where there was no lock down.

Tim’s pre-recorded keynote was a lot of fun. He filmed the entirety of it around New Zealand, which was especially nice since we all missed the opportunity to travel there. Tim walked through the numerous principles underlying his vision for CS Unplugged, which he viewed less as a way of promoting transfer, and more as a way of helping teachers have a gentle ramp to appreciating algorithms. While the presentation played, attendees chatted in Discord about the talk and posted questions in Zoom; the session chair that asked some remaining questions live with Tim.

I catch up with some friends, colleagues, and newcomers at a break.

After the keynote was a break. I had a great chat catching up with past collaborators, Ph.D. students who had participated when I chaired the doctoral consortium in past years and are now faculty, and also meeting some new people.

After the break, we resumed in Zoom, with two papers in the single track session. I was the session chair for both, which was a new and challenging task: I was listening to the presentation for the first time, monitoring Discord discussion for questions, and also monitoring questions in Zoom. The first presentation was led by Lauren Margulieux, and was a talk about metacognition in programming, pulling together an immense amount of literature in and outside of computing education. The paper found that we draw upon a lot of theory, we don’t use it very consistently, but that there are an increasing number of powerful interventions for promoting self-regulation if timed well.

Lauren’s pre-recorded talk.

The second talk was An Analysis of Use-Modify-Create Pedagogical Approach’s Success in Balancing Structure and Student Agency, presented by Merijke Coenraad. They found that an intervention that provides starter projects had a substantial impact on what learning students engaged in and how much of a lesson they completed, but that this was largely explained by who was teaching.

Merijke’s pre-recorded talk.

A busy break room.

The breaks continued to be popular after the paper session. Groups in topical discussion channels self-organized and chose breakout rooms to meet in. One group excited about K-12 education and teacher preparation gathered in Room K, others organized around broadening participation, and others still around shared institutions and experiences.

The last session of the day was a parallel track session. In these, there were five papers being discussed in parallel, each at 5 different tables. The format was more like a viewing party than a presentation: we instructed everyone to gather at a table, nominate a session chair, stream the pre-recorded presentation in parallel, and then discuss. While this happened, paper authors would bounce between the five tables and participate in the discussion and answer questions.

I played the role of an author in this session, for the paper I co-authored with Ph.D. Yim Register:

Yim’s presentation on learning machine learning with personal data.

Yim did an amazing job presenting a clear and well-paced explanation of their work. After my group of 6 streamed it, they asked a series of great questions about the implications of our work, from teaching machine learning development, to teaching more conceptual explanations of the risks and benefits of machine learning. Yim was good about moving between rooms; I forgot, so they came and tagged me to go to another room, where I had another great large group discussion.Given the loose structure of this session, and the lack of an appointed session, I was worried about how many of these sessions just falling apart, but based on feedback, it sounded like most worked pretty well.

Day 2 started with a lightning talk session in Zoom.

Day 2: Lightning talks, papers, posters, oh my

The second day of the conference started off busy again. The community surfaced needs for new affinity groups and topical discussions in Discord, and we were holding a Europe social for people to network with colleagues in Europe at a more reasonable hour. There were up to 16 people at the gathering in our Great Hall break room, mostly people in Europe, some from elsewhere.

The beginning of the bulk of Day 2 was the lightning talk session. We held this in a Zoom webinar, with the track chair steering a single slide deck, and 6 lightning talk speakers giving three minutes on their slide. Six presenters talked for three minutes each, and then the chairs fielded a few questions, extracted from the Discord discussion on the talks and the Zoom Q&A. The logistics worked reasonably well, and the content felt much like other past ICER lightning talk sessions: a medley of interesting preliminary work and sharing of provocative ideas.

Immediately after the lightning talk session was the single track research paper session of the day. The first paper was a fascinating paper on CS education in South Asia, including Bangladesh, Nepal, Pakistan, and Sri Lanka. The paper did a great job setting up the intricate complexities of each country, the gaps in infrastructure in internet, teacher preparation, and and funding. They modified an established instrument to survey teachers and adapted it for cultural differences and deployed it. They found that most of their classes were boys, and most students were viewed as “gifted”, that most of the teachers viewed the their critical resources as computers and question banks, and that many classes had students program in text-based languages without computers. The survey was just scratching the surface at experiences in this country, but took a bold move toward more global perspectives on CS education.

