We received 116 abstract submissions on EasyChair for AMP 2020. Of these submissions, we accepted:
Our overall acceptance rate was 66% (= 76/116), or 14% when considering oral presentations alone (= 16/116). Two posters were withdrawn before the conference began, but otherwise all accepted talks and posters were ultimately presented at AMP 2020.
Apart from accepted oral presentations and posters, we had three invited plenary speakers (Anne-Michelle Tessier, Juliet Stanton, and Karthik Durvasula), along with three invited presentations at our tutorial on intonation (Jon Barnes, Emily Elfner & Anja Arnhold, and Naomi Tachikawa Shapiro & Arto Anttila).
There are well-known demographic disparities in academia, and in linguistics specifically, concerning graduate education, hiring, conference presentations, publication rates, invitations to present research, and so on (see e.g. the LSA's 2019 Annual Report on the State of Linguistics in Higher Education on some of these issues). For that reason, we wanted to be transparent about how well the demographics of authors in our submission pool matched the demographics in our accepted pool of talks and posters, particularly as regards self-identified gender and ethnicity.
Our primary interest was whether minoritized groups were as well-represented in our accepted pool of talks and posters as they were in our pool of submissions. For our purposes, 'author belonging to a minoritized group' was defined as any author self-identifying as non-white, and/or self-identifying with a gender other than cis male. We chose these definitions because they correspond fairly well to demographic categories which have traditionally been underrepresented in academia, including the field of linguistics. We recognize that these are rather coarse ways of thinking about demographics and equity, but felt that we needed to rely on these imperfect categories in order to begin assessing issues of equal representation in our submission and acceptance pools.
We collected demographic information for each author by visiting research websites (when possible) and/or drawing on information known to us from our personal and professional relationships with authors. In the case of multiple authorship, we collected demographic data for first and/or equal authors. As a result of these data collection methods, our information on author demographics likely suffers from inaccurate and missing data. With that caveat in mind, here are the demographic statistics of our submitted and accepted abstract pools:
The author demographics of accepted presentations thus tracked the author demographics of submitted abstracts quite closely, even without any post-hoc tinkering on our part. However, we note that the proportion of submitted abstracts from authors self-identifying as non-white and/or not cis male was clearly below 50% (at least in our estimation). This suggests to us that early outreach is needed to encourage future AMP abstract submissions from researchers who self-identify as belonging to these categories.
We would like to acknowledge again that these demographic statistics are a rough and limited way to address more complex and far-reaching issues related to diversity, equity, and inclusion in our field. For some related statistics on conference participation (particularly question-asking), we encourage readers to visit Claire Moore-Cantwell's reports on question-asking at AMP 2016, AMP 2017, and other conferences.
There were about 550 unique registered participants for AMP 2020. These registrants came from at least 47 different countries (including every continent except Antarctica) and 34 different US states. These countries are indicated in red in the following map, and US states with registrants are in purple.
We also collected information on the job description of registrants. These numbers are only approximate, as several categories of self-description overlap, and some participants accidentally registered multiple times:
283 distinct participants attended oral presentations at AMP 2020 over the three days of the conference (210 distinct participants on Day 1, 182 on Day 2, and 155 on Day 3, with overlap between the three days). The maximum number of simultaneous participants was about 115, with attendance averaging about 80-90 people for accepted oral presentations and 100+ for invited oral presentations and tutorials.
Unfortunately we do not have attendance information for the poster sessions (or associated videos and office hours) because the platform that we used to host these sessions (Padlet) does not keep track of such information.
Oral presentations were recorded using Zoom's native recording function, and posted as unlisted videos on a YouTube channel associated with the conference. These videos were shared with registered conference participants via a password-protected list of links.
At the point when these videos were taken down (about one week after end of the conference) these videos had been viewed over 600 times in total. This suggests that registered participants were indeed interested in asynchronous presentations of talks at AMP 2020, either along with or instead of live participation.
Unfortunately we do not have post-conference viewing statistics for the poster sessions or associated videos because the platform that we used to host these sessions (Padlet) does not keep track of such information. Posters were also kept up for about one week following the end of the conference.
We are grateful to the organizers of the 2020 CUNY Sentence Processing conference at UMass Amherst, who wrote up a very useful, and detailed overview of their own experience organizing a virtual conference on Zoom (https://blogs.umass.edu/cuny2020/blog-organizing-cuny/). We relied on their advice, and repurposed some of their organizational materials, in planning for AMP 2020.
With respect to scheduling, our main concession to the virtual format was to eliminate the traditional lunch break, and add a large number of regular short breaks after each talk. The hope was that such small breaks would reduce Zoom fatigue. Informal feedback from conference participants suggests that having many short breaks did indeed reduce Zoom-related burnout.
We opted to eliminate the lunch break because (1) doing so would free up more time for short breaks; (2) participants came from many different timezones, and so having a dedicated break at, say, 12:00-1:30pm would not necessarily coincide with mealtime for most participants; and (3) we assumed that participants would be moving in-and-out of the Zoom room anyway in order to take breaks, eat food, teach classes, etc.
Our attendance statistics suggest that we were correct about point (3): attendance at any individual talk was missing about 30-60% of the total participants for that same day.
We used the Zoom webinar platform to host oral presentations for the conference. This platform had some advantages:
The Zoom webinar platform also had some (potential) disadvantages:
However, some of the organizers thought that the lack of information about the names and numbers of participants in each session was in fact a positive aspect of the webinar platform. (This opinion was echoed by some participants, as can be seen in the survey results presented below.) Our rationale was that showing participant counts could contribute to a competitive atmosphere, and/or discourage presenters who had relatively reduced attendance (e.g. because attendance was overall lower on Sunday, as it often is at physical conferences). These struck us as sufficiently negative possibilities that we opted to turn off the visibility of the participant count to attendees.
We hosted poster sessions on Padlet, using a $10/month Pro license which allowed users to upload files greater than 250mb in size. This was important, as we asked poster presenters to upload a 5 minute video or audio recording to walk viewers through their posters, substituting for the traditional poster spiel.
In general, participants seemed to feel positively about Padlet as a platform for the poster sessions (see survey results below). Presenters commented on the relative ease of uploading their materials, and viewers seemed to find the platform relatively easy to use. Participants also commented positively about the pre-recorded video walk-throughs, as well as the virtual "office hours" that poster presenters hosted during or after their poster session.
There were also some downsides about using Padlet. It was impossible to tell how many people attended each poster session (or each individual poster), either in real time or after the fact. It also turned out to be relatively easy for presenters to upload their materials to the wrong portion of the site, and/or accidentally swap their materials with materials uploaded by other users via drag-and-drop manipulation of the site.
Managing a virtual conference on Zoom is a very active task. AMP 2020 organizers spent a lot of time over the course of the conference responding to various kinds of technical support requests, uploading documents on Padlet, and uploading recordings of oral presentations to our conference YouTube channel. This was all on top of session management: coordinating A/V needs of presenters, recording presentations, giving presenters warnings about remaining time, moderating the Q&A interface, calling on question-askers, unmuting people, etc. Some of us found it quite difficult to focus on the talks themselves while dealing with so many organizational and logistic issues during the live conference.
Some of these issues would be mitigated if talks were recorded ahead of time rather than presented live, whether or not the talks included a live Q&A component as well.
We found it very useful to have a group WhatsApp thread so that we could text each other in real time during the conference. This allowed us to remind each other of tasks that needed to be done, and to quickly coordinate solutions to problems as they came up.
Following the end of AMP 2020 we sent out a short questionnaire to registered conference participants. With 61 respondents, there were a diversity of opinions about the questions we asked: