The main purpose of this blog is to share some lessons, experiences, and thoughts about planning an academic event (and building an academic community). The specific scope is on events in the AI and ML community (examples used are workshops co-located with major conferences) but the idea extends to any event of similar types and goals.
Before answering how to organize a good event, we first need to ask why communities and events (e.g. workshops) should exist at all?
Taking workshops at ML conferences as an example, it is a place to share ideas and engage in discussions within smaller subareas. It is not surprising that workshops become more attractive to researchers as ML conferences have grown to the size of over 10,000 attendees and have branched into many subareas these days. Even though the publication cycle in machine learning is already fast enough (2–3 months conference review period), workshop papers and presentations spread ideas even faster and have thus become the forefront of many areas.
Okay, these events seem to be useful and cool, but why would we organize them?
There are around 30–50 workshops in each year of machine learning conferences and they spread across diverse topics. Event organization is often also called community service and it takes much effort to put together an appealing program. Nevertheless, it brings the organizers (1) organization experiences, (2) exposure to/networking with other researchers in the field, and (3) the opportunity to share visions, spread ideas, and call for community discussions and actions.
Now, we reach the main theme of this blog: how to organize an appealing event?
There are several key criteria to organize a good event, in my humble and personal opinion, the most important points are based on the answers to three questions:
- What is the purpose of the event?
- What is the program to fulfill the purpose?
- Would the audience be attracted by the mission and program of the event?
Purpose. We organize different events for slightly different reasons. For some events, it has a requirement for the purpose; for others, you need to define the purpose. As an example, a ML conference workshop is supposed to focus on a topic that attracts the “right” amount of audience, (say 1–10%), not too small nor too large. The main idea is to provide this opportunity to engage in smaller community activities. Another example is a regional event (or “local meetup”) instead of a global conference, this may exploit the local structure around the region, e.g. California may have convenient access to industry, and New York may have easier access to financial companies. In the LoG NYC meetup, we were particularly proud to even extend beyond the scope of the conference as the New York area has a unique environment for studying the intersection between ML and structured data such as network science, geometric optimization, and structured probabilistic models.
Program. After the purpose is realized, the organizers need to devise an attractive program that fulfills the purpose. About this part, I feel two parts are the most important:
- Innovation is important as no one would love to attend an event with the same or well-known content. When we attend an event, we are interested in learning about new perspectives. Innovation can come from different aspects, it could be a totally new or trendy topic (imagine the first time you see a large language model (LLM) or foundation model workshop), a new perspective that people have not thought enough on the topic, a revisit to old questions with new ideas, or it even could simply be that no one has never organized an event for a super important topic.
- Diversity is always one indispensable part of event organization. Diversity can come from different perspectives. The common ones include seniority, gender, race, affiliation, and educational background. Diversity is highly related to the inclusion of the event and anticipation of who would like to attend. Imagine if you are submitting a proposal to NeurIPS saying you are going to organize a workshop with all organizers in the same institution, then people may wonder why don’t you just organize this inside your institution? Similar ideas apply to other diversity considerations. Diversity brings us great power to discuss different perspectives and attract diverse attendees which contributes to the event.
Practical implementation. What task does it involve to organize a workshop?
I would personally break this down into two parts (1) initial idea
- What the event would be about and who is the audience? Who is going to be the co-organizers? (different events usually come from different motivations. Sometimes one person starts with some great ideas and involves more people who share similar ideas along the way; sometimes it comes out of a discussion;
and (2) detailed program
- What would the program include? In general, a one-day workshop includes at least invited talks, spotlight talks (from submitted papers), poster sessions, panel discussions, and opening/closing remarks.
- What is the coherent story or idea behind the program? One simplest topic a workshop could focus on is always the frontier of XXX with XXX being the topic such as generative models, machine learning optimization, AI for Science, etc. However, this will only involve simply putting all the “big names” and popular stuff inside the program. But does this really make a workshop attractive? In the AI for Science workshop series, we focus on a different theme every time which would be an interesting discussion and message to the community. I highly recommend event organizers spend much time on the “story” or “takeaway” of the event.
- If you have submissions like the workshops in ML conferences, you need to handle submission platforms such as Microsoft CMT or OpenReview. In general, this does not take too much time even for beginners, they have very nice tutorials and the support team helps timely. For this part, you can also consider the scale of the event, e.g. if you need to break down into topics, with area chairs or even senior area chairs to make decisions.
- You also need to build and maintain a nice website to post relevant information. There are many good and simple website templates and you can simply host the website on GitHub for free.
- You can also create a social media account and Slack channel for attendees to interact with each other.
- Is it simpler than you thought? This is pretty much all the tasks you would need to do for organizing an event. Of course, in the end, you need to prepare some slides as opening/closing remarks in the event day!
First-time events. It takes quite a lot of work to organize a new event as it requires establishing everything from scratch which includes proposal templates, email templates, reviewer lists, etc. Of course, you can borrow from your friends or your favorite events.
Continued events. People may wonder if an event is continued, the organization load becomes much simpler. Indeed, you can reuse many of the materials you have prepared including a proposal template, list of reviewers, email templates, etc. However, especially in ML conferences, continued workshops are not given higher priority than other new workshops. This may come from the innovation part. A continued event is good and nice only when it constantly produces new programs that attract people. Imagine when a new workshop proposal is submitted, they often come up with new ideas (otherwise they would not do it since you are already doing it). So, IMHO, a continued successful event should always consider itself as a new event.
Sponsorship. Fundings are an important part of event organization and can be used to facilitate many things: (1) travel awards (which are also related to diversity), (2) social events (e.g. food/drink/dinner for attendees), (3) reviewer awards (this highly depends on the events, it is worthy, especially for a conference that we tested with LoG, see some analysis), (4) paper awards (often it is a certification), and more. If you are considering raising funds for your events, you should start at least 6 months early or even more if you are organizing a first time event. In our experience, we always face difficulties with the sponsorship part: (1) who to contact — with a contact in the institution who would like to support and push is very important, (2) how much to ask — often the case you need to argue how much you need and where you are going to spend them, (3) how to receive — sponsorship often cannot be made to personal bank account, some possibilities include university account (then you need again someone to help do the work) and non-profit organization, (4) send money to people — this is very hard, e.g. if you are using a university account, you cannot pay any arbitrary people and it needs to go through a proper process with many documents and taxes.
About the author: I am a PhD student with a background in CS/AI/ML. I have been involved in the organization of several continued events and the establishment of new communities, notably the AI for Science workshop series, the probabilistic machine learning workshop series, and the LoG conference. I also assisted in the organization of many other events in different roles. More about me can be found here.