The second talk in the single track session was a study of the major differences in students who use a TIPP&SEE protocol (read the Title, read the Instructions, understand the Purpose of the activity, then Play, and SEE refers to Sprites, Events, and Explore in Scratch). They analyzed several individual behaviors, but couldn’t uncover any observable differences in student behavior, suggesting that a lot of the performance differences had to do with supporting self-regulated learning.

For the parallel session, I chose to discuss the paper Collaborative Learning in Computing Education: Faculty Perspectives and Practices, by Sharon Mason at Rochester Institute of Technology. She examined an immense amount of course design materials from seven instructors. All seven modeled social constructivism, giving hints, coaching, guiding, facilitating, and modeling behavior. But collaboration policy varied; three required collaboration and four disallowed it. The paper surfaced challenging tensions between cheating and collaboration.

I joined a small group to discuss Sharon Mason’s paper on teacher perspectives in collaboration (faces blurred).

During the break, I joined a just-in-time birds of a feather on assessments and met about fifteen people who are interested in the challenges of measuring student learning and the tensions between measurement, collaboration, and support. There was such an interesting diversity of perspectives in the room: students sharing their experiences, teachers discussing the tensions, and researchers in educational measurement, all trying to find a path through challenging constraints.

The birds of a feather session was popular, bringing together scholars from across the world (faces blurred).

The last session of the day was a poster session, with 17 presenters. We set it up in Discord with a “Posters” category and 17 distinct voice channels, each allowing the presenter to stream their content and for attendees to freely mingle between channels. Aside from some challenges with streaming, and some confusion about where to find poster PDFs on the program, it actually worked quite well. I bounced between 5 different posters, lurked in some, asked some presenters questions, all the while being able to see the same distribution of attention as in a poster session, with some very popular posters attracting a couple dozen, and others with authors with no one to talk to. In some ways, it was better than a real poster session, because when I entered a channel, I wasn’t in a loud room full of distractions and food, just the authors and other interested people.

The first Discord poster session, showing me at Thuc Nhi’s poster, and >150 other attendees at other posters.

I chaired the closing session in Zoom, where we had about a third of the attendees, and shared some highlights from the day: we were up to 319 registrations, we had a pleasant mixer with attendees from Europe, the community had created new channels to organize and meet, and we had an exciting Day 3 coming, with doctoral consortium speakers, more papers, and an Asia social.

Jean presents a perfectly polished pitch.

Day 3: Doctoral consortium students shine

The kickoff for the third day began with 17 doctoral students who participated in the doctoral consortium speaking for a minute about their work, and enticing people to come to the poster session later in the day. One of the doctoral consortium co-chairs, Lauren Margulieux, led the session, which ended up being a rapid-fire tour through the broad scope of computing education research, and an excepting glimpse into the future leaders of the computing education research community. The students did an amazing job compressing their research plans in just 1 minute each. Bravo to the students, and for the chairs and mentors.

Phil Steinhort’s ICER talk.

Immediately after the DC lightning talks, we dove in to two fascinating talks on self-efficacy. The first paper was by Phil Steinhorst, titled Revisiting Self-Efficacy in Introductory Programming; he worked on designing a validated self-efficacy instrument for introductory programming courses. It was exemplary in its instrument design, conducting validation across multiple institutions, and a very granular factor analysis across institutions. It appears to be a wonderful tool for supporting research on self-efficacy in CS education; they intend to release it soon for the community to use.

The second talk was by Jaime Gorson, titled Why do CS1 Students Think They’re Bad at Programming? Unlike the first paper, which concerned programming self-efficacy, this paper tested the assumption about what affected programming self-efficacy. It specifically considered self-assessments at three universities, and which scenarios in which self-assessments happen. They provided short vignettes to students, then asked them about their agreement with self-efficacy statements. Many students demonstrated self-assessment when not knowing how to start or getting stuck, which they (incorrectly) perceived as something professional programmers do not experience. All of these findings were stable across their three universities.

At the break, I noted in the break room topics that I wanted feedback on ICER 2020, and hopped into break room A. I had a great discussion with several attendees about their experiences, and before I knew it, it was time for the parallel paper sessions.

My setup for parallel sessions, including the YouTube video, a pop out of the video chat, and our text channel (faces blurred).

I decided to learn about John Edward’s paper on the evolution of typing speeds of different natural language and programming language tokens. The work presentation revealed all kinds of promising differences, suggesting that pauses in typing might be related to cognitive load. Our group of ten had a wide ranging discussion about measurement, about the feasibility of using this as a class-wide signal for instructors to monitor learning over time, and about the difficulties of trying to do monitoring on an individual level, given the variation in individuals’ typing performance within a day and on different hardware.

