Due to the short notice lockdown, the organization was suddenly reduced to very few people, with limited administrative support. Nevertheless, we stuck to the foreseen dates, and with only one month to carry out an entire transition to the online format, it was necessary to make many bold decisions.
With an online event as the only alternative to uphold ECIR 2020, we repurposed our goals and priorities now considering this perspective. In compensation for the lack of physical interaction, we saw an opportunity to maximize the visibility, participation and impact of the conference by leveraging the lack of physical barriers. In order to further amplify this potential, we decided to make the conference and proceedings open for non-authors.
Below, we list the key aspects that we believe to have made or will make a difference.
Open access. A decision that was made after the conference moved online was to make its scientific outputs more visible and accessible. Making available the paper presentations was a first decision. The second decision, more elaborate, concerned Springer’s Open Access proceedings. Moving to open access was well received by the community and we hope that it will bring a long-lasting positive effect on ECIR.
Timezones. After the lockdown, the program was redesigned to accommodate the speakers’ time zones as well as possible. Papers with authors from distant Asia/Pacific time zones were allocated in the morning, and papers affiliated in the Americas were scheduled in the afternoon — while papers from nearing time zones were unconstrained. The keynotes and the panel were moved to the afternoon to maximize the time zone range compatibility for live viewing: from USA West coast (8am) to China (9pm). Sessions were recorded and available on YouTube, so that everyone had eventually a chance to follow the whole conference.
Zoom Webinar format. The choice of Zoom webinars proved to be a good decision — it was safer (no Zoom attacks were possible) and allowed for a better management of speakers and audience. It might be interesting, in the future, to consider an option for attendees to see each other somehow, though this may be challenging for large audiences, and is not supported by Zoom Webinars (only Zoom meetings support this).
Rehearsals with the community. Rehearsals with authors, session chairs and keynote speakers turned out to be essential to engage the core of the community. Everyone had a chance to test in advance with the organizers the Zoom features as they would be available in the session where they would speak or act as chair. Detailed instructions were additionally outlined and shared with participants for each different role (chair, speaker, attendee). This preparation made people feel more confident, comfortable, and committed — it also allowed everyone to debug their environment/posture (this was particularly well received by keynote speakers).
YouTube Live streams. YouTube Live was a great idea. In the questionnaire, there were several answers praising this option as attendees could easily watch the conference in decent hours in their corresponding time zone. People who usually do not participate with questions or other interactions in Zoom, preferred YouTube because they can go back and listen to parts they missed or did not understand. We had close to 100 viewers on YouTube on a regular basis, and the daily peak was always the keynote.
Slack. Slack was good for offline discussions and community building. We had a total of 251 registered users, and 153 users posted a total of 4340 messages — the graph below shows the Slack activity during the conference. The availability of a chat channel within Zoom during the technical sessions — in addition to the written Q&A Zoom feature — may perhaps have diverted some conversations as well. Adding this to Twitter and other social interaction media may create some duplicity of channels to some attendees. While all channels were active, there is for sure plenty of space to improve the offline interaction.
Full papers as live presentations. A key decision we made was that full paper authors would present live, rather than with a pre-recorded video. Full papers were thus all presented live, and we had a peak of 157 attendees following the sessions live. All sessions had plenty of questions, although attendees seemed more inclined to participate with written questions — using the Q&A Zoom feature — that had to be read by the session chair, rather than using their microphone to ask and interact with speakers, which Zoom webinars also support.
Short papers and demos as pre-recorded presentations. Pre-recorded presentations were played in a Zoom meeting where authors were given time to discuss their work. The peak attendance was below 100 on the first day and below 19 on the last day — maybe people stopped showing up because of the “canned” format or because it was competing with other two live presentation streams. Structuring a more detailed video schedule might help enhance the engagement in these tracks, which remains a point for future thought and exploration.
Pre-recorded vs. live presentations. The previous point does not mean that pre-recorded videos do not work — it just means that they did not work in the setup we designed. The number of viewers (and feedback) we had on the YouTube Live (recorded) stream may hint that pre-recorded videos may be used in some other way. Also, about half of papers had European affiliation, and this helped arrange a live schedule that was compatible with all the speakers’ time zones. Scaling a fully live approach to larger and more internationally-spread conferences might be more challenging.
Poster session replacement. We still do not have a clear solution for the presentation of posters and short papers. Perhaps one or two minute spotlight presentations followed by other means of interaction (e.g. on Zoom or Slack) might work.
Social event. The social event was planned as a virtual reality game and a Zoom meeting with everyone’s video turned on and with Mozilla Hubs. While Hubs did not attract many people, the social event on Zoom was a very nice experience. People actually engaged in the Zoom meeting, shared music, pictures of food and were all having nice conversations over a glass of wine/beer/juice, but, at a very long distance. Maybe because of the lockdown, the social event on Zoom showed that attendees were eager for more opportunities like this to engage with their usual ECIR friends. We should have had more events like this throughout the conference days.
Email. Regular and effective communication with participants, at this scale, is not easy. While we hired an email marketing service, there were still several hundreds of people who did not receive emails. Hence, although we had planned to send newsletters to participants in advance and during the conference, we had to give up on this idea because many people were being left out.
Sponsors. Sponsors required additional effort to be accommodated in the online format. We encouraged sponsors to go to Slack and engage with attendees. Because of the change in format, we had to make some changes to accommodate publicity expectations. This revealed something that is obvious now, but we had not thought about it: broadcasting time is the most valuable asset we have to offer sponsors in the online conference format.
2 responses to “ECIR 2020 – Delivering a virtual conference”
[…] analysis of the participants. Then follows their reflections of the challenges and successes of the virtual event. Udo Kruschwitz reports on the Doctorial Symposium at ECIR 2020, I have contributed notes on two of […]
[…] already had experience of the European Conference on Information Retrieval 2020 as a virtual event in April, a large scale Microsoft Research Event in August and am involved in […]