During the break, I caught up with friends and colleagues Kathi Fisler and Diana Franklin. We talked about a lot of things, but mostly about the conference format, and the possibilities for 2021. Several attendees joined our discussion later, mostly those who never would have been able to attend if the conference hadn’t been virtual, and it became crystal clear who precisely a return to entirely in-person conferences would exclude.

The last session of the day was the doctoral consortium poster session. Just as with Day 2, Discord really shined here. All of the different poster channels were clear, and it was very easy to see which ones were more or less busy. This time, we posted links to everyone’s posters in a DC poster channel in case someone wanted to check a PDF more carefully. I made it to about five students’ posters in 40 minutes, and got to meet several wonderful doctoral students doing some amazing work on broadening participation computing.

After a brief closing session in Zoom, I bounced over to the “Great Hall”—a break room with no limit on the number of people who could join—to chair the Asia social. I didn’t expect a huge number of attendees, but we had several participants from Australia, for whom the conference was a bit early. We had a wonderful chat about what we were making for dinner/breakfast, how our countries were doing with COVID-19, and differences in our school systems.

Takeshi answers questions in the first session.

Day 4: Papers, awards, and closing

I kicked off the fourth and final day with some reminders about the structure of the day, a prompt to fill out our conference survey, and some tips on how to better manage the parallel sessions. It was a lot of fun telling asking people to fill out a survey and telling them I wouldn’t advance to the next slide until I started seeing responses. And it worked! By the beginning of the first research paper session, we had received 111 responses, from nearly all of the 145 people active on the Zoom call and Discord at the time of the prompt. A 30% response rate for the conference, and 75% of active attendees was a good start! By the end of the day, we had 134 responses.

The first paper in the single track was by Takeshi Watanabe, titled Analyzing Viscuit Programs Crafted by Kindergarten Children. Just this year, Japan made “programming education” mandatory for elementary schools. The authors discussed the use of a rewrite rule based “viscuit” program contributed back in 2003, which uses images, without the use of letters or numbers, making it suitable pre-school. They observed 28 children aged 5–6 in a kindergarten engaging in a series of thirteeen 40 minute lessons. The paper was really a feasibility assessment, showing that programming could express their ideas and stories in this paradigm.

The second paper was by Using Design Alternatives to Learn About Data Organizations, by Xingjian Gu, Max A. Heller, Stella Li, Yanyan Ren, Kathi Fisler, and Shriram Krishnamurthi, and presented by Shriram. The paper started from the premise that data representations have many alternatives, and it’s important for students to understand alternatives. The paper described formative studies that gave students pairs of data organizations and asked them to describe the differences and their preferences, then reflect on the pair. The results demonstrated that even students with no computing experience could identify all kinds of tradeoffs and limitations, suggesting the feasibility of the method in teaching.

The last talk I attended was by Felienne Hermans, for her paper titled Hedy: A Gradual Language for Programming Education. She set up a great premise around the cognitive load of programming language learning, including syntax, semantics, and even the very notion of learning a new notation. She contributed Hedy, a programming language that gradually adds syntactic and semantic complexity across multiple stages. The paper did a great job articulating the possibilities of the system, and presented some basic descriptive data about what people did with it, but our group raised all kinds of interesting strengths and weaknesses of this approach around the brittle knowledge it might create, the challenges for English language learners, but also the opportunities for using this for first introductions to programming.

For the last break, I put a call out in the break room topics channel to meet up with regulars I hadn’t seen yet. I got to catch up with lots of regular ICER attendees. We got to chat about the pandemic, our teaching, our pets, our kids, and a bit about our research.

We announce the Chairs’ Award.

The closing session was a lot of fun. We had the whole team of chairs reflecting on the experience, thanking the rest of the organizing committee, and Adrienne Decker, SIGCSE Board Chair, to talk about SIGCSE and upcoming conferences and award nominations. We ended with a bit about our plans for ICER 2021, and a promise to keep the Discord active in perpetuity.

Oh, I and I got to show off the John Henry Award hat, a wonderfully cute penguin cap to a lucky winner that will be announced after a few weeks of voting from registered attendees.

Days 1–4: The Sub-Conference

While there was a program, and it structured each day strongly, there was quite an active “sub” conference throughout the week in the topical channels and the breakrooms. Dozens of topics emerged, and self-organized into break room mixers and discussions:

VERIFY THESE

• The anti-racist CS education conversations discussed more inclusive language in CS teaching; our global community raised some fascinating points about etymology of words, and situated the conversation in broader efforts about anti-Black racism.
• The broadening participation channel discussed strategies for creating more inclusive classrooms, and reflected on how to analyze qualitative data about student identiy.
• The conferencing while parenting channel shared photos of people at home with their kids, strategies for coordinating participation with meal time, and tips for teaching while parenting.
• The LGBTQ+ channel, and associated private channel, built solidarity amongst LGBQTQ+ attendees like myself, and with allies.
• The primary/secondary K-12 channel connected scholars doing research and program administration for pre-service and in-service CS teachers from across the world.
• The assessment channel discussed strategies for reconciling tensions between cheating and collaboration, especially during remote teaching.
• The CS education in India channel brought together researchers in India, and researchers interested in studying India, forging new collaborations.
• The non-English CS education channel brought together attendees interested in non-Western dominant tools and perspectives on CS teaching.

For many, their experience was partly defined by these four day conversations and the video chats during breaks that emerged from them. Throughout the week, I saw people having informal small group meetings about collaborations, about new opportunities for partnership, and to bring together affinity groups for social video gatherings.

One of my students caught me in a thinking moment on Day 1.

Feedback

Engagement in the conference was high. I monitored peak Zoom and Discord activity in each session, including breaks:

Peak participation for each session across the four days.

As the plot shows, at any given time, there were 150–200 people active on Discord throughout the four days, with the keynote, breaks, and paper sessions getting the most attendance, and opening and closing sessions (which were light on content) getting the least. There was also a clear trend of slightly declining attendance throughout each day, although only modestly, and a modest drop from 200 to 150 by the end of the fourth day. We never expected all 319 registered attendees to be active for the entire conference, but the Discord data shows that everyone was active for some portion of the conference, and nearly two thirds were active for most of the conference.

We ran a conference survey, and I won’t share any preliminary results here because the SIGCSE board requested to keep them private. However, numerous attendees reached out to the chairs to express praise and critique. Some of the highlights included:

• Most attendees thought that Discord worked surprisingly well, and really appreciated how well it supported agency in movement (unlike other video chat platforms), and serendipity. So many attendees shared how many unexpected encounters they had, and how well Discord showed the immense activity, energy, and passion of the community, perhaps even more so than in-person conferences.
• For many attendees, the volume of activity on Discord was too intense. We made a tough choice to keep notifications on for every channel, which was the cause of this; the other option would have been to keep them off, which might have prevented discoverability.
• Many expressed a desire to maintain a virtual or hybrid option for 2021, as they would not have been able to participate otherwise.

Reflections

While I learned a lot about the state of computing education research as usual, it was hard not to be completely absorbed by the new experiences in our virtual conference experience. Here is just a sampling of things that I felt I learned in running this virtual conference:

• Different video chat room sizes have profoundly different social affordances. To do more than just introductions, it’s essential to keep rooms to 10 or smaller, preferably 5 to 7. We actually reduced them from 15 to 10 for these reason; I felt they could have been even smaller.
• Discord is a great platform for connecting people openly, but not for having private group conversations. We managed to create a few private group channels upon request, but we didn’t manage to teach the community how to have private 1 on 1 Discord video chats (which is possible, just not obvious). It’s optimized much more for public serendipity than privacy—which I suppose was our goal!
• There were some interesting and beneficial new social cues that aren’t possible in real life. For example, Discord has a list of all server members, and we had assigned people roles like Organizer, Session Chair, and Presenter, and so it was always possible to see who was online, what their name was, where they were from (because we required affiliations). It created a unique sense of presence, seeing an entire subset of my academic community in a list. Another cue was that Discord would show me what they were doing—which breakouts they were in, who they were talking to, which back-channeled conversations they were having. I felt much more aware of what the community was talking about because I could—with effort—monitor many conversations at once.
• We encouraged—but did not require—people to put pronouns in their server nicknames, and this had the unique effect of making pronouns highly visible in every interaction. This was especially important to our many non-binary attendees who use they/them pronouns; I heard many people in video chats using they/them correctly, or self-correcting, because pronouns weren’t ignorable, like they are on conference badges. We should have encouraged even more meta data: titles and web pages would have helped people learn more about other attendees.
• Many of the same conference patterns of behavior happened online too. Newcomers hung out with people they knew and felt safe around. Regulars wanted to reconnect. The conference was shorter in terms of time, and so we all probably saw fewer people, but we did reconnect.
• Because we had notifications for all channels on by default, the default experience for an attendee was having visibility into all conversations. That was good, because it promoted discoverability and serendipity, but it was also overwhelming at times. This is likely why so many fewer people posted on Twitter, because they were busy monitoring Discord instead.
• Over and over, I heard from attendees that never would have been able to attend if not for the virtual conference format. They were so incredibly grateful for the opportunity, and also expressed just how much more interactive and engaging this experience had been over other virtual conferences.
• Naturally, it was too easily to try (and fail) to multitask. I had the incentive to pay attention as an organizer, but many mentioned trying to watch a talk, but getting distracted by an email. It really requires someone to commit their time to fully engage (and many did).
• It’s so critical to establish norms for discussion. Many of the paper discussions needed protocols like when someone wanted to speak, when people were done watching the video, and when authors had entered the room. Because we lacked session chairs for these, the experience was highly variable, and often awkward. We had provided guidance, but many attendees didn’t seem to read it; session chairs are a way to ensure this, but we would have needed 100 of them to cover the 25 table conversations each day, which was infeasible. I’d recommend trying to find session chairs from registrants, even at this scale, and training them on how to facilitate a great discussion.
• Many past ICER conference have had unofficial backchannels, such as a Slack team, a Twitter hash tag, and email threads. Having an official backchannel was definitely more inclusive, as it allowed newcomers to discover conversations rather than have to find them through a privileged social network.
• We didn’t really plan the social hour times; they were unstructured, and informal. This wasn’t really a strength in an online setting; we should have planned events like games, more structured mixers, and other activities to encourage participation and networking. I think the more extraverted attendees managed fine, but more introverted attendees likely avoided these sessions.
• The 12 minute format for talks worked really well. It was enough time to cover a paper’s motivation, methods, and key discoveries, but no more. There was broad agreement that they weren’t too long, but also shouldn’t be longer. I can’t see going back to 20 minutes plus Q&A; it was enough to fill even 25 minutes of discussion and Q&A in the parallel sessions.
• One downside of capping the size of Discord rooms was that people wouldn’t fill them in equal balance. So one room would end up with 10 people and so others were often left out, or to start a new room by themselves. When this happens in person, it’s often noticeable, and so people will split off to balance tables, but people didn’t notice in Discord because there wasn’t a social cue that someone was left searching for a seat.

What does this mean for the future of conferences? I speculated on this in May before we’d finalized our plans, and argued from a more principled perspective that inclusion and climate change were more important than in-person networking. I think this experience showed me that meaningful community building is possible, at least in the way we ran it, but it is not at all the same as in-person encounters. In some ways, it is better—increased access, more opportunities for inclusive practices, more communication through backchannels, and a much lower carbon footprint are all huge wins in my book—and in some ways it is worse. There was little time for deeper conversations and no opportunities for social outings, in which to build deeper and more meaningful relationships. And the novelty of the format this year definitely detracted from attention on the content; I think a second year would reduce this distraction, but there will always be newcomers trying to learn how to engage.

What does this mean for ICER 2021? I won’t speak in an official capacity as the ICER 2021 general and program co-chair, but I will say this: I can’t see ever going back to a time where we exclude the 170 people who were only able to join this year because we were virtual. They have the same right to participate in our community that well-funded, unencumbered attendees do. I don’t see how it’s just to restore such a blunt structural barrier to the world to engage in our discourse.

For me, my favorite idea is to plan a virtual conference again, but have two types of in-person gatherings:

• Regional gatherings, where organizers find a space to bring together people who might attend by train or car, but not plane, and participate together online, then spend the remaining hours enjoying each others’ company, thinking more deeply about the research, and adventuring in their chosen city. And if it’s not safe to hold it due to a pandemic, the virtual conference is still there as a foundation.
• International gatherings, where organizers find a more exotic, desirable location to bring together those with the resources to invest fully in networking and community building. We might use the increased revenue from online attendance to equitably subsidize doctoral student participation, and participation from countries lacking regional critical mass. Imagine ICER at Hawaii, in a resort hotel, where one spends three hours a day online engaging the global community, and ten hours on the beach with a mix of senior people, rising doctoral students, and colleagues from South Asia, South America, Africa, China, and other regions. And as with the regional gatherings, virtual conference is always there, independent of sudden travel restrictions, lockdowns, family emergencies, and other crises.

To me, this vision is the best of both worlds: including everyone, while investigating equitably in deep relationships in